Computer Networks 157 (2019) 41–67

Contents lists available at ScienceDirect

Computer Networks
journal homepage: www.elsevier.com/locate/comnet

Adopting IEEE 802.11 MAC for industrial delay-sensitive wireless

control and monitoring applications: A survey
Yujun Cheng, Dong Yang∗, Huachun Zhou, Hongchao Wang
School of Electronic and Information Engineering, Beijing Jiaotong University, Beijing, 100044, China

a r t i c l e i n f o a b s t r a c t

Article history: In recent years, wireless communication has been widely adopted in the field of industrial systems. Com-
Received 27 August 2018 pared with traditional wired control and monitoring systems, wireless control and monitoring systems
Revised 26 March 2019
are cost-effective and easy to deploy. On the other hand, IEEE 802.11-based wireless technology is widely
Accepted 9 April 2019
applied in many areas due to its popularity, flexibility and ease of management, which makes it a good
Available online 23 April 2019
candidate for various industrial wireless control and monitoring applications with different requirements.
Keywords: However, real-time wireless control and monitoring applications usually have varying degrees of time-
IEEE 802.11 liness requirements, and the basic IEEE 802.11 MAC mechanism cannot support such requirements due
Industrial wireless control and monitoring to its inherent drawbacks. Therefore, in this paper we intend to produce a comprehensive survey and
systems classification of the recent deterministic enhancement approaches in IEEE 802.11 networks, which can
Real-time be applied in wireless control and monitoring systems with different real-time requirements. We explain
MAC
each mechanism briefly, and give an extensive comparison of the features (especially the timeliness level)
of all described MAC mechanisms. Finally, we conclude this paper by identifying some open research is-
sues for future consideration.
© 2019 Elsevier B.V. All rights reserved.

1. Introduction According to the International Society of Automation (ISA), the in-

dustrial systems can be classified into six classes [9,10]. Each class
As one of the most widely deployed wireless technologies, IEEE has different requirement of Quality of Service (QoS), latency, re-
802.11 Wireless local area network (WLAN) technology has experi- liability and security requirements, as shown in Table 1. It is dif-
enced tremendous growth in the last twenty years. In fact, IEEE ficult to provide a standard metric for all use cases, because the
802.11 is a family of progressively enhanced standards, and the latency threshold may vary in specific applications. For example,
IEEE 802.11 Standard Group maintains the development of subse- some control systems (e.g. Laser Cutting) have delay requirements
quent amendments to overcome limitations as they are detected. of hundreds of microseconds. However, the delay requirements for
Due to the popularity, flexibility and ease of management, IEEE general industrial applications often range from a few milliseconds
802.11 techniques can be used in large-scale wireless sensor net- to hundreds of milliseconds [11,12], as listed in Table 1. These ap-
works in various scenarios [1–3], and it becomes a good candidate plications are also known as delay-sensitive applications, where
for various industrial wireless applications with different require- the timeliness is the vital issue to be considered.
ments [4]. Fig. 1 shows an example of a typical scenario with different
In recent years, wireless communication has been widely industrial wireless applications. In the network, different kinds
adopted in the field of control and monitoring systems. Compared of traffic have diverse real-time requirements. The alarm traffic
with traditional wired control and monitoring systems, wireless (emergency traffic) generated by the safety/emergency systems
ones do not require the installation and the regular maintenance should be guaranteed with the highest priority and hard real-time
of expensive communication cables, and therefore they are cost- guarantee. If mishandled, it may incur severe damages to a plant or
effective and easy to deploy. Wireless control and monitoring sys- even threaten a human life. It is rarely triggered when anomalies
tems have been applied in many application domains, including and hazards happen, such as severe electrical surges or risk of ex-
smart homes [5], industrial process control and monitoring [6], plosion etc. The control traffic has much higher priority compared
health-care systems [7] and intelligent transportation systems [8]. to other traffic types except emergency. Failure in communication
may result in the instability of the process control. The monitor-
∗ ing traffic is used to monitor the status of the processes which
Corresponding author.
E-mail address: [email protected] (D. Yang). have relatively less significance. The data collected by monitoring

https://doi.org/10.1016/j.comnet.2019.04.002
1389-1286/© 2019 Elsevier B.V. All rights reserved.
42 Y. Cheng, D. Yang and H. Zhou et al. / Computer Networks 157 (2019) 41–67

Fig. 1. An example of a typical scenario with different industrial wireless applications.

Table 1 marking the start of the frame. This could be a problem in applica-
Classification for industrial systems.
tions requiring highly precise synchronization [18,19]. In MAC layer,
Systems Class Operations Latency(ms) different resource allocation methods can result in distinct latency.
Safety Class 0 Emergency action < 10 For example, in the CSMA/CA based MAC, all stations in a radio
Control Class 1 Closed loop regulatory control < 10 range contend the channel resource. If a transmission experiences
Class 2 Closed loop supervisory control < 100 multiple backoff procedures, it may result in a high probability of
Class 3 Open loop control < 10 0 0 suffering a long MAC delay. In a heavily used channel, CSMA/CA
Monitor Class 4 Alerting < 10 0 0
based MAC performs poorly when the number of stations in a BSS
Class 5 Data < 10 0 0
is too large. The delay will increase dramatically (up to thousands
of miliseconds) due to the large collision probability. It is better
to adopt contention-free medium access (polling-based or TDMA)
to guarantee the real-time performance in this situation. On the
systems, is usually used for the system upgrades and improve-
other hand, when a frame is dropped or lost, the choice of re-
ments in the future. Thus, monitoring traffic has relatively lower
transmission strategy (e.g. retry limit adjustment) will also affect
timeliness and reliability requirements.
the network delay. In this paper, we mainly focus on the MAC de-
Although there are many other existing wireless protocols
lay optimization schemes based on various IEEE 802.11 standards
designed for wireless control and monitoring, such as Blue-
and amendments.
tooth [13] and Zigbee [14] for smart homes, and Wireless HART
The MAC layer fundamentally determines the basic data trans-
[15,16] and ISA100.11a [17] for industrial process control, IEEE
mission capabilities of the network as well as energy consump-
802.11 has several features which are particularly useful to meet
tion properties. Although the performance of the IEEE 802.11
the stringent constraints in the industrial context. For example,
MAC continuously improves as new amendments are published,
it supports both infrastructure and ad hoc mode, and it is able
several inherent defects still exist, such as the uncertain delay
to provide Quality of Service (QoS) differential between different
bound introduced by the channel access mechanism based on
traffic categories. In addition, several high data rates (multi-rate
carrier sense multiple access with collision avoidance (CSMA/CA)
technique), which can be dynamically selected to adapt to channel
called Distributed Coordination Function (DCF). Providing a real-
variations, are available. Nevertheless, many of the currently de-
time guarantee is still a challenging issue among the different
ployed IEEE 802.11 devices are based on legacy IEEE 802.11 (a/b/g)
IEEE 802.11 MAC mechanisms, especially for delay-sensitive ap-
PHY and MAC layers and do not make full use of the recent
plications. Therefore, there is a considerable amount of literature
PHY/MAC enhancements. As a result, the wireless systems may
aiming to fine-tune the IEEE 802.11 standards MAC layer to make
suffer from poor delay performance or low transmission rate prob-
it more appropriate for delay-sensitive applications. On the other
lem.
hand, the new features added by recent PHY/MAC enhancements
The delay sources of WLANs in industrial environment can be
(IEEE 802.11 n/ah) are expected to significantly improve the per-
diverse. In physical layer, for example, the multipath problems,
formance of wireless control systems, so that making a thorough
which caused by reflection, diffraction and scattering of electro-
investigation of these new features is also very essential.
magnetic fields, may introduce latency to the wireless network.
There already exists various surveys on the IEEE 802.11 MAC
Multipath propagation not only affects the delay by increasing bit
and its adjustment [20,21]. However, none of the previous sur-
errors, but it may also introduce delay variations because the re-
veys treats the timeliness issues in the wireless control systems
ceiver cannot be sure which of the received replicas is the one
 Y. Cheng, D. Yang and H. Zhou et al. / Computer Networks 157 (2019) 41–67 43

based on IEEE 802.11. In this survey paper, we restrict our atten-

tion to the IEEE 802.11 MAC enhancements and adjustments from
a perspective of delay-sensitive applications. The remainder of this
paper is organized as follows. Sections 2 and 3 review the main
MAC features of the IEEE 802.11 protocol family. Section 4 presents
a detailed review of the optimization techniques for the indus-
trial delay-sensitive applications in general IEEE 802.11 MAC (a/b/g
and e). Section 5 discusses the modification schemes in latest and
upcoming IEEE 802.11 MAC (n/ac/ax/ah and aa) for real-time en-
hancement. We enumerate open issues in Section 6, and finally,
the conclusions are listed in Section 7.

2. Conventional IEEE 802.11 (a/b/g and e) MAC review

In 1997, the original version of the IEEE 802.11 standard was

Fig. 2. EDCA traffic priorities mapping. High-priority traffic has a higher transmis-
released, which provides data transmission rates up to 1 Mbps sion chance than low-priority traffic.
and 2 Mbps in 2.4 GHz frequency band. In 1999, IEEE 802.11a and
IEEE 802.11b standards were published in succession. IEEE 802.11b Table 2
was conceived to support operation in the unlicensed 2.4 GHz in- IEEE 802.11E EDCA parameter set.
strumentation, scientific and medical (ISM) band with maximum AC CWmin CWmax AIFSN TAIFS TXOP
transmission rate of 11 Mbit/s, while IEEE 802.11a supports data
AC_VO 3 7 2 28 μs 1504 μs
rates up to 54 Mbps with frequency band 5 GHz. In 2003, IEEE AC_VI 7 15 2 28 μs 3008 μs
802.11g was approved with the maximum data rate of 54 Mbps in AC_BE 15 1023 3 37 μs 0
2.4 GHz frequency band. The fundamental IEEE 802.11 MAC tech- AC_BK 15 1023 7 73 μs 0
niques to share the medium between multiple stations are Dis-
tributed Coordination Function (DCF) and Point Coordination Func-
tion (PCF). However, they cannot provide real-time Quality of Ser- Once CW reaches CWmax , the contention window is maintained at
vice (QoS) guarantees since there is no differentiation of high or CWmax for the following transmission process. The CW is set to the
low priority traffic. Thus, it is unable to support quite a number CWmin every time after a successful data transmission or when the
of applications which have real-time requirements, such as multi- retransmission counter exceeds the retry limit. When the medium
media and industrial communications. To address this issue, IEEE is sensed busy, the back-off procedure is frozen, and it resumes
802.11e amendment was standardized. It enhances the DCF and decreasing once the medium is sensed idle again.
PCF through a new coordination function called Hybrid Coordina- While DCF has a good performance on general networks with-
tion Function (HCF). It defines Traffic Categories (TC), adopting traf- out any real-time requirements, it is hard to achieve predictable
fic prioritization to meet the requirement of real-time applications, behavior in systems with real-time constraints, especially for in-
including some industrial applications. Although using some novel dustrial applications. DCF experiences a nondeterministic back-off
schemes in the upper layer (e.g. traffic engineering) may improve delay due to the BEB algorithm. It may take a very long period
the real-time performance, we merely focus on the MAC layer re- before a successful transmission under high traffic loads and high
lated mechanisms in this paper. In the following, we mainly review station density situation. The fundamental cause of the poor per-
the basic 802.11 MAC features mentioned above. formance is that general (non-real-time) traffic may use bandwidth
from the real-time traffic so that QoS requirements of real-time
2.1. Distributed Coordination Function traffic are unable to be satisfied. Thus, in order to improve the
QoS for real-time applications, IEEE 802.11e task group enhances
The Distributed Coordination Function (DCF) is a contention- the DCF through a new channel access mechanism: Enhanced Dis-
based MAC mechanism adopted by IEEE 802.11 networks. It utilizes tributed Channel Access (EDCA), which is able to provide QoS for
Carrier Sense Multiple Access mechanism with Collision Avoid- traffic with real-time requirements.
ance (CSMA/CA) and Binary Exponential Back-off (BEB) algorithm
to share the medium. There are two separate and distinct carrier- 2.2. Enhanced Distributed Channel Access
sensing functions defined in IEEE 802.11 standard: Clear Channel
Assessment (CCA) and the Network Allocation Vector (NAV). CCA To provide real-time QoS, IEEE 802.11e Enhanced Distributed
is physical carrier sense which is based on energy thresholds from Channel Access (EDCA) introduces a priority-based differentiated
the radio interface. NAV is virtual carrier sense which is set up by medium access mechanism. Four traffic categories, which are
overhearing stations in order to avoid potential conflicts. also known as access categories (ACs), are defined in EDCA:
Before starting a new transmission, every station must sense voice (AC _V O), video (AC _V I), best-effort (AC _BE), and background
the status of the wireless medium. If it finds that the medium is (AC _BK) traffic, respectively (Fig. 2). With EDCA, high-priority traf-
continuously idle for an additional random back-off period plus a fic has a higher transmission chance than low-priority traffic by
DCF Interframe Space (DIFS) duration, then it is permitted to ini- differentiating the parameters of deferral and back-off for different
tiate its transmission. Otherwise, the transmission process should ACs. The default EDCA parameter settings are shown in Table 2.
be frozen until the medium is idle again. The duration of back-off To provide prioritization for different ACs, in EDCA, higher pri-
is composed of a multiple of a slot time. Each transmitting sta- ority traffic adopts shorter arbitration inter-frame space (AIFS),
tion generates a uniformly random back-off value from [0, CW − 1], which behaves similarly to DIFS in DCF. The size of contention win-
where CW stands for the contention window size. The random dow (CW) is set according to the expected traffic load in each AC.
back-off value is the number of slot times a station has to wait Besides, EDCA provides Transmit Opportunity (TXOP) mechanism.
before transmission. For the first transmission, CW is set to CWmin , A TXOP is a bounded time interval during which a station may
which is the predefined minimum contention window. It is dou- send as many frames as possible. When an internal traffic collision
bled after each unsuccessful transmission until it reaches CWmax . happens, the higher priority access category obtains the chance of
44 Y. Cheng, D. Yang and H. Zhou et al. / Computer Networks 157 (2019) 41–67

Fig. 3. A PCF superframe. Polling-based mechanism is employed in the CFP while a contention based mechanism is adopted in the CP.

data transmission, while the other ACs restart the back-off proce- and minimum PHY rate. Once the TSPEC negotiation is successful,
dures. With these features of EDCA, the QoS of industrial real-time HC adds the QSTA in the polling list, and assigns Transmission Op-
traffic can be improved when the traffic is defined as an AC with portunity (TXOP) to the QSTA.
a high priority. To a certain extent, EDCA is a good approach to With the QoS improvement mechanisms in the HCCA, real-time
provide adequate QoS guarantee for some industrial applications QoS can be configured with great precision compared with the PCF.
which do not have stringent real-time requirements. However, these two centralized channel access schemes still have
problems when applied to the industrial applications. First of all,
2.3. Point Coordination Function and HCF-Controlled Channel Access they lack inter-AP coordination. If the stations is polled by the
APs in two overlapping basic service sets (BSSs) simultaneously,
Aiming to improve the deterministic performance of IEEE 802.11 collision may occur, which could cause considerable performance
networks, the Point Coordination Function (PCF) is defined in orig- degradation. Besides, scalability is a big problem for the two cen-
inal 802.11 MAC. Its enhanced version, HCF-Controlled Channel tralized schemes if they are applied to a large-scale of network.
Access (HCCA) are also defined in 802.11e amendment. Different In addition, both of two schemes require a complicated software
from DCF and EDCA, these two schemes utilize a Point Coordina- architecture, especially HCCA, and therefore, few existing network
tor (PC)/Hybird Coordinator (HC), which usually reside in the AP, to equipments support PCF and HCCA, which prevents them from
provide contention-free transmission by a polling-based resource practical application.
allocation method.
In PCF, time is divided into superframes. A superframe in- 3. Latest and upcoming IEEE 802.11(n/ac/ax/ah and aa/ae) MAC
cludes a contention period (CP) and a contention-free period (CFP). features
Polling-based mechanism is employed in the CFP while a con-
tention based mechanism is adopted in the CP. In order to pro- Facing the demand for WLANs with higher data rates and bet-
vide PC higher priority in medium access than DCF stations, the ter QoS performance, in 2009, IEEE 802.11 working group led to
point interframe space (PIFS) is adopted in PCF, which is longer a further amendment, named 802.11n. It provides up to 600 Mbps
than a short interframe space (SIFS) but shorter than DIFS. In the data rate by using multiple-input multiple-output (MIMO) technol-
CFP, the AP sends Contention-Free-Poll (CF-Poll) packets to each ogy. To provide QoS enhancements for real-time applications, IEEE
station, one at a time. Nodes can only transmit their data after 802.11aa and IEEE 802.11ae served as two amendments to the IEEE
receiving the CF-Poll packet. If the PC receives no acknowledge- 802.11-2012 revised version. In 2014, IEEE 802.11ac was approved,
ment (ACK) from a polled station after a PIFS period, it will poll which is support up to 7Gbps data rate. In 2017, IEEE 802.11ah was
the next station. If there is no pending data in a polled station, it published which defines a WLAN system operating at sub 1 GHz
responses a null frame containing no payload to PC. Fig. 3 shows license-exempt bands. It can provide much longer transmission
the contention-free transmission procedure in CFP. range compared with the conventional 802.11 WLANs operating at
However, as for supporting real-time QoS, PCF still has some 2.4 GHz and 5 GHz bands. IEEE 802.11ah aims to address the partic-
problems. PCF does not provide prioritized differentiation of differ- ular requirements of Machine to Machine (M2M) communication,
ent traffic, thus it is unable to fulfill the deadline requirements in including industrial communication. The upcoming IEEE 802.11ax
industrial applications. In addition, once a station gets access to the will replace both IEEE 802.11n and IEEE 802.11ac as the next high-
channel, the occupied time interval of the polled nodes cannot be throughput WLAN amendment, which is due to be publicly re-
predicted, which results in nondeterministic issue and traffic un- leased sometime in 2019. In the following, we mainly examine the
fairness for industrial applications. new MAC features of these latest and upcoming WLAN networks.
Thus, HCCA is proposed to address these problems. HCCA works
a lot like PCF, however, different from PCF, the superframe in HCCA 3.1. IEEE 802.11n MAC enhancements
includes a Controlled Access Phase (CAP) where HCCA is used, and
a CP, where EDCA is used (Fig. 4). In contrast to the CFP in PCF, In the 802.11n amendment, the new MAC layer builds strongly
the CAP is allowed to be initiated at almost any time during a CP, on the former IEEE 802.11e foundations, which introduced QoS
which is more flexible for data transmission. Besides, HCCA utilizes concepts and defined TXOP. Three main MAC enhancements are
Traffic Specification (TSPEC) to characterize the traffic streams and adopted in IEEE 802.11n, including Frame Aggregation, Block Ac-
its QoS requirements. In fact, the concept of Traffic Stream (TS) is knowledgement (BA) and Reverse Direction (RD). Besides, Direct
introduced in IEEE 802.11e to describe a set of data units (MSDU) Link Setup (DLS), the optional protocol defined in IEEE 802.11e, will
that has to be delivered conforming to a corresponding TSPEC. In also be introduced in this subsection.
order to be included in the polling list of the Hybrid Coordinator
(HC), a QoS station (QSTA) must send a QoS reservation request us- 3.1.1. Frame aggregation
ing the special QoS management frame which carries traffic spec- Although the data rate of IEEE 802.11n PHY has greatly im-
ification (TSPEC) parameters, and each TS needs one unique reser- proved compared with legacy IEEE 802.11, the MAC layer overhead
vation request. The major TSPEC parameters include Mean data which is coupled with MAC header can cause significant through-
rate, delay bound, maximum service interval, nominal MSDU size put decrease. Frame aggregation (Fig. 5) allows to aggregate more
 Y. Cheng, D. Yang and H. Zhou et al. / Computer Networks 157 (2019) 41–67 45

Fig. 4. HCCA service. A superframe includes a contention period (CP) and a contention-free period (CFP). Polling-based mechanism is employed in the CFP.

3.1.3. Reverse Direction (RD)

The purpose of RD protocol is to more efficiently transfer data
between two 802.11 devices during a TXOP in bi-directional net-
work traffic services like Voice over IP (VoIP), video conference,
online gaming etc. With RD, once the transmitting station has ob-
tained a TXOP, it may essentially grant permission to the corre-
sponding station to send information back during its TXOP without
a new contention process for transmission medium. In RD protocol,
two kinds of wireless stations exist in network, namely RD initiator
and RD responder. The RD initiator sends its permission to the RD
responder using a Reverse Direction Grant (RDG) in the RDG field
of the HT Control field in the MAC frame. Once the RD responder
receives it, it responds with an RDG acknowledgement (ACK) if it
has data to transmit. If RD initiator receives ACK successfully, then
it waits for the data transmission from RD responder after a Short
Interframe Space (SIFS). RD enhances the channel utilization by al-
Fig. 5. IEEE 802.11n two-level aggregation.
locating the left time of TXOP to the RD responder, and it achieves
delay reduction in the corresponding link traffic.

frames into a single one to be transmitted, thus reducing the time 3.1.4. Direct Link Setup (DLS)
of transmitting overheads and the pending time caused by ran- The DLS protocol allows a direct frame exchange between adja-
dom backoff period during successive frame transmissions. It is cent devices within a basic service set. This mechanism is designed
helpful to further enhance channel utilization and efficiency for primarily for consumer use, where more and more devices need
a wireless station. Two aggregation schemes are adopted in IEEE to exchange local traffic. For example, the video streams generated
802.11n, namely Aggregated MAC Service Data Unit (A-MSDU) and by a player must be relayed through AP to reach the correspond-
Aggregated MAC Protocol Data Unit (A-MPDU). Multiple Logical ing display device in the AP-based WLAN. It can be more efficient
Link Control (LLC) packets called MAC Service Data Units (MSDUs), to send Wi-Fi frames directly from the video player to the display
which are to or from the upper layers, can be aggregated into a device. Moreover, if the location of the AP is far away from the de-
single A-MSDU. The aggregated MSDUs must have the same desti- vices, a direct communication may bring a benefit of a higher-rate
nation, and the maximum size of an A-MSDU is 7935 bytes or con- MCS due to a shorter distance.
taining five 1500 byte frames. After the addition of MAC header to
each MSDU, they form MAC Protocol Data Units (MPDU). MPDUs 3.2. IEEE 802.11ac MAC enhancements
can be aggregated into an A-MPDU only if they all have the same
destination, and the maximum size of an A-MPDU is 65,535 bytes IEEE 802.11ac is a successor to 802.11n in 5 GHz channel
or 64 MPDUs. frequency. It improves the maximum throughput primarily by the
following approaches: the use of 80/160MHz channels, 256QAM,
up to eight antennas and downlink multi-user MIMO. As the data
3.1.2. Block acknowledgement (BA) rates of IEEE 802.11ac are much higher, an enhanced frame ag-
The BA mechanism in IEEE 802.11n is an improved version of gregation mechanism is adopted to achieve very high throughput.
BA in IEEE 802.11e, which allowed the receiver to acknowledge the Besides, an enhanced TXOP called TXOP sharing mechanism is
transmission of multiple data units with a single frame. In IEEE utilized to support multiple downlink traffic streams to multiple
802.11n, BA mechanism is applied to aggregated frames, and all receiver stations simultaneously.
the frames transmitted during TXOP period are acknowledged by
a BA instead of an ACK frame for each frame transmitted. The BA 3.2.1. Enhanced frame aggregation
contains the acknowledgements of correct MPDUs, and the sender In IEEE 802.11n, there are two kinds of frame aggregation,
only needs to retransmit the MPDUs which are not acknowledged, which are called A-MSDU and A-MPDU. The IEEE 802.11ac im-
which can bring further channel utilization improvement. proves the frame aggregation mechanism by extending the size
46 Y. Cheng, D. Yang and H. Zhou et al. / Computer Networks 157 (2019) 41–67

Fig. 6. Illustration of multi-user EDCA and TXOP sharing. In this figure, AC_VI is assumed as the Primary AC, and AC_VO and AC_BE are Secondary ACs. Among the Secondary
ACs, traffic with higher priority gets transmitted earlier.

Fig. 7. Multiple frame transmission in a shared TXOP.

of A-MSDU and A-MPDU. The maximum size of A-MSDU has primary AC, and the transmission time is determined by the du-
been increased to 11,454 bytes, while the maximum size of a ration of the amount of data scheduled to be transmitted by the
single A-MPDU is limited to 1 Mbyte. Besides, compared with primary AC. The TXOP period is finished once the primary AC fin-
the aggregation scheme before, all frames transmitted must use ishes its data transmission, even if secondary AC has more frames
the aggregate MPDU (A-MPDU) format in IEEE 802.11ac. Even a to transmit.
single frame transmitted in one shot should be transmitted as an
aggregate frame. Thus, the length indication in the PLCP header 3.3. IEEE 802.11aa MAC enhancements
can be transmitted as part of high-data-rate payload instead of
transmitted at the lowest possible data rate. The 802.11aa specification is designed to support efficient
multicast transmission and robust video transport. It implements
3.2.2. TXOP sharing a set of mechanisms to provide QoS enhancements for real-time
IEEE 802.11ac enhances the MAC layer by extending the existing applications, including Group Cast with Retries (GCR), Stream
transmit opportunity by proposing a new technique called TXOP Classification Service (SCS) and Overlapping Basic Service Set
sharing (Figs. 6 and 7). It corresponds to the new PHY downlink (OBSS) management. The GCR improves the reliability of multicast
multi-user MIMO technique, which can support multiple downlink transmission. The SCS provides a finer granularity for the traffic
data streams to multiple receiver stations simultaneously. Each AC differentiation, and the real-time performance of high priority traf-
of an AP competes for a TXOP using its own EDCA parameters as fic will increase. The OBSS management solves the problem of the
it does in the current standard, once an AC wins a TXOP, it is con- interference of adjacent BSSs. The adjacent BSSs can cooperatively
sidered as the primary AC, while other ACs become secondary ACs. share of the medium thus improving the network performance.
The primary AC decides whether to share its TXOP with the sec- These mechanisms are described briefly below.
ondary ACs for simultaneous transmissions. If it does, the obtained
TXOP becomes a multiuser TXOP (MU-TXOP). The primary AC de- 3.3.1. Group Cast with Retries (GCR)
cides which secondary AC is permitted to share with the obtained In order to overcome the unreliability of 802.11 WLAN caused
TXOP, and which destinations to target for transmissions. Moreover, by the lack of acknowledgement frames, the 802.11aa standard de-
the TXOP duration should be determined by the TXOP Limit of the fines new Acknowledgement (ACK) policy, namely GCR Unsolicited
 Y. Cheng, D. Yang and H. Zhou et al. / Computer Networks 157 (2019) 41–67 47

Fig. 8. Stream Classification Service over IEEE 802.11aa. Two additional queues (AAC_VO, AAC_VI) are included within the existing EDCA access categories.

Retry (GCR-UR), Direct Multicast Service (DMS) and GCR Block ACK. admission control mechanism in a distributed environment; man-
In the GCR-UR scheme, multicast frames are sent several times agement of scheduled transmission opportunities between OBSSs.
without waiting for an acknowledgment after each transmission.
The number of retry limit should not exceed the predefined trans- 3.4. IEEE 802.11ae MAC enhancements
mitter threshold. The GCR-UR scheme improves the reliability by
increasing the delivery probability at the cost of increasing the With the increasing number of IEEE 802.11 amendments, the
frame overhead. However, it is useless when the channel quality category of management frame has significantly increased. Though
is good. In the DMS scheme, the multicast streams are converted these management frames often have limited size, they can still
into unicast. The converted frames will be retransmitted until re- degrade the performance of voice applications if they are fre-
ceiving an ACK from the AP, and they will be stopped as soon as quently transmitted with the highest priority as in 802.11. Thus,
the retry limit is reached. This feature provides a reliable trans- the IEEE 802.11ae amendment provides a solution to this problem
port to multicast traffic similar to that of unicast. The GCR Block called QoS management frame (QMF) service. QMF is a mechanism
ACK scheme is an extension of the Block ACK defined in 802.11e for the flexible prioritization of management frames. Instead of al-
and 802.11n standards. In this mechanism, after a station sends a ways sending management frames with the highest priority, it pro-
burst of data frames, the originator will unicast Block ACK Request vides a mapping between the management frame types/subtypes
(BAR) frame to all multicast users. Then these multicast users pro- and the EDCA ACs, and all management frames should be trans-
vide their feedback of receiving status in Block ACK (BA). mitted in an AC defined by the QMF policy (Fig. 9). Thus, QMF
enhances the real-time performance of the highest priority data
frames (voice), because there is less contention in the queue with
3.3.2. Stream Classification Service (SCS)
highest priority. Moreover, the QMF policies are flexible in general,
SCS is a new prioritization mechanism defined in IEEE 802.11aa.
and they can be established and changed by adopting a signaling
It extends the traffic differentiation concept of 802.11e EDCA access
protocol defined in 802.11ae. Diverse requirements of vendor ap-
mechanism to increase the granularity of traffic prioritization and
plications can be satisfied under this flexibility.
QoS support. While EDCA provides traffic differentiation by dis-
tinct contention parameters of each access category, such as AIFS,
3.5. IEEE 802.11ax MAC enhancements
CW and TXOP durations, IEEE 802.11aa suggests an Intra-AC Pri-
oritization to provide the ability to differentiate among individual
IEEE 802.11ax is a new standard being developed by the IEEE
streams. Two additional queues, named Alternative Access Cate-
802.11 Working Group, which will develop new physical (PHY) and
gory for Voice (AAC_VO) and Alternative Access Category for Video
MAC layer enhancements to further improve the user experience.
(AAC_VI), are included within the existing EDCA access categories
IEEE 802.11ax contains a new PHY protocol with higher modula-
to provide prioritization within both AV streams (shown in Fig. 8).
tion and coding schemes, the nominal data rates are increased up
In addition, packets are assigned to an appropriate AC by tagging
to 9.6 Gbps. Besides, IEEE 802.11ax employs OFDMA approach, an
with a Drop Eligible Indicator (DEI) bit. The DEI bit identifies a dif-
approach widely used in cellular networks, but brand new in Wi-
ferent maximum number of (short or long) retries, which deter-
Fi. With OFDMA scheme, different subsets of subcarriers are as-
mines if the packet is eligible to drop when the receiver does not
signed to different users, which allows simultaneous access to ra-
have sufficient resources.
dio resource. Apart from the PHY improvement, IEEE 802.11ax pro-
vides several crucial enhancements also at the MAC layer. The main
3.3.3. OBSS management mechanisms are shown in detail below.
The 802.11aa standard defines mechanisms to facilitate cooper-
ative sharing of the medium between APs operating in the same 3.5.1. Improving Spatial Reuse
channel, namely Overlapping BSS (OBSS). In this scheme, differ- The dense deployment is one of main scenarios for IEEE 802.11
ent APs are generally coordinated and therefore do not damage the Task Group ax (TGax). To improve the performance in the overlap-
QoS of multimedia applications even though there are interfering ping basic service set (OBSS) situation, several spatial reuse (SR)
systems. OBSS management provides the following capabilities: ad- enhancement schemes are introduced in an 802.11ax system, in-
ditional information for smart channel selection; extension of the cluding dynamic adaptation of the transmit power and CCA levels,
48 Y. Cheng, D. Yang and H. Zhou et al. / Computer Networks 157 (2019) 41–67

Fig. 9. The operation of QMF scheme. The scheme provides flexible prioritization of management frames.

BSS color and multiple NAVs. By dynamic adaptation of the trans- 802.11ac PHY layer, and the IEEE 802.11ah MAC layer extends most
mit power and CCA levels with SR, the IEEE 802.11ax network can of the main IEEE 802.11 features, adding some novel mechanisms.
get an optimal trade-off between individual transmission rates and In order to support a larger number associated STAs to a sin-
the number of simultaneous transmissions that maximize the area gle AP, IEEE 802.11ah designs a new 13-bit Association Identifier
throughput. The CCA modification scheme is called dynamic sen- (AID) to each STA, which is a unique number used by the AP to
sitivity control (DSC) algorithm, which aims to optimize the ex- communicate with STAs. Accordingly, the number of stations that
isting deployments by adjusting the carrier sense threshold (CST) it can express is up to 213 − 1 (= 8191). AID hierarchically clas-
properly for each station in a distributed manner. BSS color is an sifies stations into pages, blocks, sub-block, and station’s index in
innovative scheme to improve the performance of throughput in sub-block (Fig. 11). The characteristic of hierarchical AID structure
dense WLAN networks. The BSS color is a non-unique ID assigned makes grouping of the stations with same traffic patterns or same
for each BSS, which is transmitted in the frame preamble. When a locations much easier. An AP can simply use the block ID or the
station receives frames from a neighboring BSS, it can stop decod- sub-block ID to indicate group of stations instead of including all
ing the frames, assuming the channel is still idle, thus increasing of their AIDs.
transmission opportunities. The proposed multiple NAVs scheme Another important feature is the channel access scheme in IEEE
utilizes two NAV timers (intra-BSS NAV and regular NAV) at each 802.11ah. IEEE 802.11ah defines three different types of stations,
station. The intra-BSS NAV is increased or reset only by the frames namely, traffic indication map (TIM) stations, non-TIM stations, and
from the intra BSS. Thus, the performance of spatial reuse can be unscheduled stations. Each of them has different procedures and
improved by allowing the station to ignore RTS/CTS frames from time periods to access the common channel (Fig. 10). Among of
the alien BSS. them, TIM stations have to listen to DTIM and TIM beacons, and
it can only transmit data within a restricted access window (RAW)
period. RAW is a new type of contention-free channel access mech-
3.5.2. OFDMA random access anism. It is a time duration which is composed of three segments,
In Uplink OFDMA (UL OFDMA), TGax has designed an optional including a downlink segment, an uplink segment, and a multicast
scheme which allows random UL OFDMA transmissions. The fea- segment. It further splits the stations into groups, and only allows
ture is especially important when the AP expects that there are stations in a certain group to access the channel at certain times
some STAs having data to transmit, but does not know the spe- to restrict the number of stations that can simultaneously access
cific STAs. In the designed random access, a trigger frame can as- the channel. RAW can be used for various purposes. The QoS and
sign some resource units (RUs) for random access. A STA maintains energy efficiency are highly guaranteed in RAW. In contrast of TIM
an OFDMA backoff (OBO) counter to decide whether to transmit stations, Non-TIM stations do not have to listen to any beacons to
and in which RU. The OBO counter is different from an EDCA back- trigger data transmission. In fact, non-TIM devices directly negoti-
off counter, which is only for OFDMA-based random access. When ate with the AP to get transmission chance in a periodic restricted
the station receives a trigger frame with random access, it should access window (PRAW). During PRAW, Non-TIM stations are al-
decrease OBO by the number of RUs assigned for random access. lowed to request buffered downlink traffic or to transmit uplink
When its OBO counter reaches 0 or below, the STA could randomly traffic at any time upon waking up, thus contention among these
choose one of the random-access RUs to transmit. Readers can re- stations might result in excessive channel access delays or even
fer to [22] for a more detailed describe of the OFDMA random ac- collisions. The unscheduled stations do not have to listen any bea-
cess procedure. cons either. However, compared with Non-TIM stations, they can
send a poll frame to the AP asking for immediate access to the
3.6. IEEE 802.11ah MAC enhancements channel in both RAW and PRAW. This procedure targets stations
which just intend to join the network sporadically.
Compared with the great advances that 802.11ac and 802.11ax
will bring in terms of high speed data rate, IEEE 802.11ah amend- 4. Conventional IEEE 802.11 MAC for delay-sensitive application
ment is for the purpose of M2M communication at frequency
bands below 1 GHz. It intends addressing the particular require- In this section, we provide a novel taxonomic classification
ments of M2M networks: a large number of power-constrained sta- (Fig. 12) of the various design and optimization techniques for the
tions; a long transmission range; low data rates; small and infre- industrial delay-sensitive applications in general IEEE 802.11 MAC.
quent data messages; and a non-critical delay. The PHY layer of A large amount of literature has shown that standard
IEEE 802.11ah can be considered as a sub-1 GHz version of the IEEE DCF mechanism cannot meet the requirements of industrial
 Y. Cheng, D. Yang and H. Zhou et al. / Computer Networks 157 (2019) 41–67 49

Fig. 10. IEEE 802.11ah channel access restricted windows (RAW and PRAW) among signaling beacons.

Fig. 11. IEEE 802.11ah AID format.

Fig. 12. A classification of the real-time enhancements of IEEE 802.11 MAC for wireless control systems. The real-time optimizations based on conventional WLANs are
discussed in Section 4, while Section 5 presents the real-time enhancements in the latest and upcoming WLANs.

delay-sensitive applications because the contention-based access dancy and rate adoption are also widely adopted in the literature
method cannot provide a guaranteed delay-bound. To address this to boost the system deterministic performance. In the following
problem of the core MAC mechanism, two kinds of mainstream subsections, each kind of mechanism is described in more detail
approaches are put forward. One is still based on the contention firstly, and it is followed by descriptions of all the specific work
access mechanism but with several modifications to improve the which utilizes the given mechanism. At the end of each subsec-
transmission determinism. This kind of approach tunes the basic tion, an overview of the merits and demerits of the mechanism
DCF parameters such as contention window and inter frame spaces are provided.
(IFS), and generally makes use of prioritization to offer higher level
QoS parameters than the common best effort traffic. The other 4.1. Fine-tuning on DCF mechanism
group of approaches tries to build a contention-free MAC based
on the standard DCF. Without the random delay introduced by the The Binary Exponential Backoff (BEB) of IEEE 802.11 standard is
contention phase, the delay-sensitive sessions can obtain a guar- widely adopted due to its ease of implementation and good perfor-
anteed delay-bound through a proper resource allocation method. mance under the limited workload. However, it suffers severely de-
This kind of approaches consist of polling, time division multiple ac- graded performance owing to collision with the increasing network
cess (TDMA) and token passing scheme. In addition, other MAC op- load. Adopting a dynamic contention window size [23] (Fig. 13(a))
timization techniques such as retransmission management, redun- may provide a substantial performance improvement under the
50 Y. Cheng, D. Yang and H. Zhou et al. / Computer Networks 157 (2019) 41–67

Fig. 13. The basic BEB and EDCA tuning schemes.

heavy loaded WLAN environments. Based on the standard DCF, an- algorithm to dynamically adjust the CW size of each competing
other kind of approach can achieve semi-determinism by using sev- stations according to the turn-around-time measurement of chan-
eral collision avoidance schemes. In addition, four ACs with priority nel status. The scheme can provide a substantial performance im-
transmission queues are defined in 802.11e, and different CW pa- provement under heavy loaded and error-prone WLAN environ-
rameters are set to each queue to ensure QoS for the high priority ments.
real-time traffic (Fig. 13(b)). Improper configuration of backoff pa- In [27], two adaptive CW tuning algorithms are implemented,
rameters could introduce more internal collisions among the real- namely the Centralized Adaptive Control (CAC) and the Distributed
time traffic in the ACs, and adjusting the backoff parameters in EDCA Adaptive Control (DAC). A Proportional Integrator (PI) controller is
is one solution to minimize the collision problem. set to dynamically tune the CW configuration. The dynamical CW
configuration scheme is based on the collision probability that is
4.1.1. Dynamic contention window tuning computed and then analyzed using control theory. The two algo-
In [24], the authors propose a new backoff mechanism, called rithms are implemented with unmodified commercial off-the-shelf
Smart Exponential-Threshold-Linear (SETL) Backoff Mechanism. devices, which confirms its suitability for real deployments.
The core idea is to set a threshold to determine the behavior of In [28], a deadline-constrained MAC protocol with QoS differen-
CW after every transmission. The CW of each station is exponen- tiation is presented for IEEE 802.11 soft real-time network control
tial when the CW is smaller than the threshold, which is similar systems (NCSs). It provides a contention-sensitive backoff mech-
to the CW behavior in the standard BEB. However, if the CW size anism and an intra-traffic-class QoS differentiation mechanism to
is beyond the threshold, it is tuned linearly to prevent large trans- handle the real-time traffic. The standard 802.11 binary exponen-
mission delay caused by long backoffs. tial backoff (BEB) doubles the contention window size W following
In [25], the authors propose an adaptive backoff parameters a collision only, and keeps the W the same but freezes the backoff
mechanism based on the estimation of the number of competing timer when the channel is sensed busy. When the network traffic
stations. The estimation algorithm uses a Bayesian approach and load increases, a larger initial backoff window size is desirable and
sequential Monte Carlo methods, aiming to optimize the backoff it may introduce transmission delay. In the contention-sensitive
parameters based on the predictive distribution of the number of BEB algorithm, W is doubled not only in the case of collisions but
competing stations. also when the channel is busy. It can reduce the probability of col-
In [26], the authors present a backoff algorithm based on self- lision under heavy traffic load.
adaptive contention window update factor. The update factor, c, de-
fined in this paper is an implicit function of the number of the 4.1.2. Semi-deterministic schemes based on CSMA
competing stations. The decrease and increase sizes of CW are de- Several other backoff-tuning schemes intend to achieve re-
termined by c when the current transmission is on a success or source reservation to provide semi-deterministic data transmission.
collision. The paper then provides a set of the optimal update fac- These can also be called collision resolution schemes based on
tor values in different situation through an analytical model and CSMA. In [29], the authors propose CSMA with enhanced colli-
simulation. sion avoidance (CSMA/ECA) (Fig. 14). The scheme adopts a new
In [23], the authors identify the relationship between backoff backoff scheme called learning BEB (L-BEB), which behaves as the
parameters, contention level, and channel BER. Based on the obser- standard CSMA/CA protocol with one exception, that is, a selec-
vation and theoretical analysis, the authors propose a distributed tion of a deterministic backoff value is determined after successful
 Y. Cheng, D. Yang and H. Zhou et al. / Computer Networks 157 (2019) 41–67 51

Fig. 14. A typical design of semi-deterministic schemes (L-BEB [29], SRB [35]). This scheme uses a fixed or reselected random backoff values to achieve collision-avoidance.

transmissions. As for the failed transmission due to collisions or because the centralized allocation method can operate in a col-
frame errors, L-BEB moves back to the standard random backoff lision free state immediately without dynamic parameter adjust-
procedure. L-BEB uses a fixed or reselected random backoff val- ment.
ues to achieve collision-avoidance, and it is easy to implement The Semi-Distributed Backoff (SDB) [42] mechanism aims to
on existing platforms. The work is followed by analytical studies avoid successive collision in any loss-sensitive application. The key
[30–32] and several enhancements [33,34]. Semi-Random Backoff idea of SDB is that if a station encounters collision and performs
(SRB) [35] is similar to the L-BEB as they both a fixed backoff the sender-side backoff procedures in standard BEB, a receiver-
value to achieve collision-avoidance. However, they use different side backoff should also be executed. The information of optimal
analytic models, verification approaches and evaluation procedures. backoffs is disseminated to the contenders by adopting the colli-
Another similar work is CF-MAC [36] protocol, in which the imple- sion detection capability of receivers. Based on SDB, a novel MAC
mentation is done through more precise firmware instructions that named Semi-DCF is designed, which performs opportunistic migra-
force the stations to transmit exactly when scheduled. tion from random to deterministic backoff.
Zero-collision (ZC) [37] works based on empty and busy slots. In [43], a Reservation-Based Back-Off mechanism (ReB) is pro-
Similarly to L-BEB, it chooses a deterministic backoff after a suc- posed. It employs a fixed back-off cycle which includes one or mul-
cessful transmission. However, the procedure after a collision is not tiple time slots for the competing stations. The basic idea of ReB is
like in the standard random backoff procedure. Instead, the sta- borrowed from reservation ALOHA, but it introduces several addi-
tions observe the utilization of each slot in the last backoff period, tional mechanisms, such as a fixed backoff cycle size, random se-
and then choose their new backoff values uniformly between the lection of transmission timing and multiple transmission opportu-
slot it failed and the slots that were idle in last period. ZC avoids nities within one backoff cycle to improve the network efficiency
other busy slots that other stations may have reserved thus the in audio/video transmission.
converging to collision-free state is faster than L-BEB. The authors in [44] proposed E-MAC that uses a novel elaborate
The authors of [38] proposed modified versions of L-BEB and backoff mechanism to iteratively accomplish collision-free trans-
ZC named L-MAC and L-ZC. The Adopting learning algorithms sig- mission. In E-MAC, each station transmits at most once in a given
nificantly improve the convergence time of each original collision- time cycle, and if the collision happens, the station will choose a
free approach. L-MAC only uses the indication of transmission state more proper transmission time in the next cycle based on the col-
(successful or failed) as the learning parameters, and L-ZC adopts lision part of the collided packet.
additional information on each slot on the medium. These al- I-DCF [45] also allocates differentiated and unique initial back-
gorithms are fully decentralized, thus no information exchanges off values to each station, but the backoff value of each station
among transmitting stations or additional control frames are re- is then dynamically adapted based on the network load. By us-
quired, which would increase system complexity. ing traffic type information in the packet header, high priority traf-
Beacon-based Collision-free Channel Access (BCCA) [39] contin- fic will have a higher transmission probability during the collision
ues the idea of deterministic backoff. It initializes a determinis- resolution stages. In order to avoid overhead due to wasted idle
tic backoff period V(d), and all stations can only select a specific slots, the backoff time is shortened and the procedure of selection
backoff values between 0 and V(d). V(d) increases gradually as the of identical values for backoff period is eliminated.
number of stations raises. Besides, each station initializes its ap- In [46], the authors propose Back2F protocol to reduce the con-
proaching process upon receiving each beacon frame in BCCA. This tention and idle channel overheads. Back2F adopts a second an-
process allows each station to check whether there are empty slots tenna to listen to all the subcarriers, and migrates WiFi backoff to
in lower value. If there are, the station then chooses an empty the frequency domain. Instead of counting down from CW to 0 in
slot with a lower value. Thus, BCCA decreases the number of idle BEB, Back2F just transmits a symbol on the CWth subcarrier, which
slots and increases the channel efficiency. Uninterrupted Collision is narrowband OFDM channels used by 802.11. The second antenna
Free MAC adaptation (UCFA) [40] protocol also adopts the idea of of each station will listen to all the subcarriers from each compet-
bounded backoff value V(d). However, extended Kalman filtering ing station, and the station with the lowest value of subcarrier will
(EKF) estimation is adopted to estimate the number of competing win the contention and get transmission opportunity while others
stations to determine the optimal value of V(d). should defer its own transmission until the current transmission is
The authors of [41] propose a novel collision resolution solu- finished.
tion called centralized random backoff (CRB). The backoff state of
each station in CRB is centralized controlled by the access point 4.1.3. IEEE 802.11e EDCA tuning
(AP). The access point allocates a unique backoff value to the sta- While the QoS-aware features of IEEE 802.11e have been proved
tion making use of the ACK frame after a successful reception of to be helpful in an industrial context [47], on the other hand, some
a data frame from a station. CRB is expected to be a more effec- other literature aims to optimize the backoff and AIFS parameters
tive solution than the distributed deterministic backoff approaches, in 802.11e EDCA to improve the QoS for industrial delay-sensitive
52 Y. Cheng, D. Yang and H. Zhou et al. / Computer Networks 157 (2019) 41–67

traffic. In [48], the authors propose an algorithm to compute the number of active STAs in each AC. Thus, when the network be-
optimal configuration of the EDCA parameters based on the delay comes dense, it increases the CW for each AC to reduce the colli-
requirement of the application. The optimal configuration is cal- sion rate. Otherwise, the CW for each AC is decreased.
culated according to the proposed analytical model of EDCA per- In [58], the authors propose a two-level protection and guaran-
formance, which provides average throughput, the average delay, tee for real-time traffic in IEEE 802.11e distributed WLANs. Fast-
and the standard deviation of the delay as performance figures for backoff (BF) is proposed under the intra-frame-based scheme. It
both saturated and unsaturated scenarios. The algorithm admits as defines the window-increasing factor to be a real number larger
many traffic stations as possible, while introducing least influence than 1 instead of fixed value of two in the original MAC. When
for best-effort traffic. a worse traffic condition appears, it is required to adopt a larger
In [49], the authors present a dynamic prediction scheme for backoff stage (window-increasing factor more than 2). Compared
the backoff and AIFSN parameters to enhance the QoS level for to the original binary exponential backoff, BF becomes faster when
real-time traffic in various network conditions. The prediction the backoff stage is large.
technique is based on supervised learning using a J48 classifier In [59], the authors present two completely different ap-
tree and a M5 regression model, and the prediction makes use of proaches to adjust CW values and AIFSN named adaptive EDCA
several network metrics including application bit rate, transmission (a-EDCA) and improved EDCA (i-EDCA). In a-EDCA, the AP contin-
rate, occupancy level of the wireless medium, etc. Thus, the algo- uously monitors the network traffic, and dynamically determines
rithm always selects the optimum AIFSN combination [50,51] and new CW values using a parameter (CWmin and CWmax) adapt-
CW parameter depending on the current channel conditions. ing algorithm and broadcasts these values via beacon frames. In i-
In [52], the authors present a new cross-layer MAC aiming to EDCA, when the channel is idle, the stations firstly select a new
trade off between achieving bounded delays and maximizing the AIFS value named Random IFS (RIFS), which is uniformly dis-
network utilization. The derived model is able to accurately predict tributed in a fixed interval. The motivation behind this is to avoid
the QoS performance at each network flow using different network heavy congestion in the network by increasing the number of idle
measurements. Particularly, the protocol can guarantee the maxi- slots between transmissions.
mum delay and maximum loss for delay-sensitive applications in Adaptive Enhanced Distributed Coordination Function (AEDCF)
different network configurations. [60] effectively supports time-bound real-time applications by
In [53], an inter-station priority scheme is proposed for IEEE changing the dynamic process of content window values. After
802.11e to access the wireless medium. The scheme provides a each successful transmission, CW values of each AC is reset more
higher granular QoS guarantee than standard EDCA due to the idea slowly to adaptive values based on their current sizes and the aver-
of per-user differentiation. By associating a dynamic user-weight to age collision rate while keeping prioritization. On the other hand,
each station, it provides efficient QoS control across multiple users, after each unsuccessful transmission of packet of ACi, the CW is in-
and achieves a high performance for packets with strict-delay re- creased with a Persistence Factor (PF) instead of doubling the CW.
quirements. In this way, it can reduce the probability of a new collision and
In [54], the authors propose a simple fixed-priority access consequently decrease delay.
scheme for industrial Wi-Fi networks. The approach disables ran- DFA-EDCA [61] considers adaptively tuning the backoff parame-
dom backoff for real-time exchanges, and makes use of inter- ters based on the network load changing within a short period. The
frame spaces (IFS) to achieve a finer assignment of priorities to scheme adapts the initial CW size of each AC to the network con-
time-critical data. Transmission collisions are prevented due to the dition by using channel congestion rate measuring scheme, and it
unique gap assigned for each real-time message. is dynamically updates at regular periods. Once a few failed trans-
In [55], the authors present an adaptive channel access mecha- missions occur, the scheme assumes it does not mean the network
nism aiming to alleviate contention in 802.11e EDCA and improve is congested but due to channel variation. The CW size of each AC
the performance of the time constraint real time data traffic. The should be directly affected by channel conditions, and the mea-
dynamic backoff adjustment scheme for four ACs is based on the surement of channel congestion rate is a parameter to determine
Average Queue Occupancy of AC_VO and AC_VI. The Average Queue the CW updating.
Occupancy (avg) is a parameter reflecting the congestion degree of Common advantages and disadvantages: These approaches
each QoS queue. When the avg of AC_VO is beyond the minimum above are usually based on standard DCF and EDCA with slight
threshold and below the maximum threshold, the CWmax of the modifications, which means they are compatible with the cur-
other three ACs will be increased by the multiple factor of n1. If rent IEEE 802.11 standards and easy for implementation based on
the avg of AC_VO is beyond the maximum threshold, the CWmax the Commercial-Off-The-Shelf (COTS) hardware and devices. How-
of the other three ACs will be increased by the multiple factor of ever, the proposed approaches still adopt (in whole or in part) the
n2. And if the avg of AC_VO is below the minimum threshold, then contention-based channel access, and therefore, they can just re-
the CWmax of the other three traffic categories will be reset to the lieve but cannot eliminate the influence of non-deterministic de-
original values. lay introduced by the backoff mechanism. These kind of schemes
D2D [56] is a delay-aware distributed dynamic CW adaption are quite enough to support the soft real-time applications,
approach. The CW adaption process is based on delay deviation which do not have a tight delay-bound and stringent real-time
and channel busyness ratio, which reflects the present delay and requirement.
channel congestion status of the network. Instead of the “blind”
increment-decrement scheme of CW in standard BEB, each station
computes the probability of increment-decrement (q) by an adap- 4.2. Contention-free MAC
tion formula precisely in D2D. In addition, a new medium access
control protocol named D2D Channel Access (D2DCA) is proposed Building a contention-free MAC is the most straightfor-
based on the D2D for the IEEE 802.11e networks. ward approach to meet the requirement of the delay-sensitive
In [57], the authors present an adaptive contention window traffic. A more stringent deadline constraint can be easily
backoff algorithm for the IEEE 802.11e EDCA protocol called Q- satisfied through a precise resource reservation. The typical
ABACW. The proposed scheme has two main features: firstly, it es- contention-free approaches include polling-based, TDMA and to-
timates the number of active STAs according to the channel state ken passing, which will be detailed described in the following
information. Then, the CW of each AC is adjusted based on the paragraphs.
 Y. Cheng, D. Yang and H. Zhou et al. / Computer Networks 157 (2019) 41–67 53

4.2.1. Polling-based approaches stations. The RT stations are guaranteed the highest medium ac-
HCCA is a polling-based mechanisms defined in IEEE 802.11e, cess priority by managing the AIFS and CW parameters. However,
and a number of literature [62–71] aims to provide better per- if multiple RT stations attempt to transmit at the same time, con-
formance for delay-sensitive multimedia traffic by HCCA enhance- secutive collisions still occur. And the TDMA-based Coordination
ment by improving the efficiency of the HCCA reference sched- Layer is proposed to serialize the transmissions of RT stations. All
uler in supporting QoS for multimedia traffic and reducing the the real-time stations are synchronized by the beacon frames, and
polling overhead of HCCA scheduler. Particularly, in [67], an al- they will compute their dedicated time slot boundaries by their
gorithm called adaptive multi-polling TXOP scheduling algorithm station ID. Costa et al. [79] shows the real-time performance com-
(AMTXOP) is proposed to support dynamic TXOP assignment for parison between the proposed scheme and HCCA, and the results
real-time video traffics transmission over IEEE 802.11e. In [68], a show it offers a significantly better behavior in what concerns the
scheduler named Overboost is presented. It adopts HCCA for band- maximum number of admitted streams.
width negotiation, and uses EDCA to support traffic if there are In [81], the authors present an industrial deterministic com-
no sufficient resources to meet the bandwidth requirement of the munication mechanism based on WLAN. To provide guarantee for
stream. real-time traffic, it adds a deterministic scheduling layer between
In [72], the authors propose two feedback-based bandwidth al- the MAC layer and upper layers, which just needs software chang-
location algorithms to be used within the HCCA for providing ser- ing. The data transmission is synchronized by beacon frames, and
vice with guaranteed bounded delays, namely feedback based dy- the transmission cycle is composed of a TDMA stage and polling
namic scheduler (FBDS) and proportional-integral (PI)-FBDS. The stage, which are designed for periodic real-time traffic and non-
two schemes are designed using classic discrete-time feedback periodic real-time traffic.
control theory. The proposed algorithms can be easily imple- SchedWiFi [82] is an approach aiming to support multiple traf-
mented in COTS wireless cards and have low computational costs. fic classes with different requirements in industrial automation. In
Having noticed the high complexity of HCCA, in [71], a de- SchedWiFi, a particular class of real-time traffic named Scheduled
terministic backoff (DEB) method is proposed to achieve virtual Traffic (ST) is protected from any interference brought by non-ST
polling via carrier sense on the wireless channel. The DEB actu- traffic by the concept of ST-Window and Time-Aware Shaper (TAS).
ally works on the carrier sense mode such as EDCA. However, the During the ST-Window, only the messages of the ST flow assigned
backoff counter of each station is arranged differently with oth- to the ST-Window can be transmitted. The TAS function is used to
ers so that the stations can be polled without collision. Like HCCA, prevent the traffic that could interfere with the transmission of ST
stations are served sequentially in a predefined order in DEB with messages. SchedWiFi allows a flexible transmission of the non-ST
a deterministic feature. Besides, guard interval and calibration are flows, which can transmit as long as they do not interfere with the
used to enhance the robustness against inter-AP interference. ST Window of any ST flow.
In [73], the authors present a hybrid-MAC scheme that consists In [83], the authors propose an intelligent TDMA scheduler
of a CSMA-based contention period and a TDMA-based trans- which allows to plan the individual packets of the different
mission period. Different from the typical hybrid MAC where streams, and minimizes the physical layer interference. The sched-
contention period is used to transmit non-real-time traffic, the uler adopts a heuristic algorithm, including three steps namely
contention period is used for transmission opportunity compe- input, connecting and scheduling. Two different strategies are
tition in the proposed scheme. Then, the devices that compete considered in the scheduling step: sequentially allocated and peri-
successfully will transmit the data during each TDMA slot. Sup- odically allocated. Considering the worst case delay, the proposed
posing N devices succeed in contention period (CP), each device TDMA strategies perform significantly better than the CSMA/CA
will secure a transmission slot in the contention-free period. (up to 99.8%).
The work in [84] borrows the scheduling idea in Time-Triggered
4.2.2. TDMA-based approaches Ethernet (TTE), and extends it with wireless capabilities based
TDMA is adopted in a variety of existing industrial wireless pro- on IEEE 802.11. The proposed wireless MAC protocols support the
tocols, such as WirelessHART [15,16] and ISA100.11a [74]. They are same type of traffic classes as TTE. To support the wireless real-
specially designed for real-time industrial process control applica- time traffic, it defines new first-order logic constrains reflecting the
tions, however the data rate of these protocols is relatively low. wireless medium particularities. Three different MAC protocols are
Thus, it is a good solution to build TDMA MAC based on current proposed to mitigate the problem that in wireless condition, in-
802.11 protocol and hardware for the delay-sensitive application cluding pre-scheduled time-slots, contention-based time-slots and
which has a high-speed requirement. contention-based phases.
An isochronous MAC (IsoMAC) protocol [75] is developed based In [85], a frequency domain polling MAC protocol (FDP-MAC)
on TDMA for industrial traffic. Before an IsoMAC node associates is proposed to meet the deterministic requirements in the field
to the AP, it sends a resource request frame which includes a de- of factory automation. It makes use of the orthogonal frequency-
tailed traffic specification to the AP. Then, the AP decides whether division multiplexing (OFDM) technology to achieve simultane-
the new node can be accepted or not according to the remaining ously polling. Based on the polling method, a dynamic TDMA
channel resource. The communication cycle consists of two differ- scheduling scheme is also proposed, which improves the utilization
ent phases, namely a contention phase for best-effort traffic and a of time slots. In addition, a downlink acknowledgement aggrega-
scheduled phase for real-time traffic. The beacon frame is modified tion is adopted to alleviate the inefficiency of the wireless physical
to contain the scheduling information from for the nodes that have layer in industrial applications.
real-time traffic. Another issue of TDMA implementation based on IEEE 802.11 is
In [76], a TDMA-based mechanism is proposed, which priori- how to take advantage of Commercial-Off-The-Shelf (COTS) hard-
tizes real-time traffic in IEEE 802.11 networks. It adopts central- ware so that it can be used practically. A majority of these work
ized architecture, using the beacon frame from the coordinator to chooses Atheros WiFi chipset-based commodity 802.11 hardware
synchronize all the real-time stations and assign fixed time-slot for due to the nature of open source driver compared with other ven-
them. Costa et al. [77,78] extend the previous work, and a FCR MAC dors. One of the groundbreaking work is SoftMAC [86], where the
(Forcing Collision Resolution MAC) with a TDMA coordination layer authors clearly identified the modification that should be done
is described. It is assumed an infrastructure network topology, and to implement TDMA MAC with commodity hardware. In addi-
a central coordinator manages a set of real-time and non-real-time tion, a precise synchronization method is also proposed based on
54 Y. Cheng, D. Yang and H. Zhou et al. / Computer Networks 157 (2019) 41–67

Fig. 15. An example TDMA structure from RT-WiFi [80]. How to take advantage of Commercial-Off-The-Shelf (COTS) hardware is a big issue focused by a great deal of work.

the hardware timer. The work is followed by MultiMAC [87] and network topology, which makes it flexible enough when deployed
FreeMAC [88], where a more detailed and precise control scheme is in the in industrial field. The proposed scheme is compared with
proposed. Other similar work includes WildNet [89], which adopts the IEEE 802.11s standard [100] under the real industrial environ-
a 802.11 overlay TDMA MAC to support long distance communica- ment, and it achieves better deterministic performance in all test
tion in rural area. Besides, in [90], the authors developed a scheme scenarios.
to leverage the capabilities of Atheros wireless cards to achieve In [101], the authors present a design framework of fault-
a TDMA implementation almost entirely in hardware. The solu- tolerant wireless networked control systems that is developed
tion focuses on low maintenance, easy setup and production use. for industrial automation applications. The experimental platform
In [91], the authors thoroughly analyze the feasibility of hrtimer WiNC [Active fault-tolerance in wireless networked control sys-
[92] as the clock source of station synchronization, and propose a tems] supports IEEE 802.11a/b/g standards, and the MAC protocol
TDMA MAC based on the hrtimer. optimized for industrial wireless communication is accomplished
Soft-TDMAC [93,94] is a TDMA MAC protocol with tight syn- using SoftMAC. The work is followed by [102], in which a fault es-
chronization achieving by a system interface for 802.11 overlay timation algorithm based on TDMA mechanism for the application
TDMA MAC protocols (SySI-MAC) development. SySI-MAC is able to of W-NCSs on industrial automatic control.
provide near deterministic timers and transmission protocol stack,
because parts of the code having real-time requirement of SySI-
MAC reside in the Linux kernel, while the rest reside in the Linux 4.2.3. Token-based approaches
user space. Besides, the Linux RT patch [95] is also adopted to Wireless token passing approach is another type of contention-
allow tweaking the real-time priorities of the running processes. free medium access. Wireless token passing networks are formed
Based on these features of SySI-MAC, Soft-TDMAC achieves tight by a set of stations that configured in a logical ring. In the logical
network-wide synchronization by pairwise synchronization mech- ring, a token will be passed around and only the station which has
anism, and it can schedule arbitrary TDMA transmission patterns. the token can transmit the data or can control the other stations
RT-WiFi [80] aims to combine the advantages of 802.11 physical to transmit the data, which ensures no collision will happen in the
layer and TDMA MAC to provide flexible real-time high-speed com- network.
munication for wireless cyber-physical control applications (shown Wireless Token Ring Protocol (WTRP) [103] offers delay guaran-
in Fig. 15). Different trade-offs can be customized among sampling tee and the bandwidth guarantee. Various modules are proposed
rate, jitter level, reliability, and compatibility with existing Wi-Fi to deal with various challenges of wireless media, such as mobil-
networks. The timer design is based on the Timing Synchroniza- ity, interference avoidance and collision avoidance. The transmis-
tion Function (TSF) timer in IEEE 802.11, and the stations synchro- sion in the logic ring of WTRP can be only in one direction. Each
nize to the AP periodically through the beacon frames. Based on station can hold certain amount of time named token hold time
this global time synchronization, a link scheduler is designed to (THT), during the time the station can transmit its data. Due to the
coordinate the channel access among the stations. Besides, the au- distributed and token ring based feature, it is robust to single node
thors notice the coexistence problem with the normal Wi-Fi de- failure, and can support flexible network topologies.
vices, thus, two mechanisms, including performing carrier sense Ripple [104] is a wireless token-passing protocol that adopts
and shortening inter-frame spacing, are proposed to improve its a decentralized controlled-access approach to protect nodes from
co-existence performance with regular Wi-Fi networks. The work unintentional packet collisions. In ripple, six types of frames
is followed by several enhancements [96–98], which includes jitter namely DATA, ready to send (RTS), clear to send (CTS), acknowl-
minimization, unknown delay compensation. edgement (ACK), ready to receive (RTR) and NULL. Except for the
Det-WiFi [99] is also a TDMA MAC protocol implementation fixed length DATA frame, all the frames are the same as the stan-
based on commercial off-the-shelf hardware. It is able to provide dard specified in IEEE802.11. Most important, RTS and RTR frames
a relative high speed as well as deterministic feature for industrial are used as tokens. In ripple, each station has four states namely
control applications. Moreover, it is said to support multi-hop transmission state, reception state, listening state and idle state.
 Y. Cheng, D. Yang and H. Zhou et al. / Computer Networks 157 (2019) 41–67 55

Virtual Token Passing- Carrier Sense Multiple Access (VTP- D stand for centralized and distributed; in Hops column, S, M and
CSMA) [105] is a virtual token passing approach which is comple- R stand for single-hop, multi-hops and ring topology, respectively;
mented by an underlying forcing collision resolution mechanism in Evaluation column, the, sim, exp stand for the validation meth-
that prioritizes the real-time traffic over the non-real-time traf- ods of each proposed schemes, namely theory, simulation and ex-
fic. VTP-CSMA only controls the traffic generated by real-time sta- periment.
tions rather than controls all traffic generated by all stations, which
simplifies the protocol. VTP-CSMA mainly implements four pro- 4.3. Other MAC mechanisms to improve deterministic behavior
cedures including initialization, transmission procedure, listening
procedure and the main procedure. Each station initiates the lis- Besides the channel access optimization schemes mentioned in
tening procedure when the channel is idle, and executes the trans- Sections 4.1 and 4.2, other MAC optimization techniques, such as
mission procedure when it holds the token. retransmission management, redundancy and rate adoption, are
In [106], the authors present a token and self-policing-based also adopted to improve deterministic behavior in IEEE 802.11 net-
scheduling scheme. The scheme proposes three kinds of token work. In this subsection, some representative optimization tech-
named permission tokens. Different tokens are circulated among niques are described in detail, and the comparison of these
the nodes with different requirements without priority, including schemes is shown in Table 4.
video nodes, voice nodes and data nodes. When a node hold the
token, it is permitted to transmit data when the channel is avail- 4.3.1. Retransmission management
able. The proposed scheme works in a distributed manner with- Retransmission management is an important issue for real-time
out a central controller, and the current token holder decides its industrial applications. A higher retry limit increases the chance
successor. To implement token-passing scheme on 802.11 MAC, the of successful packet transmission but introduces long transmis-
CW is eliminated, and AIFS values from AC_VI to AC_BE are also sion delay; while a small retry limit will reduce the transmission
adjusted. delay however result in a larger packet loss. It is a tradeoff to
SoftToken protocol [107] aims to offer differentiated services in choose a packet retransmission limit under different application
a deterministic manner by adding a token-passing procedure on scenarios. In fact, a part of approaches [28,80,110] mentioned in
top of 802.11. In SoftToken, a central entity (master) is used to co- Sections 4.1 and 4.2 combine a specific retransmission manage-
ordinate transmissions of other wireless nodes (slaves). Once the ment scheme with its channel access method. For example, in [28],
network is initialized, the master sends token request messages to the retry limit is set differently according to the different dead-
slaves based on the transmission schedule. Specifically, each slave line requirements of each new TC; in [80], the authors propose an
maintains a separate queue for different TCs to achieve traffic dif- in-slot retransmission mechanism, which is invoked immediately if
ferentiation and to provide QoS accordingly to the TC. After the the sender does not receive an ACK message from the receiver; In
transmission over, the slave returns the token by a token response [110], the retransmissions times of real-time packets are discussed
message. in case the underlying wireless network is WLAN-based. The paper
Group Sequential Communication (GSC) [108] is a scheme for argues that the main reason of non-deterministic behavior of wire-
use in industrial wireless communication networks. It reduces the less network is the combination of current MAC layer retransmis-
protocol overheads by compelling the sequential transfer of mes- sions and the corresponding backoff procedures. In the following
sages among the members of a real-time group (RT-group) with- paragraph, we will introduce several independent work of retrans-
out the traditional polling frames. To achieve sequential transfer, mission management.
GSC adopts a Virtual Token Passing (VTP) mechanism. In addition, In [111], the authors propose a control algorithm that dynam-
a real-time block acknowledgment frame is used to achieve error ically adjusts the MAC retransmission limit for networked control
recovery mechanism for all RT messages transmission. systems. In the design, a NCS performance error threshold is set
Token-DCF [109] is a distributed and dynamically adaptive MAC according to the application requirement, and the MAC controller
using opportunistic token passing method to reduce idle and colli- will adjust the optimal retry limit to keep the transmission error
sion times. The token assignment process in Token-DCF is done by rate below the threshold. If the threshold is too small that no retry
transmitting station to its neighboring node, and the neighboring limit value is feasible, the controller will choose the optimal value
node is assigned with highest priority. Once the previous transmis- of the retransmission limit to ensure the NCS performance error is
sion completed, it starts to transmit after short time period when minimized.
the channel is found to be idle. If there are multiple neighboring In [112], the authors present a retransmission scheme combin-
nodes, the station will estimate the queue length of the neighbors, ing with real-time worst-case scheduling analysis that can offer
and the token is always given to a neighbor with a non-zero queue hard real-time support. The retransmission of erroneous packet is
length, which improves the utilization of the medium. supported when the following two criterions are met: the dead-
Common advantages and disadvantages: Contention-free line of the packet is not reached and the extra network capacity
schemes can usually provide satisfied delay-bound for industrial consumed by retransmission process does not jeopardize already
soft real-time applications, even hard real-time applications. The guaranteed delay bounds of other packets. The scheme limits the
reason is that the consuming time of each state of the nodes can number of possible retransmissions, and the real-time scheduling
be controlled precisely, and the maximum delay-bound can be eas- analysis prevents the deadline miss brought by retransmission of
ily analyzed theoretically. Thus, it is relatively simple to meet the other packets.
deadline requirements for each real-time application when adopt- In [113], the authors present a fast retry limit adaptation
ing contention-free schemes. However, it is usually complicated to method for video traffic over IEEE 802.11e distributed networks.
implement these schemes based on the COTS hardware. Besides, The method derives the optimal number of retransmissions by re-
these schemes often have poor compatibility with the existing IEEE lating the drop probability to the perceived video distortion, and
802.11 devices. The real-time performance may have great degra- the packet delay to the expiration time. The retransmission strat-
dation if the regular stations and contention-free stations exist in egy does not require any feedback distortion or delay information
the same area. from the destination. In addition, the algorithms in the method
The comparison of these real-time enhancement MAC schemes are developed with limiting computational cost, which is easy to
is shown in Table 3. Note that in Table 3, CB and CD stand for implement on the commercially available 802.11 network interface
contention-based and contention-free; in Topology column, C and cards without relatively low computational resources.
56 Y. Cheng, D. Yang and H. Zhou et al. / Computer Networks 157 (2019) 41–67

Table 3
Comparison of the MAC protocols for delay-sensitive applications.

Scheme Topology CB CF Basic MAC Timeliness level Hops Evaluation Year

SETL [24] D CW tuning DCF enhanced QoS S the/sim 2011

DCF with competing terminals estimation [25] D CW tuning DCF enhanced QoS S the/sim 2006
DCF CW optimization [23] D CW tuning DCF enhanced QoS S the/sim 2008
self-adaptive backoff [26] D CW tuning DCF enhanced QoS S the/sim 2016
CAC/DAC [27] C/D CW tuning DCF enhanced QoS S exp 2012
deadline-constrained DCF [28] D CW tuning DCF soft real-time S the/sim 2016
L-BEB [29] D semi-determinism DCF soft real-time S the/sim 2008
SRB [35] D semi-determinism DCF soft real-time S the/sim 2013
CF-MAC [36] D semi-determinism DCF soft real-time S exp 2014
ZC [37] D semi-determinism DCF soft real-time S the/sim 2008
L-MAC/L-ZC [38] D semi-determinism DCF soft real-time S the/sim 2013
BCCA [39] D semi-determinism DCF soft real-time S the/sim 2013
CRB [41] C semi-determinism DCF soft real-time S the/sim 2017
SDB [42] D semi-determinism DCF soft real-time S the/sim 2015
ReB [43] D semi-determinism DCF soft real-time S the/sim 2010
I-DCF [45] D semi-determinism DCF soft real-time S sim 2010
optimal EDCA configuration [48] D enhanced EDCA EDCA enhanced QoS S the 2010
dynamic AIFSN tuning [49] D enhanced EDCA EDCA enhanced QoS S sim 2017
cross layer MAC design [52] D enhanced EDCA EDCA enhanced QoS S the/sim 2017
dynamic user-weight classification [53] D enhanced EDCA EDCA enhanced QoS S sim 2011
fixed-priority access [54] D enhanced EDCA EDCA soft real-time S exp 2016
efficient back-off [55] D enhanced EDCA EDCA enhanced QoS S/M exp 2013
D2D [56] D enhanced EDCA EDCA enhanced QoS S the/sim 2016
Q-ABACW [57] D enhanced EDCA EDCA enhanced QoS S the/sim 2016
two-level protection [58] D enhanced EDCA EDCA enhanced QoS S the/sim 2016
i-EDCA/a-EDCA [59] D enhanced EDCA EDCA enhanced QoS S sim 2008
DFA-EDCA [61] D enhanced EDCA EDCA enhanced QoS S/M the/sim 2008
AMTXOP [67] C enhanced HCCA HCCA enhanced QoS S the/sim 2016
Overboost [68] C enhanced HCCA HCCA enhanced QoS S the/sim 2016
feedback-Based Control [72] C enhanced HCCA HCCA soft-real time S the/sim 2007
DEB [71] C enhanced HCCA HCCA enhanced QoS S the/sim 2011
robust hybrid-MAC [73] C enhanced HCCA HCCA enhanced QoS S sim 2015
IsoMAC [75] C TDMA 802.11 guaranteed deadline S sim/exp 2011
TDMA-Based WiFi [76] C TDMA 802.11 guaranteed deadline S the 2008
FCR MAC [77,78] C TDMA 802.11 guaranteed deadline S the/sim 2010
industrial deterministic WLAN [81] C TDMA 802.11 guaranteed deadline S exp 2009
SchedWiFi [82] C TDMA EDCA soft real-time S/M sim 2015
intelligent TDMA [83] C TDMA 802.11 soft real-time S the/sim 2018
time triggered WLAN [84] C TDMA 802.11 guaranteed deadline S sim 2016
frequency Polling MAC [85] C TDMA 802.11 guaranteed deadline S sim 2016
SoftMAC [86] C TDMA 802.11 guaranteed deadline S exp 2005
MultiMAC [87] C TDMA 802.11 guaranteed deadline S exp 2005
WildNet [89] C TDMA 802.11 guaranteed deadline S/M exp 2007
Soft-TDMAC [93,94] C TDMA 802.11 guaranteed deadline S/M the/exp 2012
RT-WiFi [80,96–98] C TDMA 802.11 guaranteed deadline S the/exp 2017
Det-WiFi [99] C TDMA 802.11 guaranteed deadline S/M exp 2017
WiNC [101] C TDMA 802.11 guaranteed deadline S exp 2012
WTRP [103] D token ring 802.11 guaranteed deadline R the/sim 2004
Ripple [104] D token ring 802.11 guaranteed deadline R the/sim 2006
VTP-CSMA [105] D token ring EDCA guaranteed deadline R the/sim 2007
self-policing-based scheduling [106] D token ring EDCA guaranteed deadline R sim 2011
SoftToken [107] C token ring 802.11 guaranteed deadline R exp 2015
GSC [108] C token ring DCF guaranteed deadline R the 2013
Token-DCF [109] C token ring DCF guaranteed deadline R sim 2012

4.3.2. Redundancy parallel redundant transmission paths. The system consists of

Adopting redundancy in Wi-Fi networks is also a good candi- two wireless cards using different protocol (IEEE 802.11a and
date to significantly increase the timeliness and reliability for in- IEEE 802.11b) and a PRP interconnecting device. In addition,
dustrial automation systems. One of this kind of approach cou- the multi-radio diversity architecture applies some other basic
ples Wi-Fi and parallel redundancy protocol (PRP) specified in IEC diversity techniques to improve the overall quality, including
62439-3 [130] for industrial Ethernet. In PRP (also called seamless space diversity and a maintaining functionality when a redundant
redundancy), each frame is sent by transmitters with two copies structure fails named graceful performance degradation (GPD). The
on different paths. The one which reaches the destination firstly work is followed by Rentschler and Laukemann [115], which gives
is retained, and is sent to the upper layer; while the second one a detailed analytical results via a Markov model and a series of
should be discarded. It is obvious that PRP can reduce the packet measurements on the system.
loss and transmission delay in the network, because the frame is In [116–119], the authors present an improved channel redun-
lost only when they are dropped on both networks and the trans- dancy scheme named Wi-Fi Redundancy (Wi-Red). The main draw-
mission delay is always the minimum among the different copies back of current basic wireless PRP is that every frame is always
from different paths. sent on both channels. As a result, the overall traffic over the
In [114], the proposed Wi-Fi based seamless redundancy air is doubled, and bandwidth utilization is heavily reduced as
scheme uses two non-overlapping IEEE 802.11 channels as the the bandwidth in ISM band is a scarce resource. Moreover, the
 Y. Cheng, D. Yang and H. Zhou et al. / Computer Networks 157 (2019) 41–67 57

Table 4
Comparison of several other MAC mechanisms for delay-sensitive applications.

Type scheme Brief Evaluation Year

Retransmission management [28] For each TCs, the retry limits are set differently according to the different the/sim 2016
deadline requirements.
[80] To fit the TDMA based network, a specific in-slot retransmission exp 2013
mechanism is proposed.
[111] The algorithm dynamically adjusts the MAC retransmission limit based on the/sim 2010
the channel estimation.
[112] The scheduler only retransmits the packet that is within the deadline and the/sim 2009
does not influence other packets to meet their deadline.
[113] It is a fast retry limit adaptation method based on the packet drop the/sim 2016
probability and delay.
Redundancy [114,115] The seamless redundancy scheme uses two non-overlapping IEEE 802.11 exp 2012
channels as the parallel redundant transmission paths.
[116–119] The scheme reduces the amount of duplicate frames sent on air to avoid the/exp 2018
the heavy reduction of bandwidth utilization
ER [120] ER takes advantage of broadcast nature of wireless medium and minimizes the/sim/exp 2007
the number of transmissions.
XORR [121] XORR tries to solve several practical challenges by considering not only NC sim/exp 2009
but also the MAC layer characteristics.
[122] It presents a QoS-aware retransmission scheme based on cooperation sim 2011
between NC and EDCA.
[123] The minimum completion delay problem for IDNC is formulated as the/sim 2015
stochastic shortest path (SSP) problem.
[124] The minimum completion delay (MCD) problem for IDNC is formulated as the/sim 2016
a bin-packing problem.
Rate adaption [125] By using lowest rate in different situations, the two new rate adaption the/exp 2012
techniques can cope with the specific requirements of industrial
communication systems.
[126,127] The relevant sequence of rates for the next transmission are calculated the/exp 2016
according to the perceived SNR.
[128] The rate selection algorithm is based on the amount of the network traffic the/sim 2015
[129] The transmission rate is controlled by an adjustor according to the channel the/sim 2011
variation and system status

transmission queue may overflow if one channel temporarily ex- communication. In “Sort by time”, a packet enqueued to retrans-
perience severe disturbance, which causes frames to be dropped. mission queue earlier is processed earlier; In “Maximum clique”,
Hence, Wi-Red reduces the amount of duplicate frames sent on a coding graph is created to show which candidate packets can be
air. The idea is if an ACK frame is already sent back to transmit- encoded together. However, this minimum clique partition based
ter side, which means at least one data frame has been sent to selection is proved to be NP-hard, thus a heuristic algorithm must
the destination successfully, and if the other duplicate frame still be adopted.
deferred in the transmit queue, this duplicate frame should be re- XOR Rescue (XORR) [121] tries to solve several practical chal-
moved. The idea is achieved by a modified DCF and ACK mecha- lenges in NC-based retransmission, including overheads for recep-
nism named abortable DCF and cross acknowledgment. The ana- tion information, heterogeneous and time-varying networks and
lytical results of proposed scheme is shown in [131] and improved frame scheduling. XORR addresses these challenges by considering
in [119]. Moreover, the Wi-Red is implemented combining with not only NC but also its integration with the MAC layer character-
transmission and retransmission scheduling and bandwidth man- istics. In XORR, the reception status is probabilistically estimated
agement on IEEE 802.11 COTS components in [132]. The enhanced using Bayesian-learning estimation technique to eliminate the ex-
approach can meet the deterministic requirements under the cir- tra signaling overheads caused by reception reports. The problems
cumstance with heavy disturbance and interference. of frame reordering and head-of-line blocking can be easily solved
On the other hand, Network Coding (NC) has attracted exten- by adopting a per-station queue for each station.
sive research attention recently [133–136], because it can provide In [122], the authors propose a QoS-aware retransmission
significant improvement on throughput, delay and reliability in scheme based on cooperation between NC and EDCA. The coop-
wireless networks. NC exploits the inherent broadcast nature of eration encounters two problems, namely transmission scheduling
wireless medium, and it can transmit multiple packets with a sin- problem and constitution problem for the NC packet. Thus, when
gle transmission by allowing intermediate nodes to encode alge- several packets encoded into an NC packet, the packets should
braically packets [137,138]. In order to improve retransmission ef- be encoded considering priority class. Entire network performance
ficiency in WLAN, a lot of NC-based retransmission schemes have can be enhanced while keeping QoS differentiation between the
been proposed in place of ARQ. In NC-based retransmission, several traffic with different priorities.
lost packets with different destinations can be encoded into one Paper [123] considers the problem of minimizing the comple-
packet to retransmit, rather than retransmitting each lost packet tion delay for instantly decodable network coding (IDNC) in wire-
individually. less broadcast and multicast scenarios. The authors first extend the
In [120], a NC-based retransmission scheme called ER is IDNC graph to efficiently operate in both multicast and broadcast
designed. By effectively combining packets via NC, ER takes ad- scenarios. Then the minimum completion delay problem for IDNC
vantage of broadcast nature of wireless medium and minimizes is formulated as stochastic shortest path (SSP) problem. Based on
the number of transmissions. Two selection methods, namely these the theoretical guidelines, a maximum weight clique selec-
“Sort by time” and “Maximum clique”, is proposed as for unicast tion algorithm is designed to reduce the IDNC completion delay in
58 Y. Cheng, D. Yang and H. Zhou et al. / Computer Networks 157 (2019) 41–67

polynomial time. Besides, a quadratic-time heuristic clique selec- retransmission-induced latency. Under the heavy traffic condition,
tion algorithm is also designed, which supports the operation in the algorithm balances the queueing delay and the service time,
real-time applications. and relieves the initial queueing delay for the large data burst.
According to the construction features of IDNC, the authors in The RA algorithm combines with low-latency frame aggregation
[124] provide a general solution to the minimum completion de- scheduling and retransmission dispatching to further reduce the
lay (MCD) problem with a bin-packing problem model. Some new queueing and service delays.
NC constructors are also presented and a lower bound of comple- In [129], an intelligent rate control algorithm is proposed for
tion delay is given for the MCD solution in the general IRNC-based delay sensitive real-time traffic in WLAN. It achieves system perfor-
retransmissions. Moreover, the MCD problem is solved with low mance improvement by adaptively adjusting the transmission rate
computational complexity by a high-efficiency greedy algorithm, to the current channel status. The algorithm consists of two ele-
which guarantees the practicability of the algorithm. ments namely a link condition estimator and transmission rate ad-
justor. By collecting local ACK information, the condition estimator
4.3.3. Rate adaption evaluates the channel and system status. Then, the transmission
One important feature of IEEE 802.11 network is able to pro- rate adjustor controls the rate by adaptively learning the channel
vide multiple transmission rates that can be dynamically selected. variation and system status. The detailed results show the pro-
For example, IEEE 802.11g standard makes several transmission posed algorithm can significantly improve the system performance
rates available, ranging from 6 to 54 Mbps. The higher transmission in terms of packet loss rate and end-to-end delay.
rate can support high-speed application however may increase In [142], the authors aim to optimize the adaptive rate fallback
packet loss significantly, while the lower rate provides more robust (ARF) algorithm in WLANs for industrial wireless network. Instead
transmission. The IEEE 802.11 standard does not specify any rate of only considering the packet loss, the proposed algorithm dy-
adaptation (RA) technique, and most existing RA schemes aim to namically adjusts the transmission rate according to signal to noise
maximize the network throughput or minimize the average delay ratio (SNR) information and acknowledgment or block acknowledg-
[139,140] without considering deterministic features and reliability, ment (ACK/BACK) packet. Compared with SARF in [127], the algo-
which is prominently important for real-time industrial communi- rithm checks SNR periodically and automatically adjust the modu-
cations. The following paragraph will discuss the main RA schemes lation once the packet loss occurs.
with deterministic consideration in detail.
To validate the performance of different RA schemes under in- 4.4. Summary and insights
dustrial environments, in [125], a widespread RA technique namely
the automatic rate fallback (ARF) is compared with those of the This section provides a comprehensive overview of the con-
RA techniques actually implemented on two commercially avail- ventional IEEE 802.11 MAC for delay-sensitive applications. In in-
able IEEE 802.11 devices. The results show that these techniques dustrial wireless network field, contention based access schemes
are characterized by a relevant number of packet retransmissions can hardly be adopted in the control applications because of the
that may introduce a considerable performance degradation on the non-deterministic feature. However, it is acceptable to serve as a
service time. Hence the paper propose two new RA techniques, suitable mechanism in non-critical monitoring and data accumula-
namely static retransmission rate ARF (SARF) and fast rate reduc- tion applications. Besides, when the networks are less congested,
tion ARF (FARF), to cope with the specific requirements of indus- CSMA/CA based MAC can provide a suitable communication delay
trial communication systems. The SARF technique behaves like ARF. reduction. On the other hand, the contention-free based MAC is
However, to limit the number of transmission attempts, the re- introduced to ensure guaranteed channel access, and to eliminate
transmission packet adopts the lowest rate; In addition, the suc- the uncertainty brought by the channel contention. Recently, an in-
cessful retransmission event does not reset the number of consec- creasing number of work focuses on TDMA implementation based
utive failures. In FARF, a transmission failure will force the next on IEEE 802.11 hardware. However, their implementation complex-
transmission selecting the lowest rate. The work is extended in ity and poor scalability are still the problems. Apart from the core
[141], where the considered RA algorithms are tuned with some MAC channel access scheme, other MAC mechanisms, such as rate
enhancements conceived to enforce their timeliness. A smarter adaption, redundancy and retransmission management, also have
controller is also proposed to differentiate the previous commu- a considerable influence on the network delay performance. These
nication history according to the suitable destination node, and kind of optimizations are often used together with the contention-
only the statistics relevant to that destination node are used when based channel access, and the timeliness and reliability of the net-
choosing the transmission rate. work can be improved to varying degrees.
Rate selection for industrial networks (RSIN) algorithm
[126,127] is a RA algorithm specifically conceived for the real-time 5. Latest and upcoming 802.11 protocol for delay-sensitive
industrial scenario with the goal of minimizing the transmission applications
error probability and satisfying the packet deadline requirement.
In RSIN, before the transmission of each packet, an additional In the earlier sections, we have discussed the major real-time
field is contained to record the SNR relevant to the last received optimization schemes for the industrial delay-sensitive applica-
packet from a specific destination. Besides, the stations is aware tions in conventional IEEE 802.11 MAC. In this section, we present
of the relationship between the Packet Error Rate (PER) and SNR an overview of the recent existing literature of determinism en-
for any transmission rate in advance. In order to minimize the hancement on latest and upcoming WLANs for delay-sensitive ap-
residual transmission error probability while ensuring the packet plications, including IEEE 802.11n/ac/ax/ah. Besides, the work re-
is delivered within its deadline, the number of attempts and the lated to IEEE 802.11aa is also discussed.
relevant sequence of rates for the next transmission are calculated
according to the perceived SNR. 5.1. IEEE 802.11n for delay-sensitive applications
LLRA [128] is a latency-aware rate adaptation scheme aiming to
reduce the tail latency for delay-sensitive applications. The goal of Although adopting IEEE 802.11n in industrial scenario is still at
the RA algorithm is to find a rate setting with the lowest delay at a very initial stage, a part of researchers have done quite a few
the give percentile. In the presence of light traffic, the algorithm constructive work in this field. In [143], the authors present a de-
selects the highest rate setting with low PER, which reduces the tailed description of the innovations provided by the IEEE 802.11n
 Y. Cheng, D. Yang and H. Zhou et al. / Computer Networks 157 (2019) 41–67 59

Table 5
Comparison of the real-time enhancements of frame aggregation schemes in IEEE 802.11n.

Type scheme Brief Evaluation Year

Frame Aggregation [146] To meet the specific QoS requirements, the scheduler dynamically adjusts the sim 2012
aggregated frame size based on the access category of each frame.
[147] Aggregate whenever there are backlogged frames in the queue which can be the/exp 2011
transmitted within a single TXOP.
[148] Set an aggregation delay limit to meet the predefined delay threshold of an sim 2016
application.
SRA-MSDU [149] Enable the retransmission of the corrupted MSDUs to achieve better throughput and the/sim 2013
delay performance in erroneous channels.
CA-DFA [150] Dynamically adjust the amount of frame aggregation according to the level of the/sim 2012
congestion in the network.
AFA-CAC [151] Set constraints of the throughput and delay. sim 2014

amendment, and give several recommendations for appropriate de- sequence control based on their relative index in the SRA-MSDU
ployment of IEEE 802.11n WLANs in industrial communication sys- frame. These optimization schemes enable the retransmission cor-
tems. Both papers [144] and [145] analyze the IEEE 802.11n per- rupted subframes to be retransmitted without any duplication, and
formance for industrial real-time multimedia traffic in a real world preserve the subframes order at the receiver buffer.
scenario. In addition, the work in [80,127,132], which is described Congestion aware-delayed frame aggregation (CA-DFA) [150] is
in detail in Section 4.3, are implemented based on IEEE 802.11n an adaptive frame aggregation algorithm to improve QoS only us-
hardware platform for different network determinism improving ing information available at MAC layer to optimize the amount
approaches. In IEEE 802.11n amendment, and the most important of 802.11n aggregation. The algorithm dynamically adjusts the
MAC feature is the two-level aggregation mechanism. It is useful amount of frame aggregation according to the level of congestion
for many time-sensitive applications which have small frame size, in the network. The paper argues that the benefits of aggregation
such as the multimedia and the industrial monitoring applications, clearly depend on the amount of network congestion degree. In
because a reasonable aggregation scheme can reduce the packet a more congested network, the aggregation approach makes more
overhead, which improves the channel efficiency and resolves the efficient use of wireless resources. Thus, the CA-DFA stations will
channel contention. In the following paragraphs, some related and reduce their amount of MAC aggregation if the estimated conges-
representative researches will be reviewed in detail. A comparison tion is below a certain threshold, otherwise the amount of MAC
of these schemes is provided in Table 5. aggregation is increased to reduce the level of congestion in the
In [146], the authors give the relationship between latency and network. Two simultaneous metrics are used to estimate conges-
frame aggregation size for voice and video application respectively tion, namely the average service period duration and the average
through several simulations. The results show that the delay is in- access delay.
creased when rising the size of aggregated frame because of the Adaptation of Frame Aggregation (AFA-CAC) [151] is an aggre-
waiting time of other packets to construct the A-MPDU frame. gation mechanism of IEEE 802.11n with considering the service
Thus, the paper proposes a frame aggregation scheduler for multi- differentiation to provide QoS satisfaction for delay sensitive
media applications with specific QoS requirements by dynamically application. The principle of AFA-CAC is to set constraints of the
adjusting the aggregated frame size based on the access category throughput and delay. If the two constraints are not satisfied, the
of each frame. Each AC is defined with minimum and maximum number of aggregated A-MPDU will be decreased. Once adjusting
aggregation sizes, and the aggregation scheme is selected accord- the number of A-MPDU, all the parameters and QoS metrics should
ing to the AC value. be re-estimated. In the worst case, if the number of aggregated
In [147], a backlogged queue (BQ) aggregation approach is pro- frame reaches zero after successive adjustments, and the two
posed to avoid the waiting time in the nave aggregation schemes. constraints still cannot be satisfied, the flow will be rejected.
The idea of the algorithm is to aggregate whenever there are back-
logged frames in the queue which can be transmitted within a 5.2. IEEE 802.11ac for delay-sensitive applications
single TXOP. The backlogged queue length is used as an indica-
tor for delay-optimal framing. Besides, the algorithm has a good The IEEE 802.11ac amendment mainly aims to enhance the total
compatibility with the existing rate/power control and other non- network throughput of IEEE 802.11n as well as individual link per-
aggregating stations, which makes it easy to implement and de- formance. A lot of research has discussed the new features, and an-
ploy. alyzed the performance of each enhancements [165,166]. Besides,
Paper [148] presents an A-MSDU real-time scheduler for time a few of the studies have discussed the optimization on network
sensitive applications, and satisfies their QoS requirements. It ex- real-time performance. These representative researches (Table 6)
ploits the A-MSDU attributes and enables retransmission to obtain will be reviewed in the following paragraphs.
aggregation with small size. The algorithm sets an aggregation de- In [152], a multi-polling controlled access (MPCA) scheme is
lay limit to meet the predefined delay threshold of an application. proposed to guarantee the latency in transmitting video frames
In addition, the received MSDUs are queued in the queue based on while reducing frame overhead in IEEE 802.11ac WLANs. MPCA
their smallest life time, and the highest priority will be provided defines two transmission types, namely reserved transmission
to the two MSDUs having the same life time. (RES_TX) and normal transmission (NOR_TX). In RES_TX, the chan-
In [149], the authors present an Enhanced A-MSDU frame ag- nel resource is handled by the hybrid coordinator like in HCCA,
gregation scheme with selective retransmission (SRA-MSDU) in which is designed for high real-time requirement traffic. The chan-
802.11n to achieve better throughput and delay performance in er- nel access method of NOR_TX is the same as EDCA but each QSTA
roneous channels. It enables the retransmission of the corrupted has to choose a backoff value to prevent collision. The MPCA is
MSDUs at the MSDU level by adding control fields in the A-MSDU combined with the MU-MIMO feature in IEEE 802.11ac to enhance
subframe header without changing the MAC header of the A-MSDU the reliability. Compared with EDCA, MPCA shows a lower packet
frame. Besides, the paper proposes an implicit MSDU level frame delay, lower packet loss, and higher peak signal-to-noise ratio.
60 Y. Cheng, D. Yang and H. Zhou et al. / Computer Networks 157 (2019) 41–67

Table 6
Comparison of the real-time improving schemes based on IEEE 802.11ac.

Type scheme Brief Evaluation Year

Real-time Improvement MPCA [152] A multi-polling controlled access (MPCA) scheme is proposed to guarantee the latency in sim 2017
video transmission.
[153] The paper investigates the efficiency of using RTS/CTS in the face of the QoS constraints, in the/sim 2015
different ACs and amounts of aggregation at the 802.11ac MAC layer.
[154] The authors alleviates the HC problem and reducing the error rate in 802.11ac network. the/sim 2015
[155] A new method to control the frame aggregation size for MU-MIMO based on the traffic sim 2017
variation.
MUMAM [156] The paper proposes a hybrid access mechanism named multi-user multi-cast access sim 2017
mechanism (MUMAM) to improving downlink multiuser transmissions with intra-AC
differentiation.
[157] An enhancement to the IEEE 802.11ac for channels supporting multipacket reception (MPR) sim 2014
for QoS provisioning.
[158] An adaptive traffic-aware backoff scheme for IEEE 802.11ac. sim 2016

In [153], the authors investigate the efficiency of using RTS/CTS duty cycle. According to different traffic loads, the two functions
in the face of the QoS constraints, in different ACs and levels of ag- can adapt the CSMA/CA parameters to a value to optimize the net-
gregation at the 802.11ac MAC layer. The results show that RTS/CTS work performance. The traffic-aware scheme achieves better de-
could bring significant delay in the Video and Voice ACs of the size lay and throughput performance compared with the conventional
of several tens of milli-seconds. To meet the different QoS require- scheme using constant backoff parameters.
ment, the parameter of RTS/CTS should be adjusted.
Paper [154] aims to solve a problem called “Hidden Channel 5.3. IEEE 802.11aa for delay-sensitive applications
(HC)” in 802.11ac network. The HC problem occurs because the
heterogeneity of devices use different bandwidths. The different The IEEE 802.11aa amendment mainly focuses on enhancing the
power density also affects channel sensing/reserving ranges. The QoS provisioning for robust audio-video (AV) streaming. Thus, the
authors analyzed the impact of CSMA contention parameter vari- majority of the existing work [167–169] on 802.11aa aims to im-
ations on this issue, and proposed a low complexity heuristic prove the performance of timeliness and reliability, based on the
algorithm to solve the problem. By alleviating the HC problem, three new mechanisms (GCR, SCS and OBSS management) dis-
the error rate and throughput of the network are both improved cussed in Section 3.3. The representative works are summarized
compared with the basic channel allocation scheme (RSSI based). in Table 7.
Paper [155] proposes a new method to control the frame ag- Paper [159,160] focuses on the improvement of real-time trans-
gregation size for MU-MIMO in IEEE 802.11ac WLAN. The scheme mission over WLANs based on IEEE 802.11aa. The proposed scheme
considers both channel utilization and delay of data frames suf- uses the two traffic selection algorithms (strict priority algorithm
fering from in transmission queues, and determines the frame ag- and the wireless credit-based shaper algorithm (WCBSA)) to prior-
gregation size in response to the traffic variation among spatial itize traffic within the voice and video ACs of EDCA in two video
streams. When the traffic variation is small, the aggregation strat- transmission queues (primary and alternate). The proposed WCBSA
egy is set according to the minimum policy. When the traffic varia- scheme adds the possibility of controlling the throughput of the
tion is large, the aggregation size adopts the average policy. The re- two VI transmit queues, which reduce frame losses and improve
sults show the proposed scheme achieves a better delay time per- the multimedia transmission in WLANs.
formance than the default one. In [161], the authors deploy a mid-size real-life testbed to per-
In [156], the authors propose a hybrid access mechanism form an extensive Quality of Experience (QoE) evaluation of GATS
named multi-user multi-cast access mechanism (MUMAM). The in IEEE 802.11aa under a variety of scenarios, including a real-life
scheme improves downlink multiuser transmissions with intra- scenario under no interference, under increasing UDP interference
AC differentiation. The scheme combines both benefits of IEEE and uncontrolled interference during a typical working day. Be-
802.11aa and IEEE 802.11ac to provide robust transmission for sides, the paper shows how the usage of 802.11e EDCA differen-
multicast and multi-user transmissions. Using the MUMAM tech- tiation impacts the general network performance with interfering
nique, unicast and multicast flows are transmitted in the alter- data traffic. The results can be used to derive optimal network con-
nate and primary queues respectively. The primary queues are pro- figuration guidelines for 802.11aa networks.
vided with high priority, and the alternates queues are provided In [162], an analytical model is proposed to evaluate the per-
with lower priority. Compared with EDCA TXOP technique, MU- formance of the mechanisms included in the 802.11aa standard in
MAM serves flows with lower delays and better throughput, es- a mixed scenario with multicast streams and unicast traffic. Each
pecially for secondary and complementary ACs. of the mechanisms offers different tradeoffs between reliability,
In [157], the authors propose an enhancement to the IEEE efficiency and complexity, depending on the considered scenario.
802.11ac for channels supporting multipacket reception (MPR) for Moreover, a novel algorithm is presented to select the best mech-
QoS provisioning. When adopting MPR, service differentiation can anism under a specific scenario. After providing the input of the
be provided by adding two new parameters (threshold and counter number of unicast and multicast receivers in the WLANs, the al-
decrement value) with CWmin , CWmax and AIFSN. The performance gorithm determines which of the mechanisms defined in the stan-
evaluation shows better results on different metrics (throughput, dard provides the best performance.
delay and jitter). In [163], the authors propose a modified GCR-BA scheme for
In [158], the authors propose a traffic-aware backoff scheme for video streaming in IEEE 802.11aa standard. The scheme overcomes
IEEE 802.11ac. The paper first measures the traffic statistics (av- the scalability limitation of GCR-BA scheme and limits the control
erage ON/OFF periods and duty cycle on the channel), and then overhead. If the packet belongs to I-frame (defined in H.264 stan-
design two functions, namely arctangent and Weibull functions, dard), each receiver should transmit the BA. On the contrary, if the
to adapt backoff parameters in CSMA/CA, based on the estimated packet belongs to the P-frame and B-frame, there is no BAR and
 Y. Cheng, D. Yang and H. Zhou et al. / Computer Networks 157 (2019) 41–67 61

Table 7
Comparison of the real-time improving schemes based on IEEE 802.11aa.

Type Scheme Brief Evaluation Year

Real-time Improvement [159,160] Two traffic selection algorithms are proposed to prioritize traffic sim 2014
within the voice and video ACs of EDCA.
[161] Evaluation of GATS in IEEE 802.11aa under a variety of scenarios. exp 2018
[162] A novel algorithm is presented to select the best mechanism under the/sim 2014
a specific scenario (unicast or multicast).
[163] A modified GCR-BA scheme is proposed for video streaming in sim 2014
IEEE 802.11aa standard.
VQD [164] A new cross-layer design named virtual queue dropping (VQD) is the/sim 2016
proposed for packet scheduling between AC_VI and AAC_VI.

Table 8
Comparison of the real-time improving schemes in IEEE 802.11ax.

Type scheme Brief Evaluation Year

Real-time Improvement [170] The paper highlights the impact of the contention window enlargement on different sim 2016
classes of traffic.
COSB [171] A novel backoff scheme named COSB is proposed to handle the high-density contention the/sim 2018
challenges in IEEE 802.11ax.
CSMA/CDA [172] The paper presents a collision detection and avoidance media access mechanism named sim 2015
CSMA/CDA for IEEE 802.11ax WLANs.
[173] A scheme combining adoption of DSC along with an intelligent scheme to enable RTS/CTS the/sim 2016
on selected stations, which reduces the collisions significantly.
[174] A novel scheduler, called MUTAX is designed. It can adaptively choose the best strategy, the/sim 2017
and significantly outperform existing popular schedulers on the performance of delay
and throughput.
OHCA [175] The paper presents a pragmatic and efficient OFDMA-based hybrid channel access (OHCA) sim 2018
scheme for IEEE 802.11ax.

no BA. This scheme helps to minimize the control overhead and tending STAs. Furthermore, the performance of the proposed solu-
improve the scalability of GCR-BA. tion is analyzed by a recursive discrete-time Markov chain model
Paper [164] proposes a new cross-layer design named virtual (R-DTMC). Compared with the other modified contention schemes,
queue dropping (VQD) for packet scheduling between AC_VI and COSB can achieve better delay and throughput performance when
AAC_VI. Aiming to improve the real-time performance of video the number of contending STAs increases.
transmission, VQD adopts conditional priority weighting to en- Paper [172] presents a collision detection and avoidance me-
hance the priority of I-packets from AAC_VI, which make them can dia access mechanism named CSMA/CDA for IEEE 802.11ax WLANs.
fairly compete with P-packets from AC_VI. During network conges- When a wireless station requires channel access, it should acquire
tions, VQD is adopted to make the dropping probabilities of lower- the channel via Random Access (RA) procedure firstly. When trans-
priority packets not only larger but also affected-and-increased by mitting data, a Random Register Number (RNN) would be added
those of higher-priority ones. VQD can maintain priority for AC_VI in the header of the data packet. Then AP is able to use it to an-
and fairness for AAC_VI simultaneously, which guarantees the real- nounce the preselected time slot of this station for next transmis-
time and reliable video transmissions. sion. The RNN messages carry the prescheduling channel informa-
tion, and the stations can adjust its backoff timer to reduce the
5.4. IEEE 802.11ax for delay-sensitive applications occurrence of collision.
To increase the spectral reuse in 802.11ax, Dynamic Sensitiv-
The upcoming IEEE 802.11ax-2019 amendment aims to address ity Control (DSC) algorithm is proposed by modifying current CCA
dense scenarios and satisfy tremendous demands for user expe- schemes. However, it leads to more frame collisions as well as
rience with the new physical layer design and subsequent novel more serious hidden node problem. In [173], the authors propose
MAC features [176,177]. To cope with the QoS requirements of the combined adoption of DSC along with an intelligent scheme to
the significantly increasing audio and video content, a number of enable RTS/CTS on selected stations in dense WLAN networks. Sev-
researches intend to optimize the WLAN performance based on eral methods are presented to select stations for the 4-way hand-
the current proposed MAC features [22,178]. The representative re- shake. Compared with legacy IEEE 802.11, DSC along with 4-way
searches (Table 8) will be reviewed in the following paragraph. handshake can reduce collisions significantly, and provide through-
In [170], the authors investigate the proposal of contention win- put and fairness gains.
dow enlargement [179], particularly by the reduction of time slot. In [174], the authors analyze problems that arise when adapt-
The proposal suggests the CW increasing the number of slots from ing classic schedulers to OFMDA systems in IEEE 802.11ax. Specif-
16 to 32 slots on initialization, and the slot duration is decreased ically, the authors have considered the scenario of uplink schedul-
from 9 to 4.5 μs. The paper highlights the impact of the proposal ing which targets the minimization of the average upload time
on different classes of traffic, including data priority and respec- with high number of active users, and the schedulers designed
tive window time of each access category. The results indicate that for a similar problem are not optimal for IEEE 802.11ax networks.
the proposal mitigates the collisions in the backoff procedure, and Thus in this paper, a novel scheduler, called MUTAX is designed.
decreases frame loss. It can adaptively choose the best strategy, and significantly outper-
In [171], a novel backoff scheme, called self-scrutinized chan- forms existing popular schedulers on the performance of delay and
nel observation-based scaled backoff (COSB), is proposed to han- throughput.
dle the high-density contention challenges in IEEE 802.11ax. The Paper [175] provides an accurate analytical model to derive
scheme adaptively scales up/down the size of contention window the theoretical upper bound of efficiency of random access pro-
by an estimated channel collision probability observed by the con- tocol in 802.11ax. Then, based on the de-facto collision resolution
62 Y. Cheng, D. Yang and H. Zhou et al. / Computer Networks 157 (2019) 41–67

Table 9
Comparison of the real-time enhancements of RAW in IEEE 802.11ah.

Type scheme Brief Evaluation Year

RAW enhancement [181] When collision of alarm-triggered traffic happens, it not only freeze its backoff timer, but sim 2012
further extends its timer value by a random duration.
[182] The scheduler calculates the optimal RAW parameters based on the current traffic sim 2017
conditions.
TAROA [183] The algorithm adapts the RAW parameters based on the current traffic conditions. the/sim 2017
[184] It uses a periodically reoccurring pool of reservation slots to alleviate the collisions and get the/sim 2016
guaranteed delay-bound.
DEARF [185] The delay-sensitive devices are allocated contention free slots to avoid delay, while the the/sim 2017
delay tolerant devices are allowed to transmission only in the contention-based basic
0.11ah access mechanism.

algorithm in 802.11ax (proposed in [180]), the authors present formulation and analytic result, an energy-delay aware RAW win-
a pragmatic and efficient OFDMA-based hybrid channel access dow control algorithm is proposed to dynamically adjust RAW size
(OHCA) scheme for IEEE 802.11ax. The scheme appropriately ad- to jointly optimize energy efficiency and delay. At first, the algo-
justs the CWO to its optimal value, and mitigates intensive colli- rithm estimates the number of time slots and internal slot dura-
sion among stations in dense WLAN scenario. tion in one RAW for different groups. After the estimated parame-
ters are obtained, a gradient descent approach is applied to get an
optimal solution for global network performance.
5.5. IEEE 802.11ah for delay-sensitive applications
In [183], the authors propose a traffic-adaptive RAW optimiza-
tion algorithm (TAROA), which adapts the RAW parameters based
IEEE 802.11ah amendment is defined to support for many
on the current traffic conditions. The algorithm first estimates the
IoT applications, such as smart grid, environmental/agricultural
packet transmission interval based on the packet transmission in-
monitoring and industrial process sensors [186,187]. However the
formation from AP of last transmission period. Then, it calculates
foreseen access mechanisms of IEEE 802.11ah rarely considers the
the optimal RAW parameters and assigns stations to RAW slots.
issue of deadline constraint for time-sensitive applications. For
The algorithm is executed at the start of each beacon interval in
example, most of the researches on enhancing restricted access
real-time, and the instantaneous adaptation significantly improves
window (RAW) [188–191] only aim to improve the throughput or
the network throughput and delay.
the energy efficiency without considering the real-time constraint
Paper [184] proposes an adaptive access mechanism for IEEE
for delay-sensitive applications. Hence, a few of the recent studies
802.11ah that supports several kinds of industrial traffic, namely
give several thorough analyses on this problem, and propose dif-
periodic, on-demand, and alarm reporting. The main idea of the
ferent solutions to cope with these requirements. In the following
mechanism is to use a periodically reoccurring pool of reserva-
paragraph, the related researches will be reviewed in detail, and a
tion slots with a fixed period, and the reoccurring pool consists
comparison of these schemes is shown in Table 9.
of a pre-allocated pool and a common pool. Each reservation slot
In [192], an analytical approach is proposed to investigate and
in pre-allocated pool is dedicated to one stations. If collision hap-
customize many mechanisms developed by IEEE 802.11ah, espe-
pens, every collision slot from the pre-allocated pool expands into
cially RAW. The paper gives the time distribution of the successful
frame with several slots in the common pool, where the collided
transmission for all stations in the network by a Markov model.
stations continue to compete the channel. If the collisions still ex-
Based on the distribution, the minimal RAW slot duration can be
ist, it repeats the slot expansion process until the collision is totally
calculated out. In [193], the authors mainly focus on the grouping
eliminated or the expansion limit is reached. Thus, the mechanism
strategy of RAW mechanism (no-crossing boundary case) in a satu-
is both efficient and reliable to resolve all reporting traffic while
rated network. The paper proposes an analytical model to track the
meeting the deadline requirement of the application.
impact of grouping strategy on the network performance. Besides,
Delay and Energy Aware RAW Formation (DEARF) [185] is pro-
Both papers [194] and [195] also evaluate the performance of RAW
posed for delay sensitive machine to machine (M2M) traffic in
in IEEE 802.11ah. In [194], the obtained evaluation results show
IEEE 802.11ah networks. The scheme adopts a hybrid channel ac-
that the RAW mechanism can provide substantial improvements
cess consists of a contention-based and a contention-free period.
in delay performance, especially in highly-loaded dense network
The delay-sensitive devices are allocated contention free slots to
scenarios. Paper [195] analyzes the performance of IEEE 802.11ah
avoid delay, while the delay tolerant devices are allowed to trans-
in an actuation use case from the perspective of latency. The re-
mission only in the contention-based basic.11ah access mechanism.
sults show that several techniques in IEEE 802.11ah have differ-
To implement the scheme, four types of RAWs, four type of RAWs
ent degrees of impact on the network latency performance, includ-
are defined with different size and functionalities. The proposed
ing beacon frame intervals, channel bandwidth, grouping based on
scheme is able to give reliable packet delivery for both delay-
transmission indication map.
sensitive devices and delay tolerant devices.
In [181], the authors propose an enhanced DCF mechanism
called limited local extension (LLE) in IEEE 802.11ah network for 6. Open issues and future research directions
alarm-triggered traffic in industrial automation. To handle the un-
predictable and time-critical alarm-triggered traffic, the local back- In the previous section, we have summarized the key features
off timer is modified in LLE. When a station with a pending alarm- of the reviewed real-time enhanced IEEE 802.11 based protocols
report detects another alarm-report from other station, it not only for delay-sensitive applications. In this section, open research prob-
freezes its backoff timer as in the standard BEB, but further ex- lems and future directions of the topic are provided.
tends its timer value by a random duration. With LLE, the channel
contention can be significantly reduced. 6.1. Problems in current contention-based protocols
Paper [182] first studies a RAW optimization problem in IEEE
802.11ah by adopting a novel retransmission scheme that utilizes Most of the contention-based protocols focus on adopting some
the next empty slot for retransmission in the uplink. Based on the schemes to make sure the delay as low as possible. These schemes
 Y. Cheng, D. Yang and H. Zhou et al. / Computer Networks 157 (2019) 41–67 63

usually use average delay as the reference metric. However, they including rate adaption [198], contention parameters optimization
can only decrease end-to-end latency while the delay sensitive ap- [199], dynamic MAC protocol adaption [200], bandwidth adaption
plications care more about the worst-case bound of delivery time, [201] and so on. The real-time performance of WLAN will benefit
though the performance of the contention-based protocols is still a lot from these schemes.
proved to be good for soft real-time applications (e.g., industrial
multimedia) when the network load is relatively light. Very a few 7. Conclusion
work [28] notices this issue, and provides probabilistic or worst-
case delay guarantee. However, the delay guarantees in these ap- This paper has presented a broad overview of research related
proaches cannot be achieved in the presence of a lossy channel. to MAC enhancement based on IEEE 802.11 protocols for wireless
This kind of approaches still needs further improvement to ensure real-time control applications. We propose a novel taxonomic clas-
a guaranteed deadline for time-sensitive applications. sification of the most representative real-time enhancement mech-
anisms in IEEE 802.11 network, and outline these mechanisms one
6.2. Problems in current contention-free protocols by one. The survey shows that the most of the contention-based
protocols have the advantage of ease to implement and deploy,
The contention-free protocols can naturally provide end-to-end while the contention-free protocols can support delay-sensitive ap-
delay-bound on transmission delay due to their deterministic fea- plications that have more stringent timeliness requirement. Be-
ture. However, the implementation complexity of these protocols sides, this paper describes the state of art of other three kinds of
is still a big problem, especially for the hybrid protocols that al- enhancement schemes, namely retransmission management, rate
ternate contention and contention-free periods, such as the HCCA adaption and seamless redundancy in detail. As for the recent IEEE
based approaches. Another problem for the contention-free proto- 802.11 MAC protocols, we pay more attention to IEEE 802.11n and
cols is the compatibility with the regular IEEE 802.11 terminals. In IEEE 802.11ah amendments. Although to design the real-time MAC
TDMA protocols, the regular IEEE 802.11 terminals that do not fol- mechanisms based on the two standards is still at a very initial
low the directives from the scheduler can be considered as the in- stage, the recent increasing number of mechanisms and protocols
terference. When a certain amount of regular IEEE 802.11 terminals described in this survey reflect the future trend in this field. The
are in the communication range of the TDMA terminals, the net- real-time enhancement would probably be one of the main future
work performance will suffer significant degradation. In addition, research directions for IEEE 802.11ah.
scalability is also an important issue in polling-based and TDMA
schemes due to the centralized network architecture. To design so- Acknowledgment
lutions that can scale to operate with a large number of nodes is
still a challenge. This work was supported partly by the National Key Research
and Development Program of China under grant 2018YFB1702001,
6.3. Real-time issues in IEEE 802.11ah/ax partly by National Natural Science Foundation of China under grant
61771040.
The IEEE 802.11ah amendment is proposed mainly for IoT
applications, including smart grid, environmental monitoring, Conflict of interest
industrial process sensors, most of which have varying degrees
of real-time requirements. The new access mechanism, RAW, can The authors declared that they have no conflicts of interest to
provide better QoS (delay, packet loss, throughput) in a dense this work. We declare that we do not have any commercial or as-
network compared with the basic DCF and EDCA. However, there sociative interest that represents a conflict of interest in connection
is no scheme to provide guaranteed delay-bound for the real-time with the work submitted.
traffic in IEEE 802.11ah standard. Most of the existing research
References
on enhancing RAW only aims to improve the throughput or the
energy efficiency without considering the real-time constraint. [1] N. Wang, Y. Zhu, W. Wei, J. Chen, S. Liu, P. Li, Y. Wen, One-to-multipoint laser
The real-time enhancement in RAW is important but still at a remote power supply system for wireless sensor networks, IEEE Sens. J. 12
(2) (2011) 389–396.
very initial stage, and the worst-case delay is only considered in
[2] S. Rinaldi, P. Ferrari, A. Flammini, F. Gringoli, M. Loda, N. Ali, An applica-
a few works. On the other hand, IEEE 802.11ax mainly aims to tion of IEEE 802.11ac to smart grid automation based on IEC 61850, in:
solve the problems (spectral efficiency and area throughput) in Industrial Electronics Society, IECON 2016 - Conference of the IEEE, 2016,
densely deployed Wi-Fi environments. Although the IEEE 802.11ax pp. 4645–4650.
[3] Y. Cheng, D. Yang, H. Zhou, Det-LB: a load balancing approach in 802.11 wire-
standard is expected to be finished by 2019, there has been a less networks for industrial soft real-time applications, IEEE Access (2018).
variety of work discussing its performance, enhancements and [4] I.E. Commission, et al., Industrial Networks-Wireless Communication Network
applications. However, most delay-related work in 802.11ax only and Communication Profiles-ISA 100.11 a, International Electrotechnical Com-
mission, 2014.
considers the performance of average latency as well. [5] G. Hemant, M. Subhas, G. Xiang, S. Nagender, WSN- and IoT- based smart
homes and their extension to smart buildings, Sensors 15 (5) (2015) 10350.
6.4. Improving real-time performance with New Tools [6] S. Yin, S.X. Ding, X. Xie, H. Luo, A review on basic data-driven approaches
for industrial process monitoring, IEEE Trans. Ind. Electron. 61 (11) (2014)
6418–6428.
Recently, some SDN-based schemes [196,197] have emerged to [7] J.A. Stankovic, Research directions for the internet of things, IEEE Internet
improve the real-time performance of networks based on WLAN. Things J. 1 (1) (2014) 3–9.
[8] G. Xiong, F. Zhu, X. Liu, X. Dong, W. Huang, S. Chen, K. Zhao, Cyber-physi-
In fact, using a SDN controller is more like an upgraded version of
cal-social system in intelligent transportation, IEEE/CAA J. Autom. Sin. 2 (3)
the common centralized MAC protocol (polling-based and TDMA (2015) 320–333.
etc.). Different traffic can be scheduled more precisely and all the [9] L. Zheng, Industrial wireless sensor networks and standardizations: the trend
of wireless sensor networks for process autometion, in: SICE Annual Confer-
transmissions in the WLAN can be determined. Thus the network
ence 2010, Proceedings of, IEEE, 2010, pp. 1187–1190.
can easily provide a guaranteed delay-bound with various schedul- [10] P. Zand, S. Chatterjea, K. Das, P. Havinga, Wireless industrial monitoring and
ing strategies for the industrial delay-sensitive applications. On the control networks: the journey so far and the road ahead, J. Sens. Actuator
other hand, as a powerful tool for prediction, machine learning Netw. 1 (2) (2012) 123–152.
[11] R. Candell, M. Kashef, Industrial wireless: problem space, success consider-
is also adopted widely to enhance the WLAN performance. These ations, technologies, and future direction, in: 2017 Resilience Week (RWS),
work aims to solve different problems in IEEE 802.11 MAC layer, IEEE, 2017, pp. 133–139.
64 Y. Cheng, D. Yang and H. Zhou et al. / Computer Networks 157 (2019) 41–67

[12] M. Weiner, M. Jorgovanovic, A. Sahai, B. Nikolié, Design of a low-latency, [43] Y. He, J. Sun, R. Yuan, W. Gong, A reservation based backoff method for video
high-reliability wireless communication system for control applications, in: streaming in 802.11 home networks, IEEE J. Sel. Areas Commun. 28 (3) (2010)
2014 IEEE International Conference on Communications (ICC), IEEE, 2014, 332–343.
pp. 3829–3835. [44] H. Zhao, J. Wei, N.I. Sarkar, S. Huang, E-MAC: an evolutionary solution for
[13] Bluetooth, (http://www.bluetooth.com). collision avoidance in wireless ad hoc networks, J. Netw. Comput. Appl. 65
[14] Zigbee, (http://www.zigbee.org). (2016) 1–11.
[15] J. Song, S. Han, A. Mok, D. Chen, M. Lucas, M. Nixon, W. Pratt, WirelessHART: [45] I. Paudel, B. Jouaber, I-DCF: improved DCF for channel access in IEEE 802.11
applying wireless technology in real-time industrial process control, in: IEEE wireless networks, in: Vehicular Technology Conference, 2016, pp. 1–5.
Real-Time and Embedded Technology and Applications Symposium, 2008, [46] S. Sen, R.R. Choudhury, S. Nelakuditi, No time to countdown: migrating back-
pp. 377–386. off to the frequency domain, in: International Conference on Mobile Comput-
[16] A. Saifullah, Y. Xu, C. Lu, Y. Chen, Real-time scheduling for WirelessHART net- ing and Networking, 2011, pp. 241–252.
works, in: Real-Time Systems Symposium, 2010, pp. 150–159. [47] G. Cena, L. Seno, A. Valenzano, C. Zunino, On the performance of IEEE 802.11e
[17] ISA100, (http://www.isa.org/isa100.). wireless infrastructures for soft-real-time industrial applications, IEEE Trans.
[18] A. Mahmood, R. Exel, H. Trsek, T. Sauter, Clock synchronization over ieee Ind. Inf. 6 (3) (2010) 425–437.
802.11a survey of methodologies and protocols, IEEE Trans. Ind. Inf. 13 (2) [48] P. Serrano, A. Banchs, P. Patras, A. Azcorra, Optimal configuration of 802.11e
(2017) 907–922. EDCA for real-time and data traffic, IEEE Trans. Veh. Technol. 59 (5) (2010)
[19] R. Exel, Clock synchronization in ieee 802.11 wireless LANs using physical 2511–2528.
layer timestamps, in: 2012 IEEE International Symposium on Precision Clock [49] E. Coronado, J. Villaln, A. Garrido, An adaptive medium access parameter pre-
Synchronization for Measurement, Control and Communication Proceedings, diction scheme for IEEE 802.11 real-time applications, Wirel. Commun. Mob.
IEEE, 2012, pp. 1–6. Comput. 2017 (2017) 1–19.
[20] A. Malik, J. Qadir, B. Ahmad, K.L.A. Yau, U. Ullah, QoS in IEEE 802.11-based [50] E. Coronado, J. Villaln, A. Garrido, Dynamic AIFSN tuning for improving the
wireless networks: a contemporary review, J. Netw. Comput. Appl. 55 (2015) QoS over IEEE 802.11 WLANs, in: Wireless Communications and Mobile Com-
24–46. puting Conference, 2015, pp. 73–78.
[21] X. Yu, P. Navaratnam, K. Moessner, Resource reservation schemes for IEEE [51] E. Coronado, J. Villalon, L.D.L. Ossa, A. Garrido, An AIFSN prediction scheme
802.11-based wireless networks: a survey, IEEE Commun. Surv. Tutorials 15 for multimedia wireless communications, in: International Conference on
(3) (2013) 1042–1061. Computer Communication and Networks, 2015, pp. 1–8.
[22] D.-J. Deng, S.-Y. Lien, J. Lee, K.-C. Chen, On quality-of-service provisioning in [52] A. Nafaa, A. Ksentini, On sustained QoS guarantees in operated IEEE 802.11
IEEE 802.11 ax WLANs, IEEE Access 4 (2016) 6086–6104. wireless LANs, IEEE Trans. Parallel Distrib. Syst. 19 (8) (2008) 1020–1033.
[23] D.J. Deng, C.H. Ke, H.H. Chen, Y.M. Huang, Contention window optimization [53] A.R. Rebai, S. Hanafi, A dynamic multimedia user-weight classification
for IEEE 802.11 DCF access control, IEEE Trans. Wirel. Commun. 7 (12) (2008) scheme for IEEE 802.11 WLANs, Int. J. Comput. Netw. Commun. 3 (2) (2011)
5129–5135. 1–3.
[24] C.H. Ke, C.C. Wei, K.W. Lin, J.W. Ding, A smart exponential-threshold-lin- [54] G. Cena, S. Scanzio, L. Seno, A. Valenzano, A fixed-priority access scheme
ear backoff mechanism for IEEE 802.11 WLANs, Int. J. Commun. Syst. 24 (8) for industrial Wi-Fi networks, in: Industrial Electronics Society, IECON 2016
(2011) 1033–1048. - Conference of the IEEE, 2016, pp. 4633–4638.
[25] A.L. Toledo, T. Vercauteren, X. Wang, Adaptive optimization of IEEE 802.11 [55] T. Sanguankotchakorn, A. Gopalasingham, N. Sugino, Adaptive channel ac-
DCF based on bayesian estimation of the number of competing terminals, cess mechanism for real time traffic over IEEE 802.11e Wi-Fi network, in: In-
IEEE Trans. Mob. Comput. 5 (9) (2006) 1283–1296. ternational Conference on Intelligent Systems Modelling & Simulation, 2013,
[26] C. Zhang, P. Chen, J. Ren, X. Wang, A.V. Vasilakos, A backoff algorithm based pp. 486–491.
on self-adaptive contention window update factor for IEEE 802.11 DCF, Wirel. [56] M. Khatua, S. Misra, D2D: delay-aware distributed dynamic adaptation of
Netw. (2016) 1–10. contention window in wireless networks, IEEE Trans. Mob. Comput. 15 (2)
[27] P. Serrano, P. Patras, A. Mannocci, V. Mancuso, A. Banchs, Control theoretic (2016) 322–335.
optimization of 802.11 WLANs: implementation and experimental evaluation, [57] I. Syed, S.H. Shin, B.H. Roh, M. Adnan, Performance improvement of QoS-en-
Comput. Netw. 57 (1) (2013) 258–272. abled WLANs using adaptive contention window backoff algorithm, IEEE Syst.
[28] G. Tian, S. Camtepe, Y.C. Tian, A deadline-constrained 802.11 MAC protocol J. PP (99) (2017) 1–11.
with QoS differentiation for soft real-time control, IEEE Trans. Ind. Inf. 12 (2) [58] Y. Xiao, F.H. Li, S. Choi, Two-level protection and guarantee for multimedia
(2016) 544–554. traffic in IEEE 802.11e distributed WLANs, Wirel. Netw. 15 (2) (2009) 141–161.
[29] J. Barcelo, B. Bellalta, C. Cano, M. Oliver, Learning-BEB: avoiding collisions in [59] T. Nilsson, J. Farooq, A novel MAC scheme for solving the QoS parameter ad-
WLAN, Int. Feder. Inf. Process. (1) (2008). justment problem in IEEE 802.11e EDCA, in: International Symposium on A
[30] J. Barcelo, B. Bellalta, C. Cano, A. Sfairopoulou, M. Oliver, Carrier sense multi- World of Wireless, Mobile and Multimedia Networks, 2008, pp. 1–9.
ple access with enhanced collision avoidance:a performance analysis, in: In- [60] L. Romdhani, Q. Ni, T. Turletti, Adaptive EDCF: enhanced service differentia-
ternational Conference on Wireless Communications and Mobile Computing: tion for IEEE 802.11 wireless ad-hoc networks, in: Wireless Communications
Connecting the World Wirelessly, 2009, pp. 733–738. and Networking, 2003. WCNC 2003, 2, 2003, pp. 1373–1378.
[31] J. Barcelo, B. Bellalta, A. Sfairopoulou, C. Cano, CSMA with enhanced collision [61] A.A. Shakir, C.K. Ng, N.K. Noordin, A. Ismail, Efficient back-off mechanism for
avoidance: a performance assessment, in: Vehicular Technology Conference, multimedia support in 802.11e EDCA wireless ad-hoc networks, Wirel. Pers.
2009. Vtc Spring 2009. IEEE, 2009, pp. 1–5. Commun. 71 (1) (2013) 439–466.
[32] L. Sanabria-Russo, A. Faridi, B. Bellalta, J. Barcelo, Future evolution of CSMA [62] H. Luo, M.L. Shyu, An optimized scheduling scheme to provide Quality of Ser-
protocols for the IEEE 802.11 standard, in: IEEE International Conference on vice in 802.11e wireless LAN, in: IEEE International Symposium on Multime-
Communications Workshops, 2013, pp. 1274–1279. dia, 2009, pp. 651–656.
[33] J. Barcelo, B. Bellalta, C. Cano, A. Sfairopoulou, M. Oliver, J. Zuidweg, Traffic [63] Z.A.M.B. Noh, T. Suzuki, S. Tasaka, A packet scheduling scheme for au-
prioritization for carrier sense multiple access with enhanced collision avoid- dio-video transmission over error-prone IEEE 802.11e HCCA wireless LANs,
ance, in: Communications Workshops, 20 09. ICC Workshops 20 09. IEEE Inter- in: Personal Indoor & Mobile Radio Communications IEEE International Sym-
national Conference on, 2009, pp. 1–5. posium on, 4, 2009, pp. 1527–1531.
[34] J. Barcelo, B. Bellalta, C. Cano, A. Sfairopoulou, M. Oliver, Dynamic Parameter [64] G. Cecchetti, A.L. Ruscelli, A. Mastropaolo, G. Lipari, Dynamic TXOP HCCA re-
Adjustment in CSMA/ECA, Springer Berlin Heidelberg, 2010. claiming scheduler with transmission time estimation for IEEE 802.11e real–
[35] Y. He, J. Sun, X. Ma, A.V. Vasilakos, R. Yuan, W. Gong, Semi-random backoff: time networks, in: ACM MSWIM, 2012, pp. 239–246.
towards resource reservation for channel access in wireless LANs, IEEE/ACM [65] G. Cecchetti, A.L. Ruscelli, A. Mastropaolo, G. Lipari, Providing variable TXOP
Trans. Netw. 21 (1) (2013) 204–217. for IEEE 802.11e HCCA real-time networks, in: Wireless Communications and
[36] L. Sanabria-Russo, F. Gringoli, J. Barcelo, B. Bellalta, Implementation and ex- NETWORKING Conference, 2012, pp. 1508–1513.
perimental evaluation of a collision-free MAC protocol for WLANs, in: IEEE [66] A. Jansang, A. Phonphoem, Adjustable TXOP mechanism for supporting video
International Conference on Communications, 2014, pp. 1036–1042. transmission in IEEE 802.11e HCCA, EURASIP J. Wirel. Commun. Netw. 2011
[37] J. Lee, J.C. Walrand, Design and analysis of an asynchronous zero collision (1) (2011) 158.
MAC protocol, Comput. Sci. (2008). [67] M.A. Al-Maqri, M. Othman, B.M. Ali, Z.M. Hanapi, Adaptive multi-polling
[38] M. Fang, D. Malone, K.R. Duffy, D.J. Leith, Decentralised learning MACs for scheduler for QoS support of video transmission in IEEE 802.11e WLANs,
collision-free access in WLANs, Wirel. Netw. 19 (1) (2013) 83–98. Telecommun. Syst. 61 (4) (2016) 773–791.
[39] M.F. Tuysuz, H.A. Mantar, A beacon-based collision-free channel access [68] A.L. Ruscelli, G. Cecchetti, A. Alifano, G. Lipari, Enhancement of QoS support
scheme for IEEE 802.11 WLANs, Wirel. Pers. Commun. 75 (1) (2014) 155–177. of HCCA schedulers using EDCA function in IEEE 802.11e networks, Ad Hoc
[40] M.F. Tuysuz, H.A. Mantar, G. Celik, M.R. Celenlioglu, An uninterrupted colli- Netw. 10 (2) (2012) 147–161.
sion-free channel access scheme over IEEE 802.11 WLANs, in: Wireless Com- [69] N.A. Ali, A. Bourawy, H. Hassanein, Selectivity Function Scheduler for IEEE
munications and Networking Conference, 2013, pp. 386–391. 802.11e HCCA Access Mode, John Wiley and Sons Ltd., 2013.
[41] J.D. Kim, D.I. Laurenson, J.S. Thompson, Centralized random backoff for col- [70] C.W. Chou, C.J. Lin, T.H. Lee, On efficient multipolling with various service
lision resolution in Wi-Fi networks, IEEE Trans. Wireless Commun. PP (99) intervals for IEEE 802.11e WLANs, in: Wireless Communications and Mobile
(2017). 1–1 Computing Conference, 2011, pp. 1906–1911.
[42] S. Misra, M. Khatua, Semi-distributed backoff: collision-aware migration from [71] Y. He, X. Ma, Deterministic backoff: toward efficient polling for IEEE 802.11e
random to deterministic backoff, Mob. Comput. IEEE Trans. 14 (5) (2015) HCCA in wireless home networks, IEEE Trans. Mob. Comput. 10 (12) (2011)
1071–1084. 1726–1740.
 Y. Cheng, D. Yang and H. Zhou et al. / Computer Networks 157 (2019) 41–67 65

[72] G. Boggia, P. Camarda, L.A. Grieco, S. Mascolo, Feedback-based control for pro- [100] G.R. Hiertz, D. Denteneer, S. Max, R. Taori, J. Cardona, L. Berlemann, B. Walke,
viding real-time services with the 802.11e MAC, IEEE/ACM Trans. Netw. 15 (2) IEEE 802.11 s: the WLAN mesh standard, IEEE Wirel. Commun. 17 (1) (2010).
(2007) 323–333. [101] S.X. Ding, P. Zhang, S. Yin, E.L. Ding, An integrated design framework of fault-
[73] P.K. Verma, R. Tripathi, K. Naik, A robust hybrid-MAC protocol for M2M com- -tolerant wireless networked control systems for industrial automatic control
munications, in: International Conference on Computer and Communication applications, IEEE Trans. Ind. Inf. 9 (1) (2012) 462–471.
Technology, 2015, pp. 267–271. [102] Y. Wang, S.X. Ding, D. Xu, B. Shen, An H∞ fault estimation scheme of wireless
[74] ISA100, (http://www.isa.org/isa100). networked control systems for industrial real-time applications, IEEE Trans.
[75] H. Trsek, J. Jasperneite, An isochronous medium access for real-time wire- Control Syst. Technol. 22 (6) (2014) 2073–2086.
less communications in industrial automation systems - a use case for wire- [103] M. Ergen, D. Lee, R. Sengupta, P. Varaiya, WTRP - wireless token ring protocol,
less clock synchronization, in: International IEEE Symposium on Precision IEEE Trans. Veh. Technol. 53 (6) (2004) 1863–1881.
Clock Synchronization for Measurement Control and Communication, 2011, [104] R.G. Cheng, C.Y. Wang, L.H. Liao, J.S. Yang, Ripple: a wireless token-passing
pp. 81–86. protocol for multi-hop wireless mesh networks, Commun. Lett. IEEE 10 (2)
[76] R. Moraes, F. Vasques, P. Portugal, A TDMA-based mechanism to enforce real– (2006) 123–125.
time behavior in WiFi networks, in: IEEE International Workshop on Factory [105] R. Moraes, F. Vasques, P. Portugal, J.A. Fonseca, VTP-CSMA: a virtual token
Communication Systems, 2008, pp. 109–112. passing approach for real-time communication in IEEE 802.11 wireless net-
[77] R. Costa, P. Portugal, F. Vasques, R. Moraes, A TDMA-based mechanism for re- works, IEEE Trans. Ind. Inf. 3 (3) (2007) 215–224.
al-time communication in IEEE 802.11e networks, in: Emerging Technologies [106] P. Koutsakis, Token- and self-policing-based scheduling for multimedia traffic
and Factory Automation, 2010, pp. 1–9. transmission over WLANs, IEEE Trans. Veh. Technol. 60 (9) (2011) 4520–4527.
[78] R. Costa, P. Portugal, F. Vasques, R. Moraes, R.F. Custodio, A coordination layer [107] L. Eznarriaga, C. Senguly, N. Bayer, J.I. Moreno, P. Lozano, M. Simn, Expe-
to handle real-time communication in Wi-Fi networks with uncontrolled traf- riences with SoftToken: a tokenbased coordination layer for WLANs, Int. J.
fic sources, in: Local Computer Networks, 2011, pp. 263–266. Commun. Syst. 28 (3) (2015) 457–471.
[79] R. Costa, P. Portugal, F. Vasques, R. Moraes, Comparing RT-WiFi and HCCA [108] J. Raimundo Viegas, L.A. Guedes, F. Vasques, P. Portugal, R. Moraes, A new
approaches to handle real-time traffic in open communication environments, MAC scheme specifically suited for real-time industrial communication based
2012, pp. 1–8. on IEEE 802.11e, Comput. Electr. Eng. 39 (6) (2013) 1684–1704.
[80] Y.H. Wei, Q. Leng, S. Han, A.K. Mok, W. Zhang, M. Tomizuka, RT-WiFi: real– [109] G. Hosseinabadi, N. Vaidya, Token-DCF: an opportunistic MAC protocol for
time high-speed communication protocol for wireless cyber-physical control wireless networks[C], in: 2013 Fifth International Conference on Communi-
applications, in: IEEE Real-Time Systems Symposium, 2013, pp. 140–149. cation Systems and Networks (COMSNETS), IEEE, 2013, pp. 1–9.
[81] Y. Zheng, A. Xu, Y. Song, W. Zhao, M. Liu, Industrial wireless deterministic [110] L. Seno, S. Vitturi, F. Tramarin, Tuning of IEEE 802.11 MAC for improving re-
communication based on WLAN: design, implementation and analysis, in: al-time in industrial wireless networks, in: Emerging Technologies & Factory
IEEE International Conference on Communications Technology and Applica- Automation, 2012, pp. 1–8.
tions, 2009, pp. 274–278. [111] J. Bai, E.P. Eyisi, Y. Xue, X.D. Koutsoukos, Dynamic tuning retransmission limit
[82] G. Patti, G. Alderisi, L.L. Bello, SchedWiFi: an innovative approach to sup- of IEEE 802.11 MAC protocol for networked control systems, in: IEEE/ACM
port scheduled traffic in ad-hoc industrial IEEE 802.11 networks, in: Emerging Int’l Conference on Green Computing and Communications & Int’l Conference
Technologies & Factory Automation, 2015, pp. 1–9. on Cyber, Physical and Social Computing, 2010, pp. 666–671.
[83] M. Deruyck, J. Hoebeke, E. De Poorter, E. Tanghe, I. Moerman, P. Demeester, [112] M. Jonsson, K. Kunert, Towards reliable wireless industrial communication
L. Martens, W. Joseph, Intelligent TDMA heuristic scheduling by taking with real-time guarantees, IEEE Trans. Ind. Inf. 5 (4) (2009) 429–442.
into account physical layer interference for an industrial IoT environment, [113] F. Babich, M. Comisso, R. Corrado, Fast retry limit adaptation for video dis-
Telecommun. Syst. 67 (4) (2018) 605–617. tortion/delay control in IEEE 802.11e distributed networks, Ad Hoc Netw. 36
[84] P.G. Peón, E. Uhlemann, W. Steiner, M. Björkman, Medium access control for (P1) (2016) 229–243.
wireless networks with diverse time and safety real-time requirements, in: [114] M. Rentschler, P. Laukemann, Towards a Reliable Parallel Redundant WLAN
Industrial Electronics Society, IECON 2016 - Conference of the IEEE, 2016, Black Channel, 2012.
pp. 4665–4670. [115] M. Rentschler, P. Laukemann, Performance analysis of parallel redundant
[85] J. Lin, W. Liang, H. Yu, X. Yang, Polling in the frequency domain: a new MAC WLAN, in: IEEE International Conference on Emerging Technologies & Factory
protocol for industrial wireless network for factory automation, Int. J. Ad Hoc Automation, 2012, pp. 1–8.
Ubiquitous Comput. 20 (4) (2015) 211–222. [116] G. Cena, S. Scanzio, A. Valenzano, C. Zunino, An enhanced MAC to increase
[86] M. Neufeld, J. Fifield, C. Doerr, A. Sheth, D. Grunwald, SoftMAC - flexible wire- reliability in redundant Wi-Fi networks, in: Factory Communication Systems,
less research platform, in: Proceedings of the Fourth Workshop on Hot Topics 2014, pp. 1–10.
in Networks Hotnetsiv, 2005. [117] G. Cena, S. Scanzio, A. Valenzano, C. Zunino, Dynamic duplicate deferral tech-
[87] C. Doerr, M. Neufeld, J. Fifield, T. Weingart, D.C. Sicker, D. Grunwald, Mul- niques for redundant Wi-Fi networks, in: Emerging Technology and Factory
tiMAC - an adaptive MAC framework for dynamic radio networking, in: First Automation, 2015, pp. 1–8.
IEEE International Symposium on New Frontiers in Dynamic Spectrum Access [118] G. Cena, S. Scanzio, A. Valenzano, Seamless link-level redundancy to improve
Networks, 2005, pp. 548–555. reliability of industrial Wi-Fi networks, IEEE Trans. Ind. Inf. 12 (2) (2016)
[88] A. Sharma, E.M. Belding, FreeMAC: framework for multi-channel MAC devel- 608–620.
opment on 802.11 hardware, in: ACM Workshop on Programmable Routers [119] G. Cena, S. Scanzio, A. Valenzano, Improving effectiveness of seamless redun-
for Extensible Services of Tomorrow, 2008, pp. 69–74. dancy in real industrial Wi-Fi networks, IEEE Trans. Ind. Inf. 14 (5) (2018)
[89] R. Patra, S. Nedevschi, S. Surana, A. Sheth, L. Subramanian, E. Brewer, WiLD- 2095–2107.
Net: design and implementation of high performance WiFi based long dis- [120] E. Rozner, A.P. Iyer, Y. Mehta, L. Qiu, M. Jafry, ER: efficient retransmission
tance networks., in: Usenix Symposium on Networked Systems Design & Im- scheme for wireless LANs, in: Proceedings of the 2007 ACM CoNEXT confer-
plementation, 2007, pp. 87–100. ence, ACM, 2007, p. 8.
[90] S. Leffler, et al., TDMA for Long Distance Wireless Networks, White Paper, [121] F.-C. Kuo, K. Tan, X. Li, J. Zhang, X. Fu, Xor rescue: exploiting network cod-
2009. ing in lossy wireless networks, in: Sensor, Mesh and Ad Hoc Communica-
[91] W. Torfs, C. Blondia, TDMA on commercial of-the-shelf hardware: fact and tions and Networks, 2009. SECON’09. 6th Annual IEEE Communications Soci-
fiction revealed, AEU - Int. J. Electron. Commun. 69 (5) (2015) 800–813. ety Conference on, IEEE, 2009, pp. 1–9.
[92] K.D. Kwon, M. Sugaya, T. Nakajima, KTAS: analysis of timer latency for em- [122] Y. Tanigawa, J.-O. Kim, H. Tode, QoS-aware retransmission with network cod-
bedded linux kernel, Int. J. Adv. Sci. Technol. 19 (2010). ing based on adaptive cooperation with IEEE 802.11 e EDCA, in: Global
[93] P. Djukic, P. Mohapatra, Soft-TDMAC: a software TDMA-based MAC over com- Telecommunications Conference (GLOBECOM 2011), 2011 IEEE, IEEE, 2011,
modity 802.11 hardware, in: INFOCOM, 2009, pp. 1836–1844. pp. 1–5.
[94] P. Djukic, P. Mohapatra, Soft-TDMAC: a software-based 802.11 overlay TDMA [123] S. Sorour, S. Valaee, Completion delay minimization for instantly decodable
MAC with microsecond synchronization, IEEE Trans. Mob. Comput. 11 (3) network codes, IEEE/ACM Trans. Netw. 23 (5) (2015) 1553–1567.
(2012) 478–491. [124] L. Hai, H. Wang, J. Wang, Instantly decodable network coding for multiple
[95] I. Molnar, Real-time patches for linux 2.6 Kernel, 2011, (http://www.kernel. unicast retransmissions in wireless point-to-multipoint networks, IEEE Trans.
org/pub/linux/kernel/projects/rt). Veh. Technol. 65 (8) (2016) 6232–6243.
[96] Q. Leng, Y.H. Wei, S. Han, A.K. Mok, W. Zhang, M. Tomizuka, Improving con- [125] S. Vitturi, L. Seno, F. Tramarin, M. Bertocco, On the rate adaptation techniques
trol performance by minimizing jitter in RT-WiFi networks, in: Real-Time Sys- of IEEE 802.11 networks for industrial applications, IEEE Trans. Ind. Inf. 9 (1)
tems Symposium, 2015, pp. 63–73. (2012) 198–208.
[97] W. Zhang, M. Tomizuka, Y.H. Wei, Q. Leng, S. Han, A.K. Mok, Time delay com- [126] F. Tramarin, S. Vitturi, M. Luvisotto, An innovative approach to rate adaptation
pensation in a wireless tracking control system with previewed reference, in: in IEEE 802.11 real-time industrial networks, in: IEEE World Conference on
American Control Conference, 2015, pp. 3293–3298. Factory Communication Systems, 2016, pp. 1–4.
[98] W. Zhang, M. Tomizuka, P. Wu, Y.H. Wei, Q. Leng, S. Han, A.K. Mok, A dou- [127] F. Tramarin, S. Vitturi, M. Luvisotto, A dynamic rate selection algorithm
ble disturbance observer design for compensation of unknown time delay in for IEEE 802.11 industrial wireless LAN, IEEE Trans. Ind. Inf. 13 (2) (2017)
a wireless motion control system, IEEE Trans. Control Syst. Technol. PP (99) 846–855.
(2017) 1–9. [128] C.Y. Li, C. Peng, S. Lu, X. Wang, R. Chandra, Latency-aware rate adaptation in
[99] Y. Cheng, D. Yang, H. Zhou, Det-WiFi: a multihop TDMA MAC implementation 802.11n home networks, in: Computer Communications, 2015, pp. 1293–1301.
for industrial deterministic applications based on commodity 802.11 hard- [129] R. Zhu, Intelligent rate control for supporting real-time traffic in WLAN mesh
ware, Wirel. Commun. Mob. Comput. 2017 (2017). networks, J. Netw. Comput. Appl. 34 (5) (2011) 1449–1458.
[130] H. Kirrmann, M. Hansson, P. Muri, IEC 62439 PRP: bumpless recovery for
66 Y. Cheng, D. Yang and H. Zhou et al. / Computer Networks 157 (2019) 41–67

highly available, hard real-time industrial networks, in: Emerging Tech- tion-a new method of increasing the granularity of traffic prioritization in
nologies and Factory Automation, 2007. ETFA. IEEE Conference on, 2007, WLANs, in: Computers and Communication (ISCC), 2014 IEEE Symposium on,
pp. 1396–1399. IEEE, 2014, pp. 1–6.
[131] G. Cena, S. Scanzio, A. Valenzano, Experimental characterization of redundant [161] F. Gringoli, P. Serrano, I. Ucar, N. Facchi, A. Azcorra, Experimental QoE evalua-
channels in industrial Wi-Fi networks, in: IEEE World Conference on Factory tion of multicast video delivery over IEEE 802.11 aa WLANs, IEEE Trans. Mob.
Communication Systems, 2016, pp. 1–4. Comput. (2018).
[132] L. Seno, G. Cena, S. Scanzio, A. Valenzano, C. Zunino, Enhancing communica- [162] A. Banchs, A. De la Oliva, L. Eznarriaga, D.R. Kowalski, P. Serrano, Performance
tion determinism in Wi-Fi networks for soft real-time industrial applications, analysis and algorithm selection for reliable multicast in IEEE 802.11 aa wire-
IEEE Trans. Ind. Inf. PP (99) (2016). 1–1 less LAN, IEEE Trans. Veh. Technol. 63 (8) (2014) 3875–3891.
[133] N. Aboutorab, P. Sadeghi, S. Sorour, Enabling a tradeoff between completion [163] M.K. Afzal, B.-S. Kim, S.W. Kim, Modified groupcast retries block acknowl-
time and decoding delay in instantly decodable network coded systems, IEEE edgement scheme in IEEE 802.11 aa standard-based for multimedia applica-
Trans. Commun. 62 (4) (2014) 1296–1309. tions, in: Signal Processing and Communication Systems (ICSPCS), 2014 8th
[134] H. Li, B. Li, T.T. Tran, D.C. Sicker, Transmission schemes for multicasting hard International Conference on, IEEE, 2014, pp. 1–5.
deadline constrained prioritized data in wireless multimedia streaming, IEEE [164] W.-P. Lai, E.-C. Liou, Virtual queue dropping for robust real-time video over
Trans. Wirel. Commun. 15 (3) (2016) 1631–1641. IEEE 802.11 aa wireless LANs, EURASIP J. Wirel. Commun. Netw. 2016 (1)
[135] Z. Dong, S.H. Dau, C. Yuen, Y. Gu, X. Wang, Delay minimization for relay-based (2016) 150.
cooperative data exchange with network coding, IEEE/ACM Trans. Netw. 23 [165] C. Chung, Y. Jung, J. Kim, Saturation throughput analysis of IEEE 802.11 ac
(6) (2015) 1890–1902. TXOP sharing mode, Electron. Lett. 51 (25) (2015) 2164–2166.
[136] X. Wang, C. Yuen, Y. Xu, Coding-based data broadcasting for time-critical [166] B. Bellalta, J. Barcelo, D. Staehle, A. Vinel, M. Oliver, On the performance of
applications with rate adaptation, IEEE Trans. Veh. Technol. 63 (5) (2014) packet aggregation in IEEE 802.11 ac MU-MIMO WLANs, IEEE Commun. Lett.
2429–2442. 16 (10) (2012) 1588–1591.
[137] C. Fragouli, D. Katabi, A. Markopoulou, M. Medard, H. Rahul, Wireless net- [167] M.A. Santos, J. Villalon, L. Orozco-Barbosa, Evaluation of the IEEE 802.11 aa
work coding: opportunities & challenges, in: Military Communications Con- group addressed service for robust audio-video streaming, in: Communica-
ference, 2007. MILCOM 2007. IEEE, IEEE, 2007, pp. 1–8. tions (ICC), 2012 IEEE International Conference on, IEEE, 2012, pp. 6879–6884.
[138] S. Katti, H. Rahul, W. Hu, D. Katabi, M. Médard, J. Crowcroft, XORs in the air: [168] Y. Zeng, Performance analysis of the 802.11 aa intra-access category prior-
practical wireless network coding, in: ACM SIGCOMM computer communica- itization under saturated condition, in: Global Communications Conference
tion review, 36, ACM, 2006, pp. 243–254. (GLOBECOM), 2014 IEEE, IEEE, 2014, pp. 4610–4615.
[139] I. Pefkianakis, S.H.Y. Wong, H. Yang, S.B. Lee, S. Lu, Toward history-aware ro- [169] L. Qihao, J. Lei, E. Frank, Performance evaluation of the GCR block ACK mech-
bust 802.11 rate adaptation, IEEE Trans. Mob. Comput. 12 (3) (2013) 502–515. anism in IEEE 802.11 aa networks, in: Wireless Conference (EW), Proceedings
[140] F. Schmidt, A. Hithnawi, O. Punal, J. Gross, A receiver-based 802.11 rate adap- of the 2013 19th European. VDE, 2013, pp. 1–7.
tation scheme with on-demand feedback, in: IEEE International Symposium [170] R.A. da Silva, A. Santos, M. Nogueira, K. Boussetta, N. Achir, Avoiding colli-
on Personal Indoor and Mobile Radio Communications, 2013, pp. 399–405. sions by time slot reduction supporting voice and video in 802.11 networks,
[141] F. Tramarin, S. Vitturi, M. Luvisotto, Improved rate adaptation strategies for in: Global Communications Conference (GLOBECOM), 2016 IEEE, IEEE, 2016,
real-time industrial IEEE 802.11n WLANs, in: IEEE Conference on Emerging pp. 1–6.
Technologies & Factory Automation, 2015, pp. 1–8. [171] R. Ali, N. Shahin, R. Bajracharya, B.-S. Kim, S.W. Kim, A self-scrutinized backoff
[142] D.A. Nugroho, M.R. Khaefi, D.S. Kim, Dynamic rate adaptation for industrial mechanism for IEEE 802.11 ax in 5G unlicensed networks, Sustainability 10
WLAN, in: International Conference on ICT Convergence, 2013, pp. 575–580. (4) (2018) 1201.
[143] F. Tramarin, S. Vitturi, M. Luvisotto, A. Zanella, On the use of IEEE 802.11n for [172] R.-S. Cheng, C.-M. Huang, Collision detect and avoidance media access mech-
industrial communications, IEEE Trans. Ind. Inf. 12 (99) (2016). 1–1 anism for next generation 802.11 ax networks, in: Heterogeneous Networking
[144] J.S. Blanes, 802.11n performance analysis for a real multimedia industrial ap- for Quality, Reliability, Security and Robustness (QSHINE), 2015 11th Interna-
plication, Comput. Ind. 66 (C) (2015) 31–40. tional Conference on, IEEE, 2015, pp. 189–193.
[145] S. Santonja-Climent, D. Todoli-Ferrandis, T. Albero-Albero, V.M. Sempere-Payá, [173] M.S. Afaqui, E.G. Villegas, E. Lopez-Aguilera, Dynamic sensitivity control al-
J. Silvestre-Blanes, J. Alcober, Analysis of control and multimedia real-time gorithm leveraging adaptive RTS/CTS for IEEE 802.11 ax., in: WCNC, 2016,
traffic over SIP and RTP on 802.11 n wireless links for utilities networks[C], pp. 1–6.
in: 2010 IEEE 15th Conference on Emerging Technologies & Factory Automa- [174] D. Bankov, A. Didenko, E. Khorov, V. Loginov, A. Lyakhov, IEEE 802.11 ax up-
tion (ETFA 2010), IEEE, 2010, pp. 1–4. link scheduler to minimize, delay: a classic problem with new constraints, in:
[146] N. Hajlaoui, I. Jabri, M. Taieb, M. Benjemaa, A frame aggregation scheduler Personal, Indoor, and Mobile Radio Communications (PIMRC), 2017 IEEE 28th
for QoS-sensitive applications in IEEE 802.11n WLANs, in: International Con- Annual International Symposium on, IEEE, 2017, pp. 1–5.
ference on Communications and Information Technology, 2012, pp. 221–226. [175] J. Lee, OFDMA-based hybrid channel access for IEEE 802.11 ax WLAN, in: 2018
[147] G. Bhanage, D. Raychaudhuri, I. Seskar, Backlogged queue based MAC frame 14th International Wireless Communications & Mobile Computing Conference
aggregation, Pervasive Mob. Comput. 7 (4) (2011) 449–466. (IWCMC), IEEE, 2018, pp. 188–193.
[148] B. Maqhat, M.D. Baba, R.A. Rahman, A-MSDU real time traffic scheduler for [176] D.-J. Deng, Y.-P. Lin, X. Yang, J. Zhu, Y.-B. Li, J. Luo, K.-C. Chen, IEEE 802.11 ax:
IEEE 802.11n WLANs, 2016, pp. 286–290. highly efficient WLANs for intelligent information infrastructure, IEEE Com-
[149] A. Saif, M. Othman, SRA-MSDU: enhanced A-MSDU frame aggregation with mun. Mag. 55 (12) (2017) 52–59.
selective retransmission in 802.11n wireless networks, J. Netw. Comput. Appl. [177] E. Khorov, A. Kiryanov, A. Lyakhov, G. Bianchi, A tutorial on IEEE 802.11 ax
36 (4) (2013) 1219–1229. high efficiency WLANs, IEEE Commun. Surv. Tutorials (2018).
[150] D. Camps-Mur, M.D. Gomony, X. Prez-Costa, S. Sallent-Ribes, Leveraging [178] R. Ali, S.W. Kim, B.-S. Kim, Y. Park, Design of MAC layer resource allocation
802.11n frame aggregation to enhance QoS and power consumption in Wi-Fi schemes for IEEE 802.11 ax: future directions, IETE Technical Review 35 (1)
networks, Comput. Netw. 56 (12) (2012) 2896–2911. (2018) 28–52.
[151] E. Charfi, L.C. Fourati, L. Kamoun, QoS support of voice/video services under [179] G. Hiertz, F. Mestanov, S. Coffey, Enlarged Minimal Contention Window Size,
IEEE 802.11n WLANs, in: International Symposium on Communication Sys- doc: IEEE 802.11-15/0914r1, Tech. Rep., 07 2015, 2015.
tems, Networks & Digital Signal Processing, 2014, pp. 600–605. [180] C. Ghosh, R. Stacey, E. Perahia, S. Azizi, P. Huang, Q. Li, L. Cariou, X. Chen,
[152] C.-Y. Chang, H.-C. Yen, C.-C. Lin, D.-J. Deng, QoS/QoE support for H. 264/AVC R. Yang, Y. Alpert, et al., UL OFDMA-based random access procedure, IEEE
video stream in IEEE 802.11 ac WLANs, IEEE Syst. J. 11 (4) (2017) 2546–2555. 802 (2015) 1–19.
[153] O. Sharon, Y. Alpert, The combination of QoS, aggregation and RTS/CTS in very [181] X. Zhang, K.L. Yeung, LLE: A timer extension mechanism for alarm-triggered
high throughput IEEE 802.11 ac networks, Phys. Commun. 15 (2015) 25–45. traffic in IEEE 802.11ah WLANs, in: ICC 2017 - 2017 IEEE International Con-
[154] S. Jang, S. Bahk, A channel allocation algorithm for reducing the channel sens- ference on Communications, 2017, pp. 1–6.
ing/reserving asymmetry in 802.11 ac networks, IEEE Trans. Mob. Comput. 14 [182] Y. Wang, K.K. Chai, Y. Chen, J. Schormans, J. Loo, Energy-delay aware restricted
(3) (2015) 458–472. access window with novel retransmission for IEEE 802.11ah networks, in:
[155] T. Moriyama, R. Yamamoto, S. Ohzahata, T. Kato, Frame aggregation size deter- Global Communications Conference, 2017, pp. 1–6.
mination for IEEE 802.11 ac WLAN considering channel utilization and trans- [183] L. Tian, E. Khorov, S. Latr, J. Famaey, Real-time station grouping under dy-
fer delay, STA 4 (s3) (2017) s4. namic traffic for IEEE 802.11ah, Sensors 17 (7) (2017).
[156] E. Charfi, L. Chaari, S.B. Hlima, L. Kammoun, Multi-user access mechanism [184] G.C. Madueo, edomir Stefanovi, P. Popovski, Reliable and efficient access for
with intra-access categories differentiation for IEEE 802.11 ac wireless local alarm-initiated and regular M2M traffic in IEEE 802.11ah systems, IEEE Inter-
area networks, Telecommun. Syst. 64 (3) (2017) 479–494. net Things J. 3 (5) (2016) 673–682.
[157] I. Arun, T. Venkatesh, B. Dappuri, QoS provisioning with adaptive backoff al- [185] N.F. Charania, M.K. Giluka, B.R. Tamma, A. Franklin, DEARF: Delay and En-
gorithm for IEEE 802.11 ac under multipacket reception, in: Communication ergy Aware RAW Formation Scheme to Support Delay Sensitive M2M Traffic
Systems, Networks & Digital Signal Processing (CSNDSP), 2014 9th Interna- in IEEE 802.11Ah Networks, 2017.
tional Symposium on, IEEE, 2014, pp. 826–830. [186] T. Adame, A. Bel, B. Bellalta, J. Barcelo, IEEE 802.11ah: the WiFi approach for
[158] C.-H. Liu, D. Cabric, Traffic-aware backoff design for IEEE 802.11 (ac), in: Com- M2M communications, IEEE Wirel. Commun. 21 (6) (2014) 144–152.
puting, Networking and Communications (ICNC), 2016 International Confer- [187] E. Khorov, A. Lyakhov, A. Krotov, A. Guschin, A survey on IEEE 802.11ah: an
ence on, IEEE, 2016, pp. 1–6. enabling networking technology for smart cities, Comput. Commun. 58 (11)
[159] K. Kosek-Szott, M. Natkaniec, L. Prasnal, A novel IEEE 802.11 aa intra-AC pri- (2015) 53–69.
oritization method for video transmissions, in: Global Communications Con- [188] C.W. Park, D. Hwang, T.J. Lee, Enhancement of IEEE 802.11ah MAC for M2M
ference (GLOBECOM), 2014 IEEE, IEEE, 2014, pp. 1158–1163. communications, IEEE Commun. Lett. 18 (7) (2014) 1151–1154.
[160] K. Kosek-Szott, M. Natkaniec, L. Prasnal, IEEE 802.11 aa intra-AC prioritiza- [189] M. Dong, Z. Wu, X. Gao, H. Zhao, An efficient spatial group restricted access
 Y. Cheng, D. Yang and H. Zhou et al. / Computer Networks 157 (2019) 41–67 67

Dong Yang (M’11) received his B.S. degree from Central

window scheme for IEEE 802.11ah networks, in: Sixth International Confer-
South University, Hunan, China, in 2003 and Ph.D. degrees
ence on Information Science and Technology, 2016, pp. 168–173.
in communications and information science from Bei-
[190] A. Bel, T. Adame, B. Bellalta, J. Barcelo, J. Gonzalez, M. Oliver, CAS-based chan-
jing Jiaotong University, Beijing, China, 2009. From March
nel access protocol for IEEE 802.11ah WLANs, in: European Wireless 2014;
2009 to June 2010, he was a Post-Doctoral Research As-
European Wireless Conference; Proceedings of, 2014, pp. 1–6.
sociate with Jonkoping University, Jonkoping, Sweden. In
[191] J.O. Seo, C. Nam, S.G. Yoon, S. Bahk, Group-based contention in IEEE 802.11ah
August 2010, he joined the School of Electronic and In-
networks, in: International Conference on Information and Communication
formation Engineering, Beijing Jiaotong University. Cur-
Technology Convergence, 2014, pp. 709–710.
rently, he is a professor and his research interests are net-
[192] E. Khorov, A. Krotov, A. Lyakhov, Modelling machine type communication in
work technologies, including Internet architecture, indus-
IEEE 802.11ah networks, in: IEEE International Conference on Communication
trial Internet and wireless sensor networks.
Workshop, 2015, pp. 1149–1154.
[193] L. Zheng, L. Cai, J. Pan, M. Ni, Performance analysis of grouping strategy for
dense IEEE 802.11 networks, in: Global Communications Conference, 2014,
pp. 219–224. Huachun Zhou received the B.S. degree from the Peoples
[194] O. Raeesi, J. Pirskanen, A. Hazmi, T. Levanen, Performance evaluation of IEEE Police Officer University of China in 1986. He received the
802.11ah and its restricted access window mechanism, in: IEEE International M.S. in telecommunication automation and Ph.D. degrees
Conference on Communications Workshops, 2014, pp. 460–466. in telecommunications and information system from Bei-
[195] B. Badihi, L.F.D. Carpio, P. Amin, A. Larmo, M. Lopez, D. Denteneer, Perfor- jing Jiaotong University in 1989 and 2008, respectively.
mance evaluation of IEEE 802.11ah actuators, in: Vehicular Technology Con- In October 1994, he joined Institute of Automation Sys-
ference, 2016, pp. 1–5. tems, BJTU, where he is a lecturer. From Apr. 1999 to Sep.
[196] Z. Yang, J. Zhang, K. Tan, Q. Zhang, Y. Zhang, Enabling TDMA for today’s wire- 2009, he was a senior engineer at School of Electronic
less LANs, in: Computer Communications (INFOCOM), 2015 IEEE Conference and Information Engineering, BJTU, and at Network Man-
on, IEEE, 2015, pp. 1436–1444. agement Research Center, BJTU. From Oct. 2009 to now,
[197] J. Lee, M. Uddin, J. Tourrilhes, S. Sen, S. Banerjee, M. Arndt, K.-H. Kim, he is a professor in National Engineering Lab for Next
T. Nadeem, meSDN: Mobile extension of SDN, in: Proceedings of the Fifth Generation Internet Interconnection Devices at BJTU. He
International Workshop on Mobile Cloud Computing & Services, ACM, 2014, has authored more than 40 peer-reviewed papers and he
pp. 7–14. is the holder of 17 patents. His main research interests are in the area of mobil-
[198] K. Wang, T.Y. Chai, W.-C. Wong, Routing, power control and rate adaptation: ity management, mobile and secure computing, routing protocols, network man-
a Q-learning-based cross-layer design, Comput. Netw. 102 (2016) 20–37. agement and satellite network.
[199] E. Coronado, J. Villalón, A. Garrido, Ensuring QoS for IEEE 802.11 real-time
communications using an AIFSN prediction scheme, in: Security, Privacy and Hongchao Wang (M’14)received his B.S. degree in com-
Reliability in Computer Communications and Networks, 2016, p. 365. munication engineering from Beijing Jiaotong University,
[200] A.D. Shoaei, M. Derakhshani, S. Parsaeefard, T. Le-Ngoc, Learning-based hy- Beijing, China, in 2005 and Ph.D. degree in communica-
brid TDMA-CSMA MAC protocol for virtualized 802.11 WLANs, in: Personal, tion and information system from Beijing Jiaotong Uni-
Indoor, and Mobile Radio Communications (PIMRC), 2015 IEEE 26th Annual versity, Beijing, China, in 2012.From July 2012 to Au-
International Symposium on, IEEE, 2015, pp. 1861–1866. gust 2014, he was a Post-Doctoral Research Fellow in the
[201] S. Jang, K.G. Shin, S. Bahk, Post-CCA and reinforcement learning based band- School of Computer Science and technology, Beijing Jiao-
width adaptation in 802.11 ac networks, IEEE Trans. Mob. Comput. 17 (2) tong University. In August 2014, he joined the School of
(2018) 419–432. Electronic and Information Engineering, Beijing Jiaotong
University. His research interests are Internet architecture,
Yujun Cheng received the B.S. degree in communication network security and wireless sensor networks.
engineering from Beijing Jiao University, Beijing, China,
in 2014. Currently, he is working toward the PhD de-
gree in communications and information science from
Beijing Jiaotong University, Beijing, China. His specific ar-
eas of research interest mainly focus on wireless medium
access control protocols,industrial wireless networks,and
real-time computing.