Artery-C

An OMNeT++ Based Discrete Event Simulation

Framework for Cellular V2X
Extended Version∗
Anupama Hegde Andreas Festag†
Technische Hochschule Ingolstadt / CARISSMA Technische Hochschule Ingolstadt / CARISSMA
arXiv:2009.05724v1 [cs.NI] 12 Sep 2020

[email protected] [email protected]

Abstract

Cellular Vehicle-to-X (Cellular V2X) is a communication technology that aims to facilitate the communication among
vehicles and with the roadside infrastructure. Introduced with LTE Release 14, Cellular V2X enables device-to-device
communication to support road safety and traffic efficiency applications. We present Artery-C, a simulation framework
for the performance evaluation of Cellular V2X protocols and V2X applications. Our simulator relies on the simulation
framework SimuLTE and substantially extends it by implementing control and user planes. Besides the vehicle-
to-network communication via the up-/downlink interface, it provides vehicle-to-vehicle and vehicle-infrastructure
communication via the sidelink interface using the managed and the unmanaged mode of Cellular V2X (mode 3 and 4,
respectively). The simulator also implements advanced features of 5G mobile networks, such as variable numerologies.
For the transmission of of V2X messages, it adds a non-IP interface. Artery-C integrates seamlessly into the simulation
framework Artery, which enables the simulation of standardized V2X messages at the facilities layer as well as the
coupling to the mobility simulator SUMO. A specific feature of Artery-C is the support of dynamic switching between
all modes of Cellular V2X. In order to demonstrate the capabilities of Artery-C, we evaluate V2X-based platooning as a
representative use case and present results for mode 3, mode 4 and mode switching in a highway scenario.

I. Introduction ditionally realizes a direct communication among

end devices, i.e., vehicle-to-vehicle (V2V), vehicle-to-
2X communication enables the exchange of pedestrian (V2P) and vehicle-to-infrastructure (V2I)

V information among vehicles, roadside infras-

tructure and other traffic participants. The
continuous information exchange supports vehicles
communication. The direct communication, also re-
ferred to as device-to-device (D2D) communication,
allows two physically close end devices to commu-
to obtain an accurate knowledge of its surround- nicate using the sidelink interface (LTE PC-5). By
ing environment in order to improve traffic safety sidelink, devices do not have to depend on the cel-
and efficiency. Cellular V2X is a mobile network- lular access and core network for allocation of radio
based communication technology that facilitates the resources for data transmission [1].
conventional communication between vehicle and In comparison to the cellular communication via
network (V2N) to provide backend services and ad- up- and downlink, the sidelink communication in-
∗ A short version of the paper appeared in: A. Hegde and curs a shorter latency for message transfer, which is
A. Festag “Artery-C – An OMNeT++ Based Discrete Event Sim- a critical aspect for vehicle safety and automation.
ulation Framework for Cellular V2X”, 23rd International ACM The sidelink communication enables the transmis-
Conference on Modeling, Analysis and Simulation of Wireless sion of periodic and non-periodic V2X messages
and Mobile Systems (MSWiM ’20), November 16–20, 2020, Ali-
as they are defined, amongst other standards, in
cante, Spain, DOI: 10.1145/3416010.3423240
† Also with Fraunhofer Application Center “Connected Mo- the European standards for V2X communications,
bility and Infrastructure” most importantly the Cooperative Awareness Mes-

1
A. Hegde and A. Festag

Table 1: Comparison of existing simulation frameworks for Cellular V2X with ARTERY-C

SimuLTE [2] Cellular- OpenCV2X [4] Artery-C

VCS [3]
Base framework OMNeT++ ns-3 OMNeT++ OMNeT++
Protocol stack LTE LTE LTE LTE, 5G sel. features
Plane User User User Control & User
Modes UL, DL, D2Da UL, DL, SL UL, DL, SL UL, DL, SL
mode 3 mode 4 mode 3 & 4
Mode switching Cellular – D2D No No Cellular – D2D,
SL mode 3 – 4
Variable No No No Yes
numerology
V2X applications IP-based V2I, IP-based V2I, Non-IP-based IP-based V2I, V2N;
V2N V2N, V2V mode 4 Non-IP-based V2V
V2V mode 3 (mode 3 & 4)
Facilities No No Yes, with Yes, with Artery
Artery [5]
Open source Yes No Yes Plannedb
a Abbreviations: UL = Uplink, DL = Downlink, D2D = Device-to-Device, SL = Sidelink
b We plan to publish the simulation framework under an open source license.

sages (CAM) and Decentralized Environmental No- framework SimuLTE [2], implemented the control
tification Message (DENM). With CAMs, vehicles plane functions, and integrated both user and con-
periodically broadcast their status information such trol plane into Artery, resulting in Artery-C.1
as position, speed, heading to neighboring vehi- In Artery-C, the modules for radio resource allo-
cles with a periodicity of 1 to 10 Hz. DENMs are cation take into account that a vehicle can be located
triggered in critical safety situations. With the avail- in the region of cellular coverage and that the allo-
ability of up-/down- and sidelink in Cellular V2X, cation process is managed by an LTE base station
these periodic and event-driven messages can be in a centralized manner (referred to as network-
handled across the LTE Uu and PC-5 interfaces. assisted or mode 3 in 3GPP standards [6]). In case
One of the key challenges in modelling vehic- the vehicle cannot remain in the region of cellu-
ular communications for simulation-based perfor- lar coverage, it autonomously configures the ra-
mance evaluation is the required integration of a dio resources from a pre-defined pool of resources
diverse set of components, including vehicle mo- (defined as mode 4 in 3GPP standards). Hence,
bility, environmental perception, radio propagation Artery-C supports three modes in a common simu-
and related effects as well as V2X services and com- lation framework, i.e., up-/downlink with the RAN
munication protocols. The existing OMNeT++-based and the EPC, network-assisted sidelink (mode 3)
simulation framework Artery [5] – originally de- and out-of-coverage sidelink with distributed re-
veloped for standard-compliant, WLAN-based V2X source allocation and management (mode 4). As an
communication – provides a clear separation of fa- additional feature to the simultaneous support of all
cilities, application layer and vehicular scenarios, three modes, Artery-C supports dynamic switch-
which makes it an ideal base framework for Cel- ing among the modes as a part of the simulation
lular V2X simulations. To model the data plane scenario. For example, a scenario may involve a
functionalities of the LTE Radio Access Network change from mode 3 to 4, when a vehicle moves
(RAN) and Evolved Packet Core (EPC), we have uti-
lized and extended the user plane of the simulation 1 The -C in Artery-C stands for Cellular V2X.

2
 Artery-C – An OMNeT++ Based Discrete Event Simulation Framework for Cellular V2X

out-of-coverage and switch back to mode 3 when comprehensive Cellular-V2X protocol suite with
network coverage resumes. The same vehicle may dedicated control and user planes, with a sidelink
communicate simultaneously via the up-/downlink. interface for both mode 3 and mode 4, and with
Hence, the support of mode switching allows mod- support of different V2X application scenarios, we
eling more complex scenarios and studying V2X have identified several requirements.
applications under more realistic conditions.
For performance evaluation of Cellular V2X,
several simulation frameworks exist (see Tab. 1). i. Software-related Requirements
A baseline for sidelink in LTE networks has Modularity: The layers of the protocol suite are
been developed in the OMNeT++-based simulator defined as modules, which communicate with each
SimuLTE [2] and extended for network assisted other through messages sent across gates. The lay-
device-to-device (D2D) [7]. In [3], the authors stud- ers of the user plane – Packet Data and Conver-
ied Cellular V2X mode 3 using the ns-3 framework. gence Protocol (PDCP), Radio Link Control (RLC),
The OpenCV2X simulator in [4] aims to model and Medium Access Control (MAC) and Physical (PHY)
evaluate the performance of sidelink mode 4 in Cel- – are implemented as simple modules and encap-
lular V2X. However, the existing simulators so far sulated into a compound module called Network
assume only a single resource allocation mode at a Interface Card (NIC) as shown in Fig. 1. The control
given time for a given scenario. This leads to a lim- plane (RRC) is modeled as an independent mod-
itation that heterogeneous V2X scenarios with V2V, ule, which communicates with the user plane by a
V2I and V2N cannot be simultaneously studied. message-passing paradigm.
The remainder of the paper is organized as fol- Separation between protocol stack and road
lows: Sec. II describes the requirements on a Cellu- traffic model: Traffic models to study different
lar V2X simulation environment. Sec. III explains types of V2X scenarios such as V2V, V2I, V2N and
the implementation design of ARTERY-C. Sec. IV V2P are developed using the microscopic road traf-
presents simulation results for an example V2X use fic simulator SUMO2 . The Cellular V2X protocol
case using different capabilities of ARTERY-C for stack is implemented separately in the OMNeT++ sim-
the purpose of validation. For the use case, we ulation framework and specifically used to study
have chosen V2X-based platooning and study it in relevant aspects in V2X communication environ-
a highway scenario for mode 3, mode 4 and mode ments, such as resource allocation and scheduling.
switching. Sec. V concludes the paper. Message formats for V2X communication ser-
vices: Currently, the framework Artery-C includes
II. Requirements for a Cellular V2X the V2X message types CAM and DENM. The
simulation environment framework aims to support various other message
formats for V2X use cases such as infrastructure
The VANET simulator Artery [5] provides a com- messages, sensor data sharing and maneuver coor-
prehensive framework with a clear separation of dination [9].
the protocol stack and the environment model. It
allows for a smooth interaction between OMNeT++
and SUMO, and adapting the facilities layer to dif-
ii. Cellular V2X-specific Requirements
ferent access technologies. The Artery middleware Dedicated sidelink interface: In order to facilitate
enables vehicles to use multiple V2X services simul- an uninterrupted exchange of messages among ve-
taneously. hicles, infrastructure and road traffic participants, a
Originally developed for ITS-G5 type of access dedicated sidelink (PC-5) communication interface
technologies, the recent version of Artery supports is implemented. This interface co-exists with the
mobile networks, specifically up-/downlink and up- and downlink (Uu ) interfaces. The vehicle com-
network-assisted D2D communication [8]. A first municates with the infrastructure (V2N/V2I) using
approach to extend Artery for sidelink mode 4 has
been addressed in [4]. For the development of a 2 https://sumo.dlr.de (retrieved Sep 10, 2020)

3
A. Hegde and A. Festag

the up-/downlink and the sidelink is used for V2V, consumed processing time for simulation execu-
V2I and V2P applications. tion. The simulator allows achieving statistically
Sidelink resource allocation modes and dy- meaningful results within a reasonable time and
namic switching: Both sidelink resource allocation commodity computing resources.
modes, i.e., mode 3 and mode 4, should be sup- In order to meet the Cellular V2X-specific require-
ported so that vehicles can exchange messages de- ments and realize the software- and timing-related
pending on whether they are located inside or out- aspects, we have implemented the simulation frame-
side the coverage of base station. Dynamic switch- work Artery-C.
ing between sidelink resource allocation modes
causes overhead in terms of mode switching la-
tency [10]. III. Implementation design of
Support of heterogeneous traffic: Depending on ARTERY-C
the type of application, the framework enables ve-
hicles to simultaneously send both IP-based and In the OMNeT++ framework, the basic implementa-
non-IP-based data. tion unit is called a module. Modules communi-
Resource allocation and scheduling: Conven- cate with each other through event-driven messages.
tional scheduling schemes such as Round Robin Each module is characterized by a structure defined
(RR), Deficit Round Robin (DRR), MAXimum Car- via .ned files and a behavior implemented via C++
rier over Interference (MAXCI) and Proportional classes. The modules in a network can be of two
Fair queuing (PF) are already implemented in the types – stationary and dynamic. In this section, we
user plane of SimuLTE for uplink and downlink. Fol- present the salient features of our simulation frame-
lowing 3GPP standards, the sidelink uses sensing- work Artery-C that has been built as an extension
based semi-persistent scheduling (SB-SPS) for both to the user plane implementation in the simulation
mode 3 and mode 4 (see Sec. III). framework SimuLTE [2, 7].
The implementation design of the Artery-C
iii. Timing-related Requirements framework consists of two layers. The lower layer
covers the protocols of the Cellular V2X access tech-
Transmit time interval (TTI): The smallest unit nology. The implementation is aligned with the
step time for the protocol simulations is 1 ms, which Cellular V2X protocol stack. It comprises the con-
corresponds to the smallest time unit size for re- trol plane with RRC and the user plane with PHY,
source allocation in LTE. Furthermore, 5G-New Ra- MAC, RLC and PDCP. The layer on top is for gener-
dio (NR)(3GPP TS 38.300 V16.1.0) supports a flex- ation and reception of V2X messages and represents
ible size of resource units in frequency and time, the facilities of the C-ITS protocol stack.
which requires variable numerologies in the simu- The user plane has two parallel pipelines for IP-
lator. based and non-IP-based traffic (Fig. 1) with clearly
Control and user plane latency: Control plane separated functionalities. The facilities layer real-
latency is characterized by delays incurred due to izes the V2X messaging, such as the CAM service
communication between the control plane compo- modules of Artery to generate non-IP-based peri-
nents responsible for registering the end device odic messages. For IP-based V2I traffic, Artery-C
with the infrastructure and acquiring system in- re-utilizes the modules from INET and Artery. The
formation (SI). Correspondingly, the user plane la- user plane is directly linked with the SUMO traf-
tency is caused by the communication delays be- fic model via the TRaCI API in order to continu-
tween the layers of the user plane. In order to study ously track the position of the vehicle and identifies
packet end-to-end latency for various applications, whether it is located in the region of cellular cov-
resource allocation during mode switching etc, it is erage or not. The LTE base stations (eNodeB) and
important to understand the impact of both control road side units (RSUs) are modelled as stationary
and user plane latency. modules and vehicles are modelled as dynamic
Simulation run time: It refers to the amount of modules.

4
 Artery-C – An OMNeT++ Based Discrete Event Simulation Framework for Cellular V2X

the eNodeB. The eNodeB responds with the appro-

priate Master Information Block (MIB) and System
Information Block (SIB) for different communica-
tion interfaces, i.e., UL, DL and SL. Additionally,
the eNodeB pre-configures a set of time and fre-
quency resources in the form of SIB that can be
used by the UE when it operates in mode 4. The
synchronization-related information [10] between
UE and eNodeB are exchanged immediately after
the establishment of successful connection.
Mode switching control: The mode switching
control module is responsible to regulate the dy-
namic switching between mode 3 and mode 4. This
is based on the availability of network coverage, re-
ceived signal strength and traffic load. The resource
allocation modes directly correlate with the oper-
ating states of the RRC - RRC IDLE , RRC I N ACTIVE
and RRCCONN and correspond to the three-state
finite state machine (FSM) as shown in Fig. 2.
Figure 1: Implementation design of the Cellular V2X stack

i. Control Plane – Radio Resource Control

Cell search and mode selection: The control
plane/Radio Resource Control (RRC) component is
modelled in accordance with the 3GPP Release 15 Figure 2: RRC state transition
standards for 5G-NR (TS 38.331, V15.7.0). The
RRC is responsible to carry out three primary func- In the RRCCONN state, the UE is connected to
tionalities, i.e., system information acquisition, cell the eNodeB and has established its identity with
search and mode selection & mode switching con- the EPC. The eNodeB is responsible for allocating
trol. Based on the position updates of the vehicle subchannels and subframes for sidelink communi-
from the SUMO traffic scenario, the cell search mod- cation. When an UE is inside the network coverage
ule of the RRC determines the distance between an but not exchanging any information with the eN-
UE and an LTE base station (eNodeB). If the UE odeB, it transits to RRC I N ACTIVE state. In this state,
is located inside the communication range of the it is still registered with eNodeB and hence con-
eNodeB and the received signal strength meets the tinues to function in mode 3. When an UE moves
threshold limits, then mode 3 is the preferred mode outside the base station coverage, the RRC tran-
of operation. If an UE lies outside the region of a sits to RRC IDLE state and it informs the PHY and
base station’s communication range, then mode 4 is MAC component to allocate resources from a pre-
selected. configured pool of resources mentioned in appro-
System information acquisition: The process priate SIBs. The UE is now completely disconnected
of system information acquisition involves the ex- from the base station.
change of network-related messages between UE
and eNodeB that enables the UE to establish a suc-
ii. User/Data Plane
cessful connection with the eNodeB and the com-
ponents of the Evolved Packet Core (EPC). When Packet data and convergence protocol (PDCP):
an UE recognizes itself to be in the region of cel- The PDCP is the connecting component between the
lular coverage, it sends a RRCConnSetup request to Cellular V2X access technology and the networking

5
A. Hegde and A. Festag

& transport layer. It processes both IP and non-IP tional block diagram of the MAC scheduling mod-
packets. In case of IP traffic, it performs Robust ule is depicted in Fig. 3. The adaptive modulation
Header Compression (ROHC) and assigns/creates and coding (AMC) module stores channel status in-
the Connection Identifier (CID) that uniquely iden- formation, which is based on the periodic feedback
tifies, together with the UE ID, a connection in the from the UE via up-/downlink. For sidelink op-
whole network. When an IP packet arrives at PDCP, eration, the chosen modulation scheme is fixed as
a logical connection identifier (LCID) is attached to quadrature phase shift keying (QPSK). In addition
it and forwarded to the radio link control (RLC). In to the existing scheduling policies, MAXCI, PF and
case of a non-IP-based packet3 the PDCP performs DRR, we have implemented sensing-based semi
ROHC, creates an entry in the non-IP connections persistent scheduling (SB-SPS) module for sidelink
table and forwards the packet as a PDU to the RLC. scheduling.
The two pipelines are illustrated in Fig. 1. Mode 3 operation The eNodeB has complete
The PDCP consists of two separate gates for data knowledge about the registered UEs and the re-
input from IP and non-IP traffic. Additionally, it sources utilized by them. In the existing UL and DL
has a separate gate to receive control-related mes- implementation, the eNodeB creates a scheduling
sages from the control plane. For interaction with list and provides the set of available resources in
the RLC, similar to [2], three different gates are con- every TTI. When an UE wants to transmit data on
nected with the PDCP-RRC module, one for each the sidelink, based on it’s geographical location, it
RLC mode. connects to the nearby eNodeB and requests for
Radio link control (RLC): The RLC oper- time and frequency resources. The UE reports to
ates in three modes - acknowledged (AM), un- the eNodeB about the size of the data, periodicity
acknowledged (UM) and transparent mode (TM). and maximum allowed latency based on the type
The key functionality of this component is to mul- of V2X application. The “sidelink configuration”
tiplex and de-multiplex MAC SDUs to/from the submodule in the MAC of the eNodeB configures
MAC. The implementation of the RLC has not been the sidelink grant and requests the PHY compo-
modified much in reference to [2]. For sidelink nent to allocate the candidate resource pools (CSRs)
broadcast operation, an acknowledgement is not and generates a sidelink control information (SCI)
applied and we use the un-acknowledged mode. message.

iii. Medium Access Control (MAC)

The design of the MAC for the UE module has
been modified in a way that the sidelink scheduling
co-exists with the previous uplink and downlink
implementations as done in SimuLTE [2, 7]. Addi-
tionally, the MAC module has a separate gate to
receive control-related information from the control
plane and incorporates an additional sub-module
called sidelink configuration (SC). The sidelink con-
figuration sub-module interacts with the sidelink
resource allocation (SRA) submodule in the PHY Figure 3: MAC scheduler
component and together they support the process of
On receiving the CSRs from the PHY compo-
resource configuration and allocation [10]. A func-
nent, the MAC of the eNodeB updates the resource-
3 In the European C-ITS standards, non-IP packet transport related information in the sidelink grant and pro-
is realized by GeoNetworking (ETSI EN 302 636-4), an ad hoc ceeds with scheduling of resource for sidelink.
network protocol based on geographic positions, and BTP , an
Mode 4 operation: In mode 4 operation, the
UDP-like transport protocol (ETSI EN 302 636-5). Both are
implemented in the simulation framework Artery but beyond MAC of the UE handles the scheduling of resources
the scope of this paper. independently of the base station. The channel-

6
 Artery-C – An OMNeT++ Based Discrete Event Simulation Framework for Cellular V2X

related cost metrics such as signal-to-noise-ratio taining the channel-related parameters from the
(SINR), the reference signal received power (RSRP) channel modules.
and the received signal strength indicator (RSSI)
are computed in the PHY component, which play a
key role in determining the list of CSRs. On receiv-
ing the pool of CSRs from the PHY component, it
further handles the SPS information and schedules
time-frequency resources for both data and SCI.
The flow of packets in MAC buffers is shown in
Fig. 4. On obtaining the service data unit (SDU)
from the RLC, they are stored inside MAC buffers. Figure 5: Physical (PHY) component
Based on the scheduling list generated by the sched-
uler, MAC protocol data units (PDU) are created Sidelink resource allocation (SRA): The sidelink
and stored in HARQ buffers. There are separate control information (SCI) is a 32 bit sequence that
transmission and reception HARQ buffers to store is transmitted prior to the transmission of transmit
MAC PDUs that are sent and received. The HARQ block (TB). In both mode 3 and mode 4, the SCI is
buffers in eNodeB contain MAC PDU information transmitted over two resource blocks in the same
for each of its connected UEs in both uplink and subframe as TB. The TB carries the payload data,
downlink. which are generated in the facilities layer. The time
resources are allocated in the form of subframes
and the frequency resources are characterized by
subchannels, which comprise a group of physical
resource blocks (PRBs). The number of subchannels
(NsubCH ) and size of subchannels (i.e., number of
PRBs per subchannel) (NPRB ) can vary in a certain
range as specified in [6].
The set of subframes belonging to PSSCH pool
(mode 3 & mode 4) is denoted by [t0SL , t1SL , · · · tSL
max ].
This pool includes all subframes except the sub-
frames where sidelink synchronization signals
(SLSS) is transmitted. Synchronization subframes
occur periodically at every 160 ms.
Figure 4: Flow of packets in the MAC component The frequency resource pool consists of a set of
subchannels (NsubCH ) comprising of contiguously
allocated resource blocks (NPRB ). The SCI and TB
can be transmitted in adjacent or non-adjacent re-
iv. Physical (PHY) component
source blocks of the same subframe. If SCI and TB
The PHY module contains information about an- are transmitted on adjacent resource blocks, the sub-
tenna power characteristics, standard channel mod- channel m comprises a set of contiguous resource
els and cell related information (macro, micro and blocks calculated as
pico cells). The functional components of the PHY
module are depicted in Fig. 5. In the context of
our work, the key functionality of the PHY com- n PRB = nsubCHRBStart + m ∗ nsubCHsize + j + β (1)
ponent is to allocate a set of candidate resource
pools (CSRs) for sidelink broadcast communication where m = 0, 1, · · · NsubCH −1 and j =
through SB-SPS. The “sidelink resource allocation” 0, 1, · · · nsubCHsize − 1. The starting index of sub-
sub-module is responsible for computing the CSRs channel, nsubCHRBStart is indicated by the higher
by utilizing the cell-related information and ob- layer components. Here the value of β is 2.

7
A. Hegde and A. Festag

Selection of candidate single-subframe re- Eq. 2 where there is an integer j that meets
source (CSR) pool: The PHY module determines
0
the CSRs based on sensing-based semi persistent y + j ∗ Prsvp−TX = z + Pstep ∗ k (2)
scheduling (SB-SPS). In mode 3, the procedure is
carried out by the eNodeB in a centralized man- z is a subframe in the sensing window, the
ner and in mode 4, it is carried out by the UE in a resource reservation interval given by higher
0
distributed way. The time interval between the gen- layers is Prsvp−TX = Prsvp−TX ∗ Pstep /100, j =
eration of a packet Tp and the maximum allowed 0, 1, · · · Cresel − 1. If the RSRP value measured over
latency TL is known as the selection window. The any resource element in the list L1 exceeds a given
TTIs and the resource blocks (RBs) consist of sub- threshold, it indicates that another UE is currently
carrier groups in a time-frequency grid as depicted using it for its transmission.
in Fig. 6. The minimum number of RBs needed to Note that the number of resources in L2 must
transmit an SCI is two [6, 11]. The number of RBs contain at least 20 % of resources in the selection
for data transmission varies depending on the size window. Otherwise, we increase the RSRP thresh-
of the transmit data block (TB). old by 3 dB and iterate again. From the list L2 , Veh-1
ranks the resource elements in increasing order of
their RSSI values and identifies the ones with low
RSSI values thereby creating a new list L3 . These re-
source elements are preferred because a low value
of RSSI indicates that it has not been used by any
other UE during the TTI of our interest. From
this list L3 , Veh-1 autonomously chooses any of the
available resources and uses it for transmission.
Mode 3 operation: In mode 3, the subframes and
subchannels for transmission of SCI and TB are
allocated by the eNodeB. From the standards per-
spective, Eq. (2) is adapted as

0
Figure 6: Sensing-based semi-persistent scheduling (SPS) y + j ∗ PSPS = z + Pstep ∗ k (3)
0
where PSPS is the sidelink SPS interval of the
In Fig. 6, Veh-1 is our desired transmitter UE for
corresponding SL SPS configuration given by higher
which resources have to be allocated. In each TTI 0
layers, PSPS = PSPS ∗ Pstep /100.
of the selection window, Veh-1 identifies a list L1
of Mtotal candidate resources that are needed for Mode 4 operation: Utilizing the sidelink grant,
the transmission of both SCI and TB (adjacent RBs the UE monitors the resources utilized by other
are allocated for SCI and TB in Fig. 6). A candidate UEs in the previous 1,000 ms interval, which is
single-subframe resource R x,y is a set of contiguous known as sensing. The UE uses a resource re-
subchannels LsubCH . The selection window time selection counter Cresel whose value is decremented
interval is [n + TP , n + TL ], where TP ≤ 4 ms and by one every time a packet is successfully transmit-
20 ≤ TL ≤ 100 ms [6]. ted. Once the Cresel value reaches zero, the UE has
to perform sensing and allocate resources.
From the above list L1 , Veh-1 discards those re-
sources that are affected by the events below and
creates a new list L2 . Next, Veh-1 discards the re- IV. Validation of the ARTERY-C
source elements in the list L1 , which are already simulator
reserved (persistently scheduled) by another UE in
the previous 1,000 TTIs (highlighted in red color in In order to validate the implementation of the Cellu-
Fig. 6). The UE decides to discard a certain sub- lar V2X protocol stack in the simulation framework
frame y in the selection window in accordance to and to demonstrate its capabilities, we assess the

8
 Artery-C – An OMNeT++ Based Discrete Event Simulation Framework for Cellular V2X

(a) Non-IP, V2V broadcast among platoon members (PM) and platoon (b) Scenario with adjacent regions with and without eNodeB coverage
head (PH) and IP, V2I unicast data traffic between PH and roadside
unit (RSU)

Figure 7: Use case V2X-based platooning with mode switching between adjacent regions of cellular coverage and non-coverage

performance of V2X-based platooning [9]. The im- trucks in the platoon by non-IP, V2V broadcast.
plementation of the use case applies sidelink and These messages contain information about distant
a simultaneous flow of IP and non-IP based data road conditions and traffic information; their size
traffic for V2I and V2V (Fig. 7a). We test the use varies between 50 and 1,500 bytes. We note that
case in a highway-tunnel scenario with a truck pla- the scenario also contains other surrounding vehi-
toon and surrounding vehicles with three variants: cles, which are not PM and periodically generate
(i) full eNodeB coverage, (ii) no eNodeB coverage CAMs whose frequency is expected to vary be-
and (iii) adjacent regions with and without eNodeB tween λ = [1, 10] Hz. In the simulations, the nodes
coverage. The latter option implies mode switch- generate CAMs with a fixed period, which does
ing, see Fig. 7b. The scenario-related simulation not depend on the vehicle dynamics as in [12]. In-
parameters are listed in Table 2. stead, we vary the CAM period in order to control
For the platooning use case, we assume that the the data traffic load. The generated traffic load in
vehicles, which consists of N trucks, drive in a terms of payload size depends on the values of
single-lane formation with a fixed inter-vehicle dis- λ and hence we consider normalized traffic load
tance and coordinate their maneuvers. The first in our simulations. The mobility model for the
truck acts as “platoon head”, the others as “platoon platoon is adapted according to the “ACC” car fol-
member” (PH and PM, respectively). The vehicles lowing model in SUMO where the vehicles drive in
operate in sidelink mode 4 when they are outside accordance to the speed limits in Table 2 while
of coverage of an RSU. When cellular coverage is maintaining a minimum gap of 2.5 m.
available, they can switch to mode 3 or continue For the performance assessment of the “in-
to remain in mode 4 (Fig. 7a). In the chosen setup, platoon” V2X communication for both modes, we
an RSU operates as eNodeB.4 It stays connected to consider two metrics for evaluation. The proba-
the infrastructure and is responsible for network- bility of message reception Pr refers to the ratio
assisted V2I communication. of the number of messages successfully received
The platoon exchanges two types of messages (Ta- (Nr ) to the number of messages transmitted to the
ble 3): (i) all vehicles inside a platoon, including PH intended recipient (Nt ). The factors contributing
and PMs, transmit non-IP, V2V broadcast messages. to successful message exchange between UEs are:
Following ETSI standards, these messages have the (i) the periodicity of CAMs and the reception time
type CAM, a variable size of 280-330 bytes and are of alert messages from the RSU, (ii) the availability
generated periodically [12]. (ii) The PH exchanges of resources in the selection window of SB-SPS and
IP-based, V2I unicast messages with an RSU and (iii) half-duplex constraints.5 The second metric,
forwards the information from the RSU to the other end-to-end (E2E) latency, measures the time taken

4 An RSU can operate as UE- or eNodeB-type (3GPP TR 23.285 5 Due to the half-duplex constraints, a vehicle cannot receive a

V14.2.0). packet because it transmits its own packet in the same subframe.

9
A. Hegde and A. Festag

Table 2: Scenario-related simulation parameters

Parameter Road type

Highway Tunnel
Vehicle speed [km/h] 100 – 130 60 – 80
Range of sidelink broadcast [m] 100 80
Cellular coverage regions (Fig. 7b) R2, R4 R1, R3
Types of vehicles Cars, trucks
Pedestrians/slow moving vehicles No
Traffic capacity (number of vehicles) 2,000 – 6,000 2,000

Table 3: Overview of the simulated use case all the connected UEs, their traffic load and the
resources utilized by them. Also, when an alert
Use case V2X-based platooning message interrupts a CAM transmission, the eN-
Road type Highway with tunnel odeB employs a load balancing scheme to efficiently
Nodes Platoon with 6 vehicles, RSU, allocate resources for different message types. In
other vehicles with mode 4 when the CAM transmission of a vehicle is
32 vehicles/(km lane) interrupted by another message of higher priority,
Message types Non-IP based CAM, i.e., alert, the UE has to immediately configure the
IP-based Alert sidelink grant and allocate resources for the higher
Data exchange V2I: RSU platoon head (PH) priority message. If resources are still available in
V2V: platoon members (PM) the selection window, CAMs can still get transmit-
Message Unicast: RSU, infrastructure ted; otherwise they are lost. We stress that mode 4
distribution Unicast: RSU and PH suffers from half duplex constraints where mes-
Broadcast: among PMs sages are lost at the receiving UE when a sender
Interfaces Uu and PC-5 uses the same subframe to transmit its own CAM or
Mode switching Uu to PC-5 (mode 3) alert message. This constraint does not depend on
PC-5 sidelink mode 3 to 4 the distance between the transmitter and receiver
Carrier frequency V2I: 5.9 GHz, V2V: 5.9 GHz but rather on the size of the subframe and the mes-
sage rate of a vehicle. In case of mode 4, the effect
is more pronounced at higher traffic load.
for the transport of a CAM between the transmit-
ting UE and the intended recipient UE. The E2E In case of mode 3, the eNodeB decides on the
latency for alert messages is calculated as the time resource reservation and allocation based on the
taken for the transport of message from the RSU to feedback it receives from the vehicle about the dy-
a PM via the PH. The E2E latency in both cases is namically changing traffic load and the latency re-
affected by the resource allocation latency [10] of quirements. On the other hand, in mode 4, the
the SB-SPS scheme. vehicles reserve their resources for several consec-
For each of the above defined metrics, we make a utive periodic message transmissions indicated by
comparison of the mode 3 and mode 4 performance the Cresel . When the traffic load changes dynam-
as depicted in Fig. 8 and 9. In Fig. 8a, we can ob- ically, several vehicles will compete for the same
serve that at lower traffic loads, Pr is comparable radio resources, which leads to multiple iterations
for both modes. For small CAM generation fre- of resource re-selections causing additional latency.
quencies, there is sufficient time for the CAMs to However, with the vehicle density considered in
obtain resources and get transmitted. At medium- our simulation scenario, we can see that the E2E
to-high traffic load, mode 3 performs better than latency for both CAM and alert messages meet the
mode 4 because the eNodeB is constantly aware of defined limits set by the standards [9, 12] for both

10
 Artery-C – An OMNeT++ Based Discrete Event Simulation Framework for Cellular V2X

(a) Probability of successful message reception Pr (aggregated CAMs and (b) End-to-end latency for CAMs and alerts
alerts)

Figure 8: Dissemination of periodic messages among vehicles in a motorway platoon: Comparison of mode 3 & 4

mode 3 and mode 4. V. Conclusions

Fig. 9 illustrates the impact of mode switching on
the reliability of packet transmission. We note that We have presented Artery-C, an OMNeT++-based dis-
the PMs do not remain in the same mode through- crete event simulation framework for the assess-
out the simulation as in Fig. 8. When the vehicles ment of Cellular V2X protocols and the evaluation
switch from mode 4 to mode 3, connection establish- of V2X application performance. Artery-C com-
ment and time synchronization with the eNodeB is prises control and user plane of Cellular V2X, im-
faster than the vice versa process. As a result, the plements components for every layer of the Cellular
vehicles obtain resources within the expiry period V2X protocol stack and realizes up-/downlink and
of the message. This effect ensures that almost up sidelink communication. By seamless integration
to 90 % of the generated messages are successfully into the existing Artery framework, it facilitates the
received. Depending on the location of the vehicle use of microscopic mobility models from SUMO,
and the time at which it switches to mode 4, it has the simulation of the full C-ITS protocol stack in-
to wait for the subsequent cycle to exchange syn- cluding ad hoc networking, facilities, security and
chronization subframes, which occurs at a period various other advanced features. Artery-C meets
of 160 ms, with other UEs. In situations where the the requirements for a comprehensive simulation
traffic load is high and the synchronization gets framework related to software, Cellular V2X and
slower, we can observe that (Pr ) reduces to almost timing. Using Artery-C, we also have presented
60 %. This further supports the fact that it is al- performance results for V2X-based platooning in
ways preferable for a vehicle to switch to mode 3 if a highway scenario as a representative use case.
available. These results demonstrate several capabilities of the
simulator and validate technical key features.

VI. Acknowledgements
This work was supported by the German Science
Foundation (DFG) within the priority program Co-
operatively Interacting Automobiles (CoInCar) (SPP
1835). We would like to thank Mr. Raphael Riebl for
insights about OMNeT++ and Artery, Mr. Quentin
Delooz and Ms. Julia Rainer for feedback and the
developers of SimuLTE for providing their simula-
Figure 9: Impact of mode switching on reliability tion environment to the open source community.

11
A. Hegde and A. Festag

References [6] European Telecommunications Standards In-

stitute (ETSI). Intelligent Transport Sys-
[1] Rafael Molina-Masegosa and Javier Gozalvez. tems (ITS); LTE-V2X Access Layer Specifica-
LTE-V for sidelink 5G V2X vehicular commu- tion for Intelligent Transport Systems Oper-
nications: A new 5G technology for short- ating in the 5 GHz Frequency Band, 2020.
range vehicle-to-everything communications. EN 303 613 V1.1.1.
IEEE Vehicular Technology Magazine, 12(4):30–
39, 2017. [7] Antonio Virdis, Giovanni Stea, and Giovanni
Nardini. Modeling unicast device-to-device
[2] Antonio Virdis, Giovanni Stea, and Giovanni communications with SimuLTE. In IWSLS,
Nardinis. Simulating LTE/LTE-advanced net- Vienna, Austria, July 2016. IEEE.
works with SimuLTE. In M. S. Obaidat, T. Ören,
J. Kacprzyk, and J. Filipe, editors, Simulation [8] Raphael Riebl, Giovanni Nardini, and Anto-
and Modeling Methodologies, Technologies and Ap- nio Virdis. Simulating LTE-enabled vehicu-
plications, pages 83–105. Springer, Cham, 2015. lar communications. In Recent Advances in
(SimuLTE website https://simulte.com, re- Network Simulation, pages 407–423. Springer,
trieved Sep 11, 2020). Cham, 2019.

[3] Sebastian Kühlmorgen, Patrick Schmager, [9] 3rd Generation Partnership Project (3GPP).
Andreas Festag, and Gerhard Fettweis. Study on enhancement of 3GPP support for
Simulation-based evaluation of ETSI ITS-G5 5G V2X services, 2018. TR 22.886 V16.2.0.
and Cellular-VCS in a real-world road traffic
[10] Anupama Hegde and Andreas Festag. Mode
scenario. In VTC-Fall, Chicago, IL, USA, Au-
switching strategies in Cellular-V2X. IFAC
gust 2018. IEEE.
Symposium, 52(8):81–86, September 2019.
[4] Brian McCarthy and Aisling O’Driscoll.
[11] Sasssan Ahmadi. 5G NR: Architecture, Technol-
Opencv2x mode 4: A simulation extension
ogy, Implementation, and Operation of 3GPP New
for cellular vehicular communication networks.
Radio Standards. Elsevier Academic Press, 2019.
In CAMAD, Limassol, Cyprus, October 2019.
DOI: 0.1016/C2016-0-04944-6.
IEEE. (OpenCV2X website: http://www.cs.
ucc.ie/cv2x, retrieved Sep 11, 2020). [12] European Telecommunications Standards Insti-
tute (ETSI). Intelligent Transport Systems (ITS);
[5] Raphael Riebl, Hendrik GÃijnther, Christian
Vehicular Communications; Basic Set of Ap-
Facchi, and LarsS Wolf. Artery: Extending
plications; Part 2: Specification of Cooperative
Veins for VANET applications. In MT-ITS
Awareness Basic Service, 2019. EN 302 637-2
’15, pages 450–456, Budapest, Hungary, June
V1.4.1.
2015. IEEE. (Artery website http://artery.
v2x-research.eu, retrieved Sep 11, 2020).

12