Performances Analysis on Video Transmission

in a Wireless Sensor Network

Bogdan Ciubotaru, Dan Pescaru, Doru Todinca
Computer Department, Faculty of Automatics and Computers,
“Politehnica” University of Timisoara
2 Bd. V. Parvan, 300233 Timisoara, Romania

Abstract - This paper presents an analysis on transmission between different systems (sensor nodes, central points –
performance in a Wireless Sensor Network based on video sinks, processing units etc.).
sensors. Routing algorithm proposed is derived from Flooding, In this paper we will discus only the communication part of
Direct Diffusion and LAR. It is adapted for continuous video a video sensor network system. Because of the data quantity
data transfer. Unless the original algorithms, it considers quality
that has to be transferred in such of network and considering
of transmission issues required by video streams. Simulation
evaluation shows the overall estimated performance and the real-time constraints involved by video streaming and
emphasis some sensitive aspects. application requirements (security applications involves
timely alarm triggering) the traditional communications
I. WIRELESS SENSOR NETWORKS. protocols are not satisfactory.
The paper is organized as following: the next section
In the near future the wireless sensor networks (WSN) are describes some of the routing protocols specific to wireless
seen as powerful tool for human interaction with the ad-hoc networks and wireless sensor networks. Also it
environment. A wireless sensor network represents a macro- presents their drawbacks from the point of view of video
entity formed of hundreds or thousands of micro entities transfer. Next, we continue by presenting our solution for
called intelligent sensor nodes. Each node is equipped with video transfer in wireless sensor networks and performance
three basic functionalities: communication, computation and analysis of the presented solution.
sensing. Due to the high number of sensor nodes that forms a
wireless sensor network, which are able to communicate with II. OVERVIEW OF ROUTING ALGORITHMS IN WIRELESS
each other in a collaborative manner, this new type of SENSOR NETWORKS.
networks bring the possibility of a high resolution data
acquisition. Because of this data resolution the wireless Wireless sensor networks are a particular form of a
sensor networks bring a new vision of intelligent environment wireless ad-hoc network. The literature in ah-hoc network
monitoring systems. routing protocols is very reach, but the solutions developed
One of the applications that can benefit from the for wireless ad-hoc networks in general are not always suited
collaborative and pervasive characteristic of wireless sensor for wireless sensor networks. As presented in [1], ah-hoc
networks is video surveillance. Video surveillance systems routing protocols can be classified in flat routing protocols
are mostly used as security systems or as process or and hierarchical routing protocols. In case of flat routing
environment monitoring systems. The major drawback of protocols each of the network nodes has equal responsibility
these systems is the quantity of the information (visual to route the packets in the network and to maintain the
information) that they provide for the human operator to communication paths within the network by storing routing
process. It is very useful to have video cameras that record the information locally. Hierarchical routing protocols involve
activity in a specified environment but the time spent by the heterogeneous networks where nodes have different
human operators to analyze the recordings are sometime not responsibilities in terms of communication.
acceptable. This is where data centric orientation of sensor Traditional routing algorithms for ad-hoc networks like
networks plays its part. In wireless video sensor networks the Distance Vector based protocols [6] requires route
captured images will be further processed locally at sensor information dissemination between all the nodes (routers) in
nodes level and only the relevant information will be sent to the network leading to a major drawback in protocol
the sink. For example, in a site surveillance application the scalability.
sensor nodes will apply a motion detection algorithm to detect Optimized Link State Routing (OLSR) is an optimization
and separate only the changing portions of the image and send of the classical link-state protocol to be suited for wireless ad-
only this data to the gateway. hoc networks. The flooding overhead is minimized in case of
Video capturing involves dealing with important quantity OLSR by using only selected nodes for traffic information
of data that has to be processed, stored and transferred retransmission [8].

1-4244-1234-X/07/$25.00 ©2007 IEEE. 183

 SACI 2007 – 4th International Symposium on Applied Computational Intelligence and Informatics

LAR is a protocol that uses location information to find the should take care equally on quality of service issues.
routes for communication [8]. It uses location information to An earlier version of the Adaptive Routing (AR) protocol
reduce the area of search for a route, reducing in this way the presented here was proposed in [12]. It is derived from
communication overhead. Location information is taken from Flooding, Direct Diffusion [9] and LAR [10]. It is adapted for
GPS receivers and because of mobility of the nodes the video data transmission. The protocol involves two distinct
approximate position is estimated based on time and average phases. The first is called the setup-phase and it has as a main
velocity of the node. goal to fill the routing tables maintained at the level of each
Low-energy adaptive clustering hierarchy (LEACH) [2] is sensor node. The second phase involves video data transfer
a clustering based routing protocol. It assumes that each node between nodes and the sink.
has enough power to reach the base station and concentrate on
data aggregation at cluster level. Each cluster has a cluster A. The Setup Phase.
head node (randomly elected), which aggregates the date The setup phase should be done as the initial step after the
received from all the nodes in the cluster and sends the node deployment. However, this phase could by repeated time
resulting information to the base station. This routing solution to time to ensure configuration update. The main goal is to fill
is not suitable for our application because in this case the the routing tables, maintained by each node, with information
nodes are not able to reach directly the base station. about proximity and hop count to the central point. A second
Sensor Protocols for Information via Negotiation (SPIN) aim is to query the nodes for synchronously image
[3] is a flooding based routing protocol for wireless sensor information used to extract topology [11].
networks. The SPIN family of protocols represents an The central server (the sink) broadcasts a setup message
enhancement of the flooding protocol. In SPIN containing empty route path information. Each node that
implementation the data are transmitted only after a receives a setup message will determine if itself is included or
handshaking negotiation procedure where sensor nodes not in the route information. If it is included, it simply drops
advertise their data to neighboring nodes which in turn send the message. If not, it refreshes its routing table with
request for data messages to the advertiser. Flooding information from the message and broadcasts the message
algorithms are generally not suited for considered application further. Same time, it starts a timer that manage the image
because of the redundant data transmission required. SPIN capturing for the localization algorithm.
protocol reduces the redundant data transmission but do not Nodes are both routers and end points in the same time.
ensure a stable communication path to the base station. Each nodes help server requests to be propagated to the entire
Directed diffusion is a data-centric routing protocol [4]. In network by forwarding the request to its neighbors. Also
directed diffusion the sink floods a data interest to the images that are sent to the server are routed back to the server
network. Each node that receives the interest will establish in an energy-efficient manner that preserves network
gradients that will be used to drain the information from the integrity. Each node keeps simple routing information that
source node (the nodes that can gather the requested will describe all of its neighbors by them current energy level
information) to the sink. The sink will send later on and the hop-count to the server. To obtain energy-efficiency,
reinforcement messages for the paths that most efficient. This when node energy is over a threshold (e.g. 50%) no
solution creates data paths from the source to the destination restrictions are considered in electing a node as next hop.
(sink) but do not ensure a continuous data flow for large data When energy node is between a certain intervals (e.g. [20% -
packages with collision avoidance on the path due to 50%]) the node will be elected as next hop only if there is no
neighboring nodes. better alternative. If there is more than one candidate as
Greedy perimeter stateless routing (GPSR) [5] uses possible next hop, a less recently used (LRU) algorithm is
location information for message routing. Because location used to achieve equal energy consumption for all neighbors.
information will be necessary for tagging our sample images At server request the sensor nodes will send back an image
this routing protocol may be suited for our application. The captured by node's camera. The server makes a single general
problem that appears is also the problem of stable, request available for all nodes connected inside the network.
uninterruptible path to the base station. Due the deployment density, a request from server will be
received by a node possibility multiple times. In order to
III. ROUTING ALGORITHM OPTIMIZED FOR VIDEO DATA avoid sending duplicates, each request will be augmented
TRANSMISSION. with a unique ID.
The message containing node's response will be sent back
Wireless Video Sensor Networks are a special case of to the server using the most energy-efficient path with
wireless sensor networks in which large amounts of video network integrity preservation. Indeed, the most energy-
data are sensed, processed in real-time, and then transferred efficient path is considered the path with the lowest hop-count
over the wireless networks. Considering that, the routing to server having enough energy to cover communication.
protocols proposed for common WSN are less effective in Using this schema, the setup protocol is able to deliver to a
that case. Commonly they are designed to transfer small central point all necessary images in an efficient manner. At
amount of data in a reliable and power efficient manner. In the end, using the computed routes, the transformation
case of more complex Video WSN, the routing protocol parameters are sent back to nodes identified by images’ IDs.

184
 B. Ciubotaru, D. Pescaru, D. Todinca • Performances Analysis on Video Transmission in a Wireless Sensor Network

B. The Video Data Transfer Phase. simulation library that includes support for input/output,
The second phase of the routing protocol involves great statistics and data collection, graphical presentation of
effort in keeping a quasi-constant transfer rate for video data simulation data, random number generators and data
transmission. To accomplish this, a virtual channel between structures.
the node and the sink will be created and maintained. According to the objective of simulating various
One or more nodes could sense a motion in theirs deployment patterns, we build a random deployment
supervised field. Starting on that time, they will try to setup a generator. It creates the OMNeT++ NED files starting from
channel to the sink and transmit the data. The first one who basic input parameters like number of nodes and average
detects the motion will send a message and try blocks sensing density. The messages used for setup-phase can carry the
activity for all its neighbors. From the acknowledgments following information: the type of message, the hop-count
received, combined with information from local routing table, from/to the server, the message network path and a unique id.
it will select the best next hop to the sink. Using a hand- The messages for video transfer include the frame data and a
shacking algorithm, the selected node will be included on 16 bits CRC code.
transmission channel. To avoid dead channels, a timer will
start at the level of every selected node. Each frame sent V. PERFORMANCE EVALUATION
through current node will reset its timer. Same times, all
neighbors of channel nodes are forced to hibernate as We use 3 different density schemes for deployment
depicted by Fig. 1. Indeed, they have sensing activity topologies to measure transmission performance. First has a
blocked. Transition from hibernation to sensing state is done quasi-equal distribution, the second has one concentration
by receiving an end-transmission message from the sender or point and the third has two concentration points. For each
based on theirs internal timer in case of failures. topology we use the following network sizes: 3, 8, 16 and 30
nodes.
Our analyze is based on four performance metrics:
bandwidth estimation expressed by number of received
packets at sink (NPR), number of transmitted messages
(NTM), total amount of received messages (NRM) and
average dissipated energy (ADE). Comparison is done
against flooding and directed diffusion protocols. All
diagrams present average results for the considered density
schemes.
As depicted by Fig. 2, the number of successful received
packets is considerably greater using AR protocol. The
explanation resides in decreasing of number of transmitted
Channel Node Hibernating Node Sensing Node and received packets as shown in Fig. 3 and Fig. 4. However,
for large object images the frame rate is still very low. If we
Figure 1. A video transmission channel. Hibernating nodes protect consider a 128x128 image, this fits on 12 packets. Indeed, in
it against collisions.
term of frame rate this means a little bit more than 1 image
Video frames are split in 4 KB blocks at transmission time. per second. A compression algorithm is compulsory in that
Considering low resources for communication and buffering case.
at the level of network nodes, only tracked objects cropped In Fig. 3 the total number of transmitted messages is
from original image will be transmitted, instead of entire presented for different nodes densities. Regarding the total
frame. Even that, due to the limited channel bandwidth, amount of messages required to ensure information delivery
retransmission of every corrupted packet may be to the sink (gateway) node the AR protocol scales well with
unacceptable for video streaming. A possible solution resides performances in the range of Directed Diffusion.
in using a priority-based scheme as PSR-AQR [13]. Since this Directly related with the number of messages sent in the
scheme is adaptive, the video quality is automatically network is the total number of messages received by the
adjusted based on the quality of the wireless link. And, very nodes. The MAC protocol has a significant impact on the
important in our context, this introduces a quite small real- sent/received ration. Collision avoidance techniques play a
time processing overhead. Consequently, node processor is major role in the efficiency of the network communication.
able to carry all necessary tasks. The adjacent nodes blocking techniques used by the AR
ensures a collision free path from the transmitting node to the
IV. IMPLEMENTATION CONSIDERATIONS receiver (sink) leading to a high number of valid received
messages.
We choose Objective Modular Network Test-bed in C++ Fig. 5 presents the simulation result in term of average
(OMNeT++) for protocol simulation [14]. It is a component- dissipated energy. The energy consumption was estimated
based, modular simulation framework and represents a viable considering a small energy amount assigned to every
discrete event simulation environment. It offers an extensive transmission and reception for 1 bit. The average estimated

185
 SACI 2007 – 4th International Symposium on Applied Computational Intelligence and Informatics

result is 22% longer life for network using AR protocol VI. CONCLUSIONS AND FUTURE WORK
compared with flooding.
This paper presents a performance analyzes of an adaptive
20
18
routing protocol for video transmission over a wireless sensor
16 network. This analyze is made against a standard flooding
14
protocol and more sophisticated directed diffusion. It
no. of packets

12
10 demonstrates the validity of proposed protocol and
8
6 encourages as to continuing working on it.
4 As a future work we identify the necessity of a
2
0 t(s)
compression algorithm to minimize the transmitted data at the
1 2 3 4 5 level of sensor node. Also, we plan to include an adaptive
NPR Flood NPR DD transmission scheme to solve the retransmission of corrupted
NPR AR
packets problem.
Figure 2. Average NPR of Flooding, Directed Diffusion and
Adaptive Routing on a 30 nodes network.
REFERENCES
100000000 [1] J. A. Stankovic, T. F. Abdelzaher, C. Lu, L. Sha and J. C. Hou, ”Real-
Time Communication and Coordination in Embedded Sensor Networks,”
1000000 in proceedings of the IEEE, Vol. 91, No. 7, July 2003, pp. 1002-1022.
no. of msgs

[2] W.B. Heinzelman, A. P. Chandrakasan and H. Balakrishnan, "An

10000 application-specific protocol architecture for wireless microsensor
networks," in IEEE Transactions on Wireless Communications, Vol.1,
100
Iss.4, Oct 2002 Pages: 660- 670.
1 [3] W. Rabiner Heinzelman , J. Kulik and H, Balakrishnan, “Adaptive
protocols for information dissemination in wireless sensor networks,”
3 8 16 30 nodes Proceedings of the 5th annual ACM/IEEE International conference on
NTM Flood NTM DD Mobile computing and networking, p.174-185, August 15-19, 1999.
NTM AR [4] C. Intanagonwiwat , R. Govindan and D. Estrin, “Directed diffusion: a
scalable and robust communication paradigm for sensor networks,”
Figure 3. Average NTM on Flooding, Directed Diffusion and Proceedings of the 6th annual international conference on Mobile
Adaptive Routing in a logarithmic representation. computing and networking, Boston, Massachusetts, USA, August 2000,
pp.56-67.
[5] B. Karp and H. T. Kung, “GPSR: greedy perimeter stateless routing for
100000000
wireless networks,” in Proceedings of the 6th annual international
1000000 conference on Mobile computing and networking, August 06-11, 2000,
Boston, Massachusetts, United States, pp. 243-254.
no. of msgs

10000 [6] R. V. Boppana and S. P. Konduru, “An Adaptive Distance Vector

Routing Algorithm for Mobile Ad-Hoc Networks,” in Proceedings of the
100 Twentieth Annual Joint Conference of the IEEE, vol. 3, 2001, pp. 1753-
1 1762.
[7] T. Clausen and P. Jacquet, “Optimized Link State Routing Protocol
3 8 16 30 nodes (OLSR),” RFC 3626, October 2003.
NRM Food NRM DD [8] Y. B. Ko and N. Vaidya, “Location-aided routing (LAR) in mobile
NRM AR adhoc networks,” in Proceedings of The 4th ACM/IEEE Int. Conf.
Mobile Computing and Networking, Oct. 1998, pp. 66–75.
[9] C. Intanagonwiwat, R. Govindan, D. Estrin, J. Heidemann and F. Silva,
Figure 4. Average NRM on Flooding, Directed Diffusion and “Directed Diffusion for Wireless Sensor Networking,” IEEE/ACM
Adaptive Routing in a logarithmic representation. Transaction on Networking, vol. 11, no. 1, February 2003, pp. 2-16
[10] Y. B. Ko and N. Vaidya, “Location-aided routing (LAR) in mobile
adhoc networks,” in Proceedings of The 4th ACM/IEEE Int. Conf.
10
Mobile Computing and Networking, Oct. 1998, pp. 66–75.
8 [11] D. Pescaru, B. Ciubotaru, D. Chiciudean and A. Doboli, "Experimenting
6 Motion Detection Algorithms for Sensor Network Video Surveillance
mJ

Applications,” Transactions on AUTOMATIC CONTROL and

4 COMPUTER SCIENCE, Scientific Bulletin of “Politehnica” University
2 of Timisoara, ISSN 1224-600X, Vol. 50 (64), 2005, pp. 39-44
0 [12] M. Cosma, D. Pescaru, B. Ciubotaru and D. Todinca, “Routing and
Topology Extraction Protocol for a Wireless Sensor Network using
3 8 16 30 nodes Video Information,” in Proceedings of SACI’2006, pp. 353-363.
ADE Flood ADE DD ADE AR [13] F. Hou, P-H. Ho and Y. Zhang, "Performance Analysis of
Differentiated ARQ Scheme for Video Transmission over Wireless
Figure 5. Node Average Dissipated Energy (ADE) on Flooding, Networks," in Proceedings of the 1st ACM workshop on Wireless
Directed Diffusion and Adaptive Routing. multimedia networking and performance modeling, Montreal, Canada,
2005, pp. 1-7.
By preserving a collision free path to the gateway AR [14] A. Varga, “The OMNeT++ Discrete Event Simulation System,” in
protocol represents a good solution for video streaming over Proceedings of the European Simulation Multiconference (ESM'2001),
wireless sensor networks. It also achieves good performance Prague, Czech Republic, June 6-9, 2001.
from the point of view of energy dissipation.

186