Abstract—The safety and commercial benefits of Intelligent Transportation System (ITS) raised interests towards inter-vehicle networking technologies such as Vehicular Ad-hoc Network (VANET). Being an approved standard for wireless access in vehicular environments, IEEE 802.11p attracts a lot of research attentions, especially on its broadcasting performance. However, most of the previous network performance models paid little attention to vehicle distribution, or simply assumed homogeneous car distribution. It is obvious that vehicles are distributed non-homogeneously along a road segment due to traffic controls and speed limits at different portions of the road. In light of the inadequacy, we present in this paper an original methodology to study the performance of 802.11p VANETs with practical vehicle distribution in urban environment. An empirically verified stochastic traffic model is adopted, which incorporates the effect of urban settings (such as traffic lights) on car distribution and generates practical car density profiles. Based on the knowledge of car density at each location from the traffic model, the 802.11p broadcasting model is developed and a new metric, Broadcasting Performance Index (BPI), is introduced to better characterize the broadcasting performance and packet collision probability in VANETs. Furthermore, the analytical closed form for BPI is derived and its accuracy is confirmed with extensive simulation. In general, our results demonstrate the applicability of the proposed methodology on modeling protocol performance, and shed insights into the performance analysis of other communication protocols and network configurations in urban vehicular networks. Index Terms— Vehicular Ad-hoc Network (VANET), IEEE 802.11p, Stochastic Traffic Model, Markov Chain, Broadcasting Performance Index (BPI). I. INTRODUCTION With the growing number of cars on road, inter-vehicle networking technologies such as VANET attracted a lot of research effort to improve transportation safety. For example, rear-end collisions can be largely avoided if the braking car can send a warning message to the following cars. Besides, non-safety-related communication applications, such as Internet access and commercial advertising can also be enabled by VANET. In the US, a 75 MHz Dedicated Short Range Communication (DSRC) band was allocated for vehicular communications in 1999 [1]. The band is divided into seven 10 MHz channels. These include a control channel (CCH) for critical safety data exchange and six service channels (SCH) for non-safety-related data communication. To exploit the DSRC band, the original IEEE 802.11 or WiFi standard was enhanced to support inter-vehicle communications. For instance, the IEEE 802.11p amendment was published in July 2010 [2]. For the bottom networking layers, the standard adopts the PHY layer of 802.11a, and the MAC layer of 802.11e to accommodate the needs of vehicular communication. IEEE 802.11p uses CSMA/CA as the medium access method. In this method, time is divided into slots of 16 ρs and every node maintains a contention counter loaded with random integer picked between 0 and W SS . During a busy slot, the contention counter is frozen, but once the channel is sensed idle for consecutive periods or Arbitration Inter-frame Spaces (AIFS), the contention counter decreases by one for each passing idle slot. When the counter reaches zero, the packet is sent out. If two nodes start transmission in the same time slot or the transmission is interrupted by a hidden node, the transmission is failed, and the package is scheduled for retransmission with doubled contention window [0, 2W SS ], which is the well-known exponential backoff process. Retransmission upon collisions keeps occurring until the upper limit of the contention window reaches a predefined maximum value CW max . IEEE 802.11p also implements various Quality of Service (QoS) classes for traffic with different priorities by setting different values of W SS and AIFS (see Table 1). Therefore, every node maintains four queues, and each of them buffers a class of data. The queues are prioritized according to the importance of the data class. According to Table 1 again, the safety data queue in Access Class 3 (AC3) has the highest priority in contention. Lastly, the 802.11p standard requires each node to stay in the CCH and SCH alternatively for 50 ms in a switching period of length t p = 100 ms [3, 4] for ensuring service data exchange as well as safety data propagation. For further details of the IEEE 802.11p standard, the reader is referred to [1 – 4]. Table 1. Contention parameters for different access classes in 802.11p. AC Data Class CWmin CWmax AIFS 3 Video & Safety related 3 7 2 2 Voice 3 7 3 1 Best Effort 7 15 6 0 Background Traffic 15 1023 9 A data type of particular interest in inter-vehicle communications is the safety information broadcast. Cars equipped with GPS can inform their neighbors of their locations, velocity, acceleration and any emergency warnings through broadcasting. These safety information packets called the “heartbeat” of the vehicular network are of constant length (500 bytes), and are generated periodically (at a rate of 10 Hz). Because of its importance, heartbeat is broadcasted on the CCH with the highest priority. Hence, the broadcasting performance is a key evaluation parameter for VANET. However, the broadcasting node has no means to identify a collision, since no ACK packet will be sent by the receivers in broadcasting. Therefore, developing models for new broadcasting standards and identifying the optimal A Stochastic Traffic Modeling Approach for 802.11p VANET Broadcasting Performance Evaluation Harry J. F. Qiu, Ivan Wang-Hei Ho, Chi K. Tse Department of Electronic and Information Engineering, The Hong Kong Polytechnic University 07835614d@connect.polyu.hk, {ivanwh.ho, encktse}@polyu.edu.hk 2012 IEEE 23rd International Symposium on Personal, Indoor and Mobile Radio Communications - (PIMRC) 978-1-4673-2569-1/12/$31.00 ©2012 IEEE 1077