Urban Traffic Density Models

A Traffic Density Model for Radio Overlapping in Urban Vehicular Ad hoc Networks Saeed Bastani, Bjorn Landfeldt Lavy Libman, School of Information Technologies, University of Sydney, NSW 2006, Australia Networks research group, NICTA, 13 Garden St., NSW 1430, Australia sbastani@it.usyd.edu.au, bjorn.landfeldt@sydney.edu.au, lavy.libman@sydney.edu.au Abstract Vehicular traffic density is a key factor in determining the behavior of radio overlapping and thus impacts performance of data and safety message communication in Vehicular Ad hoc Networks (VANETs). In this paper, we propose a novel density model for urban traffic systems and employ this model for the purpose of spatial-temporal analysis of radio overlapping. To model traffic density, we consider a signalized junction and road segments linked to that junction as basic building blocks of urban traffic systems. The density model enables us to derive a framework to explain trends and critical regions of radio overlapping corresponding to VANET scenarios targeted for urban transportation systems, which cannot be derived from uniform density models widely used in existing literature. We apply the derived radio overlapping model to study channel load associated with periodic beaconing, a fundamental mechanism for safety message communication in VANETs. This study also provides a generic analytical framework to investigate other performance aspects of data and safety message communication in VANETs. Our experimental results using the proposed realistic density model demonstrate the need for an adaptive mechanism to adjust transmission power and data rate to reduce channel loads associated with dense traffic conditions. Keywords: Traffic Density; Vehicular Ad hoc Networks; Radio Overlapping I. INTRODUCTION Vehicular Ad hoc Networks (VANETs) have recently attracted significant research and development efforts from different players including academia, standards bodies, and industry, leading to DSRC/WAVE [1][2] as a reference standard, numerous dissemination/forwarding schemes [3], and several active projects [4]. Inter-vehicle communications open the door for a plethora of applications and services to provide safety and comfort applications through wireless vehicle-tovehicle communications. Comfort applications are expected to improve the passengers travel experience and optimize traffic efficiency, while the safety applications aim at minimizing accident levels. Safety applications can further be categorized into periodic and event-driven applications. The first category has an informative nature, as messages are disseminated among vehicles regularly to inform drivers about local parameters such as speed and position. With event-driven applications, on the NICTA is funded by the Australian Government as represented by the Department of Broadband, Communications and the Digital Economy and the Australian Research Council through the ICT Centre of Excellence program. other hand, safety messages are broadcast by vehicles whenever they experience or detect a hazard or otherwise notable event. Fundamental to the performance and reliability of these applications is the status of radio overlapping among communicating parties and interference from other parties. Under ideal channel conditions, if the radio range of one communicating party overlaps with a single party and the two parties do not attempt transmission simultaneously, any packet transmitted between these two parties is always received successfully. In case the two parties transmit simultaneously, a packet collision occurs, a phenomenon termed internal interference. In case a third radio overlaps with radios of communicating parties, a more stringent requirement exists; specifically, a transmission from vehicle to vehicle is successful if no other vehicle having radio overlapping with vehicle is simultaneously transmitting [5][6]. In the case where vehicle is outside radio range of, this phenomenon is known as external interference or hidden terminal effect. The key factor determining the intensity of radio overlapping is traffic density. In static wireless networks, due to the deterministic traffic density and distribution of nodes throughout the network area, it is straightforward to characterize the traffic behavior and thus radio overlap. On the other hand, in vehicular ad hoc networks (as in mobile ad hoc networks in general), characterizing radio overlap involves taking into account dynamic topology changes due to vehicles mobility, which in turn is affected by microscopic and macroscopic traffic parameters [7][8][9]. These parameters include, but are not limited to, traffic regulations on junctions and road segments, driver behavior, traffic flow, road capacity, etc. In vehicular networks, radio overlap behavior is expected to be even more complicated compared to other mobile ad hoc networks due to the impact of unexpected drivers behavior and variable traffic flow on vehicles mobility [8]. It is also expected that radio overlapping behaves differently between highway and urban VANET scenarios. This is partly due to the fact that traffic density is homogeneous under free and stable traffic flow regimes, which mainly dominate highways. In contrast to a highway scenario, a mixture of different traffic densities can be observed simultaneously in an urban scenario as simple as a road segment linked to a signalized junction. Moreover, an urban traffic network can be seen as a 2-dimensional network, compared to a 1-dimensional highway network. This makes the shape and dynamics of traffic density

more complicated in urban scenarios, resulting in more complex radio overlap behavior, especially near junctions. In line with the above, we are motivated to develop a traffic density model for a simple urban traffic scenario comprising a signalized junction and road segments linked to that junction. The proposed traffic density model is employed to analytically describe spatial-temporal behavior of radio overlapping throughout an urban road segment. Our findings on radio overlap are then applied towards a study on channel load associated with periodic beaconing in VANETs. To the best our knowledge, no analytical research work has been conducted so far to characterize urban traffic behavior effects on radio overlapping in VANETs. The remainder of this paper is organized as follows. Section II provides an overview of related work. Section III describes the proposed traffic density model. Using this model, in section IV we develop an analytical framework to characterize radio overlap in spatial and temporal dimensions. Section V specifies the periodic beaconing application and the channel load it imposes, which is then used in the numerical evaluation presented in section VI. Finally, section VII concludes the paper. II. RELATED WORK In the literature, the study of radio overlapping is usually performed in the context of radio interference and is mainly addressed in the form of topology control in wireless networks. Heide et al. [10] introduced an explicit definition of interference based on current network traffic and Burkhart et al. [5] proposed the following explicit definition: the interference of a link is the number of nodes covered by two disks centered at node and node with transmission range. Formally, the definition is expressed as follows: is covered by,, where denotes the set of nodes that can be affected by node and node when they communicate with each other using the minimum power needed to reach one another, is the distance between node and node, and ; denotes the disk centered at node with transmission range. This interference definition is also known as coverage-based interference. Using this definition, the authors of [5] disprove the widely used assumption that sparse topologies imply low interference. Furthermore, they propose a centralized algorithm to compute an interference-minimal connectivity-preserving topology. Rickenbach et al. [12] argued against the definition in [5] on two points: (i) It is sender-centric, i.e., the interference is considered to be an issue at senders instead of receivers. Hence, this definition deviates from the real-world interference; (ii) The definition is not robust enough to withstand the addition or removal of an individual node in the network. In contrast to this sender-centric interference definition, a receiver-centric interference model was proposed in [12]: the interference of a node corresponds to the number of nodes covering node with their disks induced by their transmission range reachable to node. Furthermore, an algorithm with approximation ratio ( being the maximum node degree) to find the optimal connectivity-preserving topology was proposed. Since the coverage area of an omni-directional antenna does not have a sharp boundary, Moscibroda et al. [13] studied the average interference problem, where the nodes were grouped into active nodes () and passive nodes (). The interference of a node is the number of nodes whose transmission ranges cover. Correspondingly, the average interference of a topology is the sum of all interferences divided by the number of passive nodes. A greedy algorithm was proposed to compute an approximation to the connectivity-preserving interference problem, where is the number of nodes in the network. Most relevant to our work is [14], where Du et al. adopted a definition of interference given by Jain et al. [15] and incorporated vehicular mobility and traffic flow characteristics to explain interference behavior in VANETs. According to [15] and assuming that is the Euclidian distance from node to node and is the transmission range of node, a transmission is successful if both of the following conditions are satisfied: (1), and (2) any other node such that or is not transmitting. As a very simple traffic scenario, the authors proposed their interference model for a number of nodes deployed with density on a freeway with free flow traffic regime. Although they evaluated the impacts of different traffic conditions on the interference behavior independently, they stopped short of explaining how the proposed interference model behaves in mixed traffic conditions where more than one traffic regime exists on the road segment simultaneously, as is the case in urban networks. In this paper, we adopt a definition of interference based on radio overlapping among nodes, defined as the number of nodes within the transmission range of a potential transmitting node. In contrast to [14], our goal is to develop an appropriate density model to characterize radio overlapping that takes into account the details of urban traffic behavior, e.g. the acceleration and queuing phases of vehicular movement between urban intersections. III. MODELING TRAFFIC DENSITY A. Assumptions of the Model A traffic density model is designed to capture the basic characteristics of an urban traffic system. More specifically, a signalized junction and a road segment linked to that junction are considered as the basic components comprising an urban traffic system (see Figure 1). Such a traffic density model is expected to express traffic density as a function of position (along the road segment ) and time during a traffic light cycle at the junction. For simplicity, we assume traffic is unidirectional with flow / approaching the junction from a -lane road segment. The maximum speed of the road segment is limited to, in line with regulations imposed on urban transportation systems. The junction is signalized and operates with fixed timing consisting of red and green phases denoted by and, respectively. Hence, the duration of a traffic light cycle is denoted by. We assume that the junction operates in near-saturation

L W L N l j L E l d l f l a l j l d l f l j l d l f L S O Figure 1: Signalized junction scenario condition, i.e. the queue forming during a red phase entirely discharges during the green phase of the same traffic light cycle. Traffic densities corresponding to under-saturation and over-saturation scenarios can be numerically compared to the near-saturation scenario. B. Traffic Density Model There are numerous studies in the literature investigating different aspects of traffic in the presence of signalized intersections. Queuing behavior at signalized intersections has been widely studied and many models have been proposed [19][20][21]. However, few works address global traffic behavior throughout an urban road segment. By global traffic model we mean a model able to represent traffic characteristics, e.g. traffic density, throughout a road segment or a large traffic system formed by a network of connected road segments. Local traffic models, on the other hand, are only capable of estimating traffic behavior in an immediate vicinity of a vehicle or a position along a road segment. An example of local traffic density models is found in a model proposed by Pipes [11] and used by Artimy et al. [16] for dynamic assignment of the transmission range in VANETs. In that model, a vehicle s instant velocity and a sensitivity factor are used to estimate local density around that vehicle. We argue that existing queue models or local traffic density models cannot adequately capture the general behavior of traffic throughout an entire road segment. Therefore, a global traffic density model is needed to explain traffic behavior at any given position throughout a road segment, and hence enable the analysis of radio overlapping along a road segment. Our proposed global traffic model is inspired by observations from 100 simulation runs conducted for single- and multi-lane scenarios in Paramics [17], a popular traffic simulator. Our findings are summarized in the following main points: a) During the red phase, three regions with different traffic densities coexist along the road segment: 1) a jam traffic density caused by vehicles building up a queue, 2) a growing traffic density caused by vehicles decelerating as they approach a queue ahead, and 3) an almost constant light traffic density caused by vehicles driving in free/stable flow traffic state. The regions with the three different densities are denoted by,, and, respectively (Figure 2). Figure 2. Traffic density defined as the number of vehicles in 40m intervals on a 3-lane road segment linked to a signalized junction (cycle length=100s, red time and green time=50s) b) During the green phase, a fourth region associated with a different traffic density emerges as vehicles in front of the previously formed queue gradually discharge the queue. This fourth region is denoted by in Figure 2. c) The shape and dynamics of the three density regions formed during a red phase and partly during a green phase are analogous to the shape and dynamic of a logistic curve [18]. Due to significant differences between the traffic dynamics associated with red and green phases of a traffic light cycle, we address traffic density corresponding to each phase separately. According to item (c) above, traffic density can be expressed by a simplified logistic curve defined as:, where: is the lower bound traffic density associated with stable/free flow traffic state; is the upper bound traffic density associated with jam traffic density (if 0 then is called the carrying capacity); is the growth rate of traffic density from lower bound () to upper bound ( ); is the reflection point of the logistic curve;, is the traffic density at position at time instant during a traffic light cycle. To model traffic density using (1), the above mentioned parameters must be determined. We start with, defined as the lower bound of traffic density, which is equivalent to traffic density in stable and/or free flow traffic state (region in Figure 2). Assuming a deterministic traffic arrival / and road speed limit, the inter-arrival time of vehicles. is measured in seconds and the headway distance is measured in meters. In a stable flow traffic state, a following vehicle must keep a safe headway distance from a leading vehicle to avoid rear end collision. Intuitively, the safe headway distance is equal to the distance that the following vehicle with velocity and average deceleration should maintain in order to be able to make a full stop when the leading vehicle brakes suddenly. The safe headway distance is (1)

also known as the braking distance and denoted by. Consequently, the headway distance in free and stable flow traffic states can be expressed as HD max, and traffic density can be expressed as: is measured in / and is the number of lanes. For traffic arrivals with other distribution functions such as Poisson, in the definition of HD and hence in (2) is replaced with the expected traffic arrival. in (1) is the traffic density associated with vehicles queued up at a junction (region in Figure 2) and is inversely related to jam headway distance, i.e.: (2) (3) where is the average jam headway distance and is a known parameter. To calculate and, we first calculate the braking distance (region in Figure 2). We previously concluded that the braking distance (denoted by ) is a function of initial velocity ( ) and the average deceleration rate ( ). The braking distance is expressed as. Also, the logistic function, can be rewritten with respect to as (we substitute, by for brevity). We define and as the start and end positions of the braking region and and denote the traffic densities corresponding to and, respectively. The difference between and must be equal to (braking distance), and, hence, Consequently: (4) The last parameter to be determined is, which is the reflection point of the logistic curve, equal to the queue length built up to the current time plus half the braking distance. We obtain: 2 (5) where represents the queue length (in meters), which is a function of the average traffic arrival rate (, current time instance (), and the headway distance corresponding to jam traffic. Finally, we consolidate the calculated parameters into the logistic function described by (1) as:, :,, 2 (6) where is measured relative to the junction position. and in 6 are scenario dependent and can be used to fine tune the mobility behavior. However, they do not greatly impact the general model. During a red phase,, defined by (6) measures traffic density at any given time instance for any position within the road segment. However, during a green phase, due to the presence of a queue discharge region (denoted by in Figure 2), the density function will be different from the red phase. The approach to model density during a green phase is to decrement from (6) the rate of density reduction due to the queue discharge, i.e.:, (7) where is a function of vehicles average acceleration and drivers average reaction time. To define we utilize a snapshot of vehicles trajectory shown in Figure 3. Assume,,,, are currently queued up at a junction. At the beginning of a green phase these vehicles are located at positions 0,, 2,, 1 respectively (Figure 3(a)), where is the jam headway distance. We define, the time elapsed since the start of the green phase. starts moving after (Figure 3(b)), starts moving after 2 (Figure 3(c)), starts moving after 3, etc. At a time such that 2, vehicle has travelled the distance and the positions of,,, are unchanged. At time where 2 3, vehicles and have travelled distance and 2, respectively and the positions of,, are unchanged. This process can be similarly continued for subsequent time instances and for the remaining vehicles in the queue. Generally, at time where 1 2 the locations of vehicles and 1 are and 1 1, and the distance between the two vehicles is 2 2 1. Consequently, the initial distance of the two vehicles and 1 queued up during a red phase, increases at time instance during a green phase where, and the increase rate is obtained as. It is straightforward to generalize the rate increase by describing the vehicle number (i.e. ) as a function of its position (i.e., namely. Bearing this in mind, and noting that the initial density of queued vehicles is, the density at position x at time t in a -lane scenario becomes and the rate of density reduction ( ) can be expressed as:. (8) 0. Having obtained by (8), and considering that the lower bound traffic density is, (7) can be rewritten as:

,., (9). buildings). Therefore, vehicles on segment do not contribute to of vehicles on segment. Other scenarios of segment overlapping, e.g. partial overlapping, fall within these two extreme scenarios. Hereafter we focus on road segment as the main road segment and derive of each position throughout this road segment. In deriving we address the two overlapping scenarios described above separately. Non-overlapping segments: Due to different descriptions of density functions corresponding to red and green phases, we derive in each phase separately. At any given time instance during a red phase ( 0 ) and for any position on road segment, substituting, in (10) with the density function expressed by (6) determines as:, (11) Likewise the description of for a red phase, substituting, in (10) with the density function described by (9), determines at position at a given time instance during a green phase (i.e. ) where is bounded as in (8). Hence, we define as the upper bound of in (8), i.e.. Depending on the relative values of and, different expressions for are derived as: Figure 3: Evolution of traffic discharge in a queue IV. RADIO OVERLAPPING To address radio overlapping we define the Radio Overlapping Number ( ) at position along the road segment as the number of unit disks covering a vehicle positioned at where each unit disk corresponds to a vehicle on the road segment. In our analysis we assume that the radius of all unit disks is equivalent to the nominal transmission range R of vehicles. RON is simply translated into the number of nodes located in the transmission range of a vehicle located at position, and is determined by calculating the integral over the density function, described by (6) and (9), corresponding to red and green phases of the traffic light cycle. Incorporating the temporal dimension into the definition of, it can be expressed as:,. (10) where,,,, and and are the coordinates of the start and end positions of the road segment under consideration. For an unbounded road segment the bounds of the above integral will simply become,. Depending on whether or not the vehicles on the two perpendicular road segments (denoted by and in Figure 1) overlap, two extreme cases can be identified: full overlapping and non-overlapping segments. Full overlapping is attributed to the situation where the area surrounding the two road segments (shadowed area in Figure 1) is an open space and no obstacles are present within this area. Hence, radios belonging to vehicles on one road segment can potentially overlap with those on the other road segment. The non-overlapping segments scenario is attributed to a situation where the entire surrounding area is covered by obstacles (e.g., Case 1:, (12) where is the integral of defined by (8), i.e. Case 2:, Case 3: 2 2 2 (13), (14) Full overlapping segments: The radio of a vehicle at position on road segment overlaps with radios of vehicles on road segment if. If this condition is satisfied, the length of the overlapping sub-segment on will be. To this end, the total radio overlapping experienced by a vehicle in position on road segment is equivalent to radio overlapping attributed to the vehicles on this road segment plus the number of vehicles currently existing on the overlapping part of road segment, i.e.:, (15) where, is determined using (11) for 0 (red phase) and (12)-(14) for (green phase). is the number of vehicles driving at time (with respect to traffic light timing of road segment ) on the 0, part of road segment. Calculation of

will be straightforward when the symmetry property of traffic light timing for road segments and is taken into consideration. By timing symmetry, we mean that if the traffic light at time instance is in red phase for road segment, at the same time it is green for road segment and an equal time duration has elapsed from the beginning of red and green phases as perceived by the two road segments. As a result, (15) can be rewritten as:,, 0 (16) is determined by calculating (10) with 0, and obtaining, from expression (9), replacing the traffic parameters of road segment with those of. In a similar manner, is determined using, by using expression (7) with the parameters corresponding to road segment. It is informative to compare the radio overlapping calculated using our proposed density model with that obtained using uniform traffic density, as adopted widely in the literature. Assume traffic is uniformly distributed with density / on road segments and. Following the same approach as above, the radio overlapping corresponding to non-overlapping and full overlapping scenarios for a uniform traffic density model can be obtained as follows: Non-overlapping segments: 2. Full overlapping segments: 2 2 2 (17) (18) V. CHANNEL LOAD ASSOCIATED WITH PERIODIC BEACONING IN VANETS The notion of, introduced in section IV, can be readily employed to determine channel load in vehicular ad hoc networks. Obviously, channel load is dependent on the characteristics of the safety or data application. In this section, we focus on channel load associated with beaconing, a basic mechanism for safety message dissemination in VANETs. The channel load of periodic beaconing perceived by a vehicle at position is the total data rate generated by vehicles within radio transmission range () of this vehicle. To determine the channel load, two parameters must be known a priori: the beacon arrival rate per vehicle ( /, and the number of vehicles existing on a stretch of road segment with radius centered at position (i.e., ). Bearing this in mind, it is easy to show that channel load of beaconing (denoted by, ) perceived at position during time instance can be written as:,, (19) where is size of each beacon message and is the beacon arrival rate. Note that in the definition of channel load, it is assumed that the channel is ideal, beacon transmissions are scheduled and thus collision-free, and retransmissions are disabled. In this sense,, can be thought of as a channel load lower bound under real channel conditions. Given that and are fixed parameters, the maximum possible channel load is obtained when, is at a maximum, which in turn occurs when the entire segment of size 2 centered at is in jam traffic condition. In this case, the maximum channel load can be determined as follows: (20) VI. NUMERICAL RESULTS To verify the density model derived in section III, we conduct a set of three different experiments in Paramics, each with a different number of lanes and traffic flows. As a typical setting for major roads in urban environments, we set traffic light timing to 50, speed limit 20 /, average driver reaction time 1, and the length of road segments 1000. In our experiments we ignore the effects of traffic belonging to road segments and on traffic density associated with and, but, clearly, they can be incorporated in the model in a straightforward manner. Traffic flow associated with any experiment is set to near saturation flow and determined according to the capacity of the junction. These traffic flow settings correspond to ideal signalized junctions. The verification metrics used are average accuracy and cross correlation of traffic density, calculated by our model and the traffic density measured in Paramics. Our results are shown in Table I. According to this table, the mean error of the model is around 14% with confidence interval 95%. In road segments with higher number of lanes, the accuracy decreases slightly. This can be attributed to a high probability of takeovers in road segments with a large number of lanes. However, except for very wide highways with a large number of lanes, the proposed model can be used safely in most urban traffic scenarios. As mentioned above, by fine tuning M and B to the specific scenario, the accuracy can also be slightly improved. This is corroborated by the fact that the proposed model mimics traffic behavior properly, as cross correlation results show. TABLE I. VERIFICATION OF TRAFFIC DENSITY MODEL 1 lane 2 lane 3 lane Traffic flow (vehicle/hour) 850 1800 2740 Mean Error 12% 13.4% 15.5% Cross Correlation 97.5% 95.6% 95.4% We continue the radio overlapping analysis using a 3-lane scenario. Figure 4 (a) and (b) show the RON corresponding to the non-overlapping and full overlapping scenarios, respectively. According to Figure 4(a), uniformly

increases as new vehicles arrive and join the queue formed at the junction. The maximum with magnitude 114 is observed at time 59 and in position 151. This can be attributed to the fact that in the beginning of a green phase, up to a time when the queue discharge rate exceeds the traffic arrival rate, the queue length continues to grow in accordance with an extension of the red phase. After 1s (= reaction time) from the beginning of the green phase, the queue starts to discharge, causing lower RONs in the front region of the queue. As time passes in the green phase, the length of this region increases, leaving more positions with low. Moreover, the region with large are shifted away from the junction towards farther positions on the road segment. This phenomenon is explained by a further queuee forming during the green phase as a result of non-zero reaction time and lack of acceleration of all vehicles queued up during the red phase. Specifically, maximum with magnitude 114 is shifted from position 151 to position 219 at time 75 and with magnitude 107. Following this shifting process, the maximum at the end of the green phase is observed at position 348 with magnitude 83. It is also important to note that in spite of the queue discharge during the green phase, the number of positions with high is larger compared to the red phase. In the remaining road segment with a flat shape, the average magnitude of is 11 with confidence interval 95%, and this region is dominated by free/stable flow traffic with no impact from the junction. Compared with the non-overlapping scenario, the full overlapping scenario (Figure 4(b)) shows similar behavior for the major part of the road segment except for positions with, which show high at all times during a cycle. The fluctuations of in this region are attributed to the dynamics of traffic density on the second road segment (i.e. ) described by Eq.(12)-(14). In addition to the global maximum produced in region during a green phase, a local maximum also emerges as a result of queue shifting in a similar way to the non-overlapping scenario. To study the channel load of beaconing in VANETs, we conduct a different set of experiments with various transmission ranges and traffic flows. Transmission ranges are selected from those specified in DSRC. Traffic flows are set in a way that they cover a range of flows from 50% below to 50% above saturation flow to reflect different classes of traffic, ranging from sparse to dense conditions. Again we use a 3-lane scenario with its associated saturation flow of 2740 vehicle/hour and vary flows with respect to this saturation point. Beaconing parameters and are set to 500 bytes and 10 /. Figure 5 (a) and (b) show maximum channel load observed in non-overlapping and full overlapping scenarios, respectively. According to Figure 5(a), in the nonchannel load overlapping scenario, the maximum corresponding to 150 transmission range grows linearly from 2.25 associated with traffic flow 1370 to 5.37 associated with traffic flow 3425. From this point up to higher flows, the channel load approaches its maximum corresponding to a jam traffic condition (described by Eq. (20)). Therefore, the channel load curve becomes flat with only slight change observed. As the transmission range exceeds 150, a larger contribution of the overlap comes from vehicles in free flow state. This (a) (b) Figure 4: Radio overlapping (a) non-overlapping, (b) overlapping segments (a) (b) Figure 5: Channel load, (a) non-overlapping, (b) overlapping segments

explains why for transmission ranges larger than 150, the growth in channel load is linear within the entire range 1370, 4110 of traffic flows. In the full overlapping scenario (Figure 5(b)), due to contribution from vehicles on segment, the maximum possible channel load on segment is reached for longer transmission ranges compared to the nonoverlapping scenario (i.e. for 300). Observe that for long transmission ranges, e.g. 1000, channel load can be as high as 17 in dense traffic conditions. VII. CONCLUSION In this paper, we proposed an accurate density model for urban traffic systems that takes into account the different phases of vehicle movement between signalized intersections. We have established that the number of overlapping radios behave in a highly dynamic manner, which cannot be adequately captured by the widely used uniform density model. Based on our density model, we have investigated radio overlapping and channel load of safety beacon messages. As a result of traffic heterogeneity, a single data rate and/or transmission power cannot be assigned to all vehicles at all times. Our evaluation of the RON model has been based on a unit disk coverage without retransmissions. This is a conservative approach, and using a more realistic propagation model with channel errors would indeed further increase the channel occupancy due to retransmissions. Furthermore, traffic on twoway road segments contributes to higher channel load compared to the one-way scenario investigated in this paper. Considering the nominal range of data rates specified for VANETs operation in the DSRC standard (3-27 Mbps), it becomes clear that the use of data rates in the lower part of that range is questionable in urban areas when the wireless transmission range is large. The problem is exacerbated when other safety applications, in addition to periodic beaconing, are taken into consideration. The scarcity of the radio resource underscores the need for careful design of applications and protocols in urban VANETs in order to mitigate channel load in dense traffic regions, based on vehicle s estimations of their local traffic density or traffic information provided by roadside infrastructure. The design of an adaptive and robust mechanism to adjust transmission power and assign proper data rates based on perceived radio overlapping is a subject of our future work. ACKNOWLEDGMENT This work was partially sponsored by the Australian Research Council under grant DP0987782. REFERENCES [1] Standard Specification for Telecommunications and Information Exchange Between Roadside and Vehicle Systems 5 GHz Band Dedicated Short Range Communications (DSRC) Medium Access Control and Physical Layer Specifications, ASTM E2213-03(2010). [2] IEEE Draft Standard for Information Technology - Telecommunications and information exchange between systems - Local and metropolitan area networks - Specific requirements - Part 11: Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) specifications, Amendment 6: Wireless Access in Vehicular Environments. IEEE Std 802.11p, July 2010. [3] K. C. Lee, U. Lee, and M. Gerla. Survey of routing protocols in vehicular ad hoc networks. Advances in Vehicular Ad-Hoc Networks: Developments and Challenges, IGI Global, October 2009. [4] S. Zeadally, R. Hunt, Y.S. Chen, A. Irwin, and A. Hassan. Vehicular ad hoc networks (VANETs): status, results, and challenges. Telecom Systems, Springer Netherlands, 2010. [5] M. Burkhart, P. V. Rickenbach, R. Wattenhofer, and A. Zollinger. Does topology control reduce interference. MobiHoc'04, Roppongi, Japan, May 2004. [6] R. Rajaraman. Topology control and routing in ad hoc network: A survey. ACM SIGACT News, 33(2):60-73, 2002. [7] M. Fiore, J. Harri, F. Filali, and C. Bonnet. Understanding vehicular mobility in network simulation. IEEE International Conference on Mobile Adhoc and Sensor Systems (MASS), 2007. [8] J. Harri, F. Filali, and C. Bonnet. Mobility models for vehicular ad hoc networks: A survey and taxonomy. IEEE Communications Surveys Tutorials, 11(4):19 41, 2009. [9] J.J. Blum, A. Eskandarian, and L.J. Hoffman. Challenges of intervehicle ad hoc networks. IEEE Transactions on Intelligent Transportation Systems, 5(4):347 351, December 2004. [10] M. A. D. Heide, C. Schindelhauer, K. Volbert, and M. Gruenewald. Energy, congestion and dilation in radio networks. 14th Annual ACM Symposium on Parallel Algorithms and Architectures (SPAA), Winnipeg, Manitoba, Canada, August 2002. [11] L.A. Pipes. An operational analysis of traffic dynamics. Journal of Applied Physics, 1953. [12] P. Rickenbach, S. Schmid, R. Wattenhofer, and A. Zollinger. A robust interference model forwireless ad-hoc networks. IEEE International Workshop on Algorithms for Wireless, Mobile, Ad-Hoc and Sensor Networks (WMAN), Denver, Colorado, USA, April 2005. [13] T. Moscibroda and R. Wattenhofer. Minimizing interference in ad hoc and sensor networks. DIALMPOMC'05, Cologne, Germany, September 2005. [14] L. Du, S. Ukkusuri, and S. Kalyanaraman. Characterizing interference in vehicle ad hoc networks on freeway segments under different traffic flow conditions. Journal of Transportation Research, 2009. [15] K. Jain, J. Padhye, V. N. Padmanabhan, and L. Qiu. Impact of interference on multi-hop wireless network performance. Wireless Networks, 11(4):471-487, 2005. [16] M. M. Artimy,W. Robertson, and W. J. Phillips. Assignment of dynamic transmission range based on estimation of vehicle density. 2 nd ACM Workshop on VANETs, Cologne, Germany, September 2005. [17] Quadstone Paramics, 2008. Available at: http://www.paramicsonline. com. [18] S. E. Kingsland. Modeling nature: episodes in the history of population ecology. University of Chicago Press, ISBN 0-226-43728-0. [19] R. Akçelik and N. M. Rouphail. Overflow queues and delays with random and platooned arrivals at signalized intersections. Journal of Advanced Transportation, 28: 227 251, 1994. [20] G. Stephanopoulos, P. G. Michalopoulos, and G. Stephanopoulos. Modelling and analysis of traffic queue dynamics at signalized intersections. Transportation Research Part A, 13(5): 295-307, October 1979. [21] F. Viti and H.J. Van Zuylen. Modeling queues at signalized intersections. Transportation Research Record 1883, 2004.