conditions and free parking lots along the streets in real time. There are some ..... View: A Scalable Traffic Monitoring System", Proc. of 2004. IEEE Int. Conf. on ...
Evaluation of Inter-Vehicle Ad-hoc Communication Protocol Masashi Saito, Jun Tsukamoto, Takaaki Umedu, Teruo Higashino Graduate School of Information Science and Technology, Osaka University {m-saito, tukamoto, umedu, higashino}@ist.osaka-u.ac.jp Abstract In this paper, we propose an inter-vehicle ad-hoc communication protocol called RMDP (Received Message Dependent Protocol) which disseminates and propagates the preceding traffic information to the following vehicles, and discuss its performance. Our proposed protocol dynamically changes the dissemination interval depending on the number of reception messages for avoiding message collision in heavy traffic jam conditions. We design the parameters of the protocol based on the simulation results. Our simulation results show that by using RMDP, a lot of vehicles can acquire the preceding traffic information within very short periods. In addition, we have also shown that selection of disseminate data affects preceding traffic information acquiring time. Each vehicle holds other vehicles’ tracks for some period. Depending on the number of preserved other vehicles’ tracks and their preserved period, propagation ratio of the preceding traffic information varies. Some simulation results are explained.
1. Introduction According to recent progress of wireless LAN technology and diffusion of GPS products, car navigation systems are going to equip GPS receivers and wireless LAN cards such as IEEE802.11 [15] to acquire the information about traffic jams, road surface conditions and free parking lots along the streets in real time. There are some researches to realize these services by installing wireless LAN base stations along streets. Although such installation is possible, considerable costs may be required. Such expenses may be reasonable for urban roads and highways; however, they are too expensive for back streets with light traffic. On the other hand, car drivers always want to access not only trunk road information but also back street one. In these cases, we do not require that all the information can be accessed from any place. It is sufficient that each local vehicle can access its neighboring traffic information. To solve these problems, we are doing research on disseminating and propagating traffic information using
inter-vehicle ad-hoc communication protocol without special communication infrastructures. In this paper, we show evaluation results of our designed application layer inter-vehicle ad-hoc communication protocol. Each vehicle disseminates its own traffic information and receives traffic information from other vehicles. Then, it relays the combined information to other vehicles. Each vehicle holds a certain number of traffic information, and propagates them. Oncoming vehicles carry preceding vehicles’ information to the following vehicles. As a result, each vehicle can acquire several kilometers preceding traffic information. In this research, we have developed a mobile ad-hoc network simulator that monitors inter-vehicle data propagation situation from the location information of all the vehicles. It produces various kinds of statistics information such as packet collision ratio and reception message amount. This simulator cooperates with a traffic flow simulator NETSTREAM [13] that produces actual traffic flow. In our previous research, we have evaluated SDRP (Speed Dependent Random Protocol) which changes the dissemination interval by vehicles’ speed [14]. We have evaluated the collision ratio and the number of reception messages for some equipping ratios of car navigation systems. In general, we do not know such equipping ratios in advance, and it is dynamically changed day by day. So, it is difficult to choose a suitable dissemination interval to maximize the reception data amount. Here, we propose RMDP (Received Message Dependent Protocol) whose purpose is to maximize the reception data amount without knowing the equipping ratio. In heavily jammed traffic conditions, although SDRP leads too many collisions and appropriate data propagation cannot be done, RMDP solves this and most vehicles can acquire the head of traffic jams in a short time.
2. Ad-hoc Network for Inter-vehicle Communication We can acquire traffic jam information by receiving broadcasting media such as FM multiplex broadcasting [16], radio-wave beacons, and digital radio, and/or by inquiring a traffic information server using car
navigation systems with cellular phones[17], [18]. On the other hand, the research to propagate and acquire such information using inter-vehicle communication is actively studied recently. For example, [1] proposes a method to propagate information, which is acquired from a base station on a road, to other vehicles using inter-vehicle communication. [9] proposes a multicast protocol for inter-vehicle ad-hoc communication, and [5] proposes a peer-to-peer communication protocol to prevent collisions at a road crossing. [8] proposes an ad-hoc communication protocol between mobile terminals and servers, which are placed at signals, kiosks in stations, or lounges at airports. To build inter-vehicle ad-hoc networks without base stations, some data dissemination methods have been proposed like [2] and [10]. [4] proposes a protocol that changes the data dissemination rate depending on the sending node density. However, in most of those methods, they assume that mobile terminals move randomly in order to simplify the problems. For each vehicle, the relevance to given traffic information depends on the distance and time from its origin; however, there exist very few research results that consider such relevance. [6] and [9] have been developed to monitor inter-vehicle mobile ad-hoc communication. Although [6] consider multiple lanes’ roads, they assume that roads themselves are straight simple ones. [10] proposes an inter-vehicle communication protocol that takes account of time and movement of vehicles on a road map; however, it does not consider traffic jams and signal waiting. In addition, they assume collision free communication so that all the data can be received successfully. In actual traffic jam conditions, broadcasting data from many vehicles may cause a broadcast storm [7], and too many collisions may prevent the inter-vehicle communication. So, we should take care to prevent the collisions. NETSTREAM [13] is a traffic flow simulator for performing the effective prediction for traffic jams and prior evaluation of ITS introduction. It defines traffic flow characteristics and the lengths of signals for all road links, and then calculates all vehicles’ actions for every second. For each road, it defines S being the distance between a front vehicle and itself, and K=1/S being the road’s traffic density; and it calculates each vehicle’s speed V corresponding to the ratio of K by Kcongestion, which is the heavily jammed traffic density and the maximum speed at free driving Vmax. V=Vmax(1-K/Kcongestion) Therefore in wide areas such as a large city, it can calculate the traffic flow with high accuracy. To evaluate traffic information propagation situations in an inter-vehicle ad-hoc network setting, we have developed a network simulator. Our simulator cooperates with NETSTREAM so that we can
evaluate our inter-vehicle communication protocol with typical traffic flow. Our simulator also supports the network collision status monitor to simulate the more realistic conditions.
3. Ad-hoc Communication Protocol for Acquiring Traffic Information 3.1. Local Traffic Information Services The services that we want to realize are to provide location dependent local traffic information. In these days, vehicles equip a lot of sensors, so we can get useful information for driving. For example, when many vehicles moving to the same direction are stopping and/or moving at very low speeds, we can predict a traffic jam. If this information can be propagated by opposite lane’s oncoming vehicles and if it is reached to the tail of the traffic jam, drivers just arriving at the tail of the jam can recognize the length of the traffic jam, and they may be able to avoid it by re-routing. Other examples are freeze road surfaces by slide conditions of tires, and rainy conditions by movement of wipers. By acquiring such information in advance, vehicles’ drivers can prepare danger situations by slowing down their vehicles’ speed. So by combining local traffic information services with global information services, which are delivered by using radio wave and cellular phones, we are able to have much safer driving environments. To realize location dependent services, we assume that the set of the vehicle information is measured at every second. We also assume that each set can be encoded within 100 bytes, and each vehicle exchanges at most 100 sets including ones that are received from other vehicles using inter-vehicle communication. In this case, the size of the disseminating data is 10 Kbytes at a time. We also assume that the bandwidth of radio communication for application protocols is 100 Kbytes/sec. Inter-vehicle communication is done as follows. Vehicle-A holds a set of its own speed, location and direction, and disseminates it to surrounding vehicles at time t. Vehicle-B, which is driving on the opposite lane, receives the information and moves to another place; and then re-disseminates vehicle-A’s information Time= t+d
Vehicle A
Vehicle C
Time= t
Disseminate
Move Vehicle B
Vehicle B
Figure 1 Inter-vehicle Ad-hoc Communication
together with vehicle-B's information at time t+d. At that time, another vehicle-C may be able to receive the vehicle-A’s information. That means vehicle-C can know its preceding traffic conditions and road surface situations using inter-vehicle ad-hoc communication (Fig. 1). The chain of information dissemination leads propagation of each location’s traffic information within a short period.
3.2. Inter-vehicle Ad-hoc Communication Protocol As a method of performing inter-vehicle ad-hoc communication, we can use IEEE802.11 IBSS [3]. In IBSS, CSMA/CA collision avoidance algorithm is defined and broadcast communication is supported. Our proposed protocol is an application protocol over UDP/IP protocol on the top of IEEE802.11. 100 vehicles’ data sets are aggregated and disseminated as one UDP packet by using broadcast facility. Here, we use the following assumptions to simplify the simulation. At first, vehicles within a circle of 100m’s radius can communicate each other; however, success probability of data exchange decreases in proportion to the square of distance. When the distance between two nodes becomes large, the receiving radio power becomes weaker and the reception error ratio is increased [12]. Here, for the influence of the large objects on the street and surrounding buildings along the street, we treat them within the approximation of the above reception probability.
3.3. Traffic Information Dissemination Protocol We have designed SDRP for traffic information dissemination, first. In SDRP, according to each vehicle’s speed v, a random transmission interval is calculated. In case of a traffic jam, since there are many vehicles around the vehicle, the transmission interval is longer. In addition, in case of high speed driving, the interval is shorter to increase the propagation probability. From our simulation results, we can specify the appropriate data dissemination interval if the equipping ratio is given [14]. However, in the real world, we do not know the equipping ratio as a priori knowledge, and it may vary from moment to moment and from place to place. So nobody knows the suitable dissemination interval that maximizes the total reception data amount. If we can estimate the suitable dissemination interval from the number of reception data in a fixed period, we can implement the dissemination protocol for any equipping ratio. For this purpose, we propose a protocol called RMDP (Received Message Dependent Protocol). In RMDP, we assume that the dissemination interval
is in inverse proportion to the number of reception messages. Here, we propose a method for obtaining the parameters to maximize the total amount of the reception data using our previous simulation results. Let P be the dissemination interval and r be the number of reception data in a fixed period. P=
α
1− β ⋅ r
(α,β>0)
Here, let the fixed time be 30 seconds. When vehicles can run more than 30 km/h, we have fixed the dissemination interval to 1 second. If their speed is less than 30 km/h, the dissemination interval is set to 1 to 16 seconds one by one: (1) When the equipping ratio is 60%, the total reception data amount is maximized when P is set to 1 second. In this case r = 14.00. (2) When the equipping ratio is 90%, the total reception data amount is maximized when P is set to 2 seconds. In this case r = 15.13. For the above equation, by substituting the values of P and r, we can obtain that the solutions for α and β are 0.14 and 0.06, respectively. So we can estimate the dissemination interval P when the number of receiving messages for 30 seconds is r: P=
0.14 1 − 0.06 ⋅ r
RMDP can perform suitable data dissemination by using this interval. Here, in the case that r is greater than or equal to 17, P is set to infinity. So RMDP suspends data dissemination. When these vehicles receive dissemination messages, they also calculate their own dissemination intervals for the next dissemination in parallel. In the following simulation, we use RMDP with these parameters.
4. Evaluating RMDP by Using Mobile Ad-hoc Network Simulator 4.1. Mobile Ad-hoc Network Simulator To evaluate traffic information propagation situations in an inter-vehicle ad-hoc network setting, we have developed a mobile ad-hoc network simulator. Our developed mobile ad-hoc network simulator inputs NETSTREAM log data, which include all vehicles’ positions, speed and direction in every second. We also give the network environment and equipping ratio of car navigation systems as the simulation parameters. Based on these parameter values, this simulator produces data propagation situation from the location information of all the vehicles, and records various kinds of statistics information such as packet collision ratio and reception data amount. In the real world, if each vehicle receives such traffic jam information, it may dynamically change its route. However, we do not take account of the influence
in this evaluation for simplicity.
4.2. Data Dissemination Simulation Results
4.3. Acquiring Traffic Jam Information Vehicles at the tail of a traffic jam are interested in the information about its traffic jam length and the location of its head. Drivers may choose another route
60%
25
50%
20
Data Amount (GBytes)
Collision Ratio
4.2.1. Collision Ratio and Reception Data Amount by RMDP Using our mobile Ad-hoc network simulator, we have evaluated the inter-vehicle traffic information propagation with the following input parameters: ・ road environment : 20 km x 20 km ・ the number of signals : 198 ・ simulation time : 60 minutes ・ location information of vehicles : every second ・ the number of vehicles : 4890 ・ network environment : 100m, 100Kbytes/sec ・ traffic density : light Fig. 2 shows the comparison of SDRP and RMDP with respect to their collision ratios and reception data amounts. We use SDRP’s dissemination interval; 1 - 16 seconds for less than 30km/h vehicles, and 1 second for more than 30km/h vehicles. From Fig.2(a), SDRP’s collision ratio is increased corresponding to the equipping ratio. On the other hand, RMDP’s collision ratio is constant, around 17%. From Fig. 2(b), RMDP’s reception data amount is slightly less than SDRP’s one, but it is almost the same as the value of 4 seconds’ interval for less than 30km/h vehicles’ case. Therefore, RMDP can also work well in the real world. 4.2.2. Collision Ratio and Reception Data Amount in Case of Heavy Traffic Jam Conditions The simulation results of Section 4.2.1 reflect the situations of wide areas with light traffic. In this section, we show the simulation results for a crossing (construction site) with a heavy traffic jam using RMDP. Evaluation parameters are follows: ・ road environment : 2 km x 2 km one four-lane road and one two-lane road which are crossed at an intersection ・ the number of signals : 1 (at an intersection) ・ simulation time : 20 minutes ・ the number of vehicles : 5480
・ equipping ratio : 90% ・ traffic density : very heavy average 700 m traffic jam In this traffic condition, using SDRP causes the collision ratio being 96%, therefore, only a small number of data can be propagated. In addition, even if we use RMDP, the collision ratio increases to 98%. The reason is very simple; when a vehicle approaches the tail of a traffic jam, it seldom receives valid messages because of collisions caused by a lot of vehicles ahead. If such collisions often occur, the number of each vehicle’s reception data becomes also small because of communication errors. This makes the dissemination interval of RMDP shorter because each vehicle may misunderstand that there are only a few vehicles near the vehicle. This makes the situation worse. We assume that MAC layer can detect such message errors. In fact, in case of IEEE 802.11 MAC layer, it detects frame errors to change their inter frame space from DIFS to EIFS for avoiding further collisions. Using this facility, we improve RMDP so that we can consider communication errors caused by collisions. Here, we also count such communication errors as the message receptions. When a given traffic condition is very heavy, we cannot estimate how many vehicles disseminate their vehicles’ information simultaneous and cause a collision. So RMDP should use an appropriate weight for each occurrence of errors. When we count a communication error as 0.5 message reception, the collision ratio is still high, 89%. The collision ratio decreases when we increase the weight. When we use the weight more than 1, the reception data amount becomes almost constant. We believe this is the effect of RMDP’s interval changing algorithm. In fact, when a vehicle receives more than 17 messages in 30 seconds, it stops disseminating.
40% 30% 20% 10% 0% 1
2
4
8
16
15 10 5 0
RMDP
SDRP Dissemination Interv al(sec)
30.0%
60.0%
90.0%
(a) Collision Ratio Figure 2
1
2
4
8
16
SDRP Dissemination Interv al (sec)
30%
60%
90%
(b) Reception Data Amount SDRP and RMDP Comparison
RMDP
Jam Head Information Acquiring Ratio
90% 80% 70% 60% 50% 0 Weight=0.5 Weight=2
Figure 3
30
60 Weight=1 Weight=4
90
120 Time (sec)
Jam Head Information Acquiring Ratio by Time
4m in
4 8min
6
80% 70% 60% 50% 0
50
100
150
200
Distance (m)
Weight=1 Weight=4
4.4. Duplicate Data Dissemination Our previous simulation results show that all vehicles cannot always acquire their preceding traffic information. In [14], we have shown that by increasing the dissemination of each vehicle’s data for a specific period, their preceding traffic information can be acquired more easily. One of the reasons is that a lot of data are disseminated and the most of them may be duplicated, and useful data necessary for each vehicle are not well passed. In fact, if we can collect one datum for a 200m square area, we need 100 data in order to collect data for a 2km square area. Using simulation environment in Section 4.2.1, we have also evaluated the duplication frequency of all the disseminating data. Fig. 5 shows the data duplication frequency by the distance between a reception data’s location and the current location of the vehicle. Fig. 6 shows the frequency by the time interval between the reception data’s time and the current time. We evaluate them by changing the period of vehicle’s own data aggregation. In both cases duplication frequency is very high at a first glance. Fig. 5 shows that when we make the own vehicle’s data period longer, duplication frequency decreases. It also shows that 12 km far location’s information can be exchanged at the maximum; data are propagated towards geographically diversified sufficiently. When we use 12 minutes period, the duplication frequency does not change by distance. This means it may be difficult to choose dissemination data by distance. Fig. 6 shows that when we make the own vehicles Duplication Frequency
(Thousand)
Duplication Frequency
70 60 50 40 30 20 10 0 2
90%
Figure 4 Jam Head Information Acquiring Ratio by Distance
when the jam length is very long. Fig. 3 and Fig. 4 show the simulation results of time and distance of vehicles that can acquire the information about the head of a traffic jam (here, we call it the head information). In this simulation, the head information means the information about a vehicle in the same direction whose position belongs to 10 m radius circle of a target crossing. The traffic jam length is about 700 m, and the number of vehicles at the jam is about 70 for each lane. Note that if we only want to know the traffic jam length, 10 m radius circle of a target crossing is very narrow (it may be 100 m radius circle). However, if we want to know the situation about the head vehicle of the traffic jam, i.e., if we want to know whether the head vehicle has some engine trouble or has some accident, it is desirable that we can check how much time we need for acquiring the head information in 10 m radius circle. When the error weight is 0.5, then 49% vehicles can acquire the information about the head of the jam within 1 second. Although the collision ratio is very high, all vehicles can acquire the head information within 45 seconds. On the other hand, when the error weight is 4, then 65% vehicles can acquire the head information within 1 second, but it takes 310 seconds all vehicles acquired the head information. This tendency is the same with respect to the distance. So using the following two criteria, we choose the weight being 1. (1) All the vehicles can acquire the head information within a short period. (2) Considering that the collision ratio should be small enough to avoid catastrophic conditions.
0
100%
Weight=0.5 Weight=2
8
10
12
(Thousand)
Jam Head Information Acquiring Ratio
100%
25 20 15 10 5 0 0
500
Distance (km)
12m in
Figure 5 Duplication Frequency by Time
4m in
8min
1000
1500
Time Diference (sec)
12m in
Figure 6 Duplication Frequency by Time
data period longer, duplication frequency decreases, too. It also shows that data are propagated towards time diversified sufficiently. The characteristic of the graph is that the duplication frequency increases to the own vehicles data period, and then decreases to five thousands. When we use 12 minutes period, the duplication frequency of the first 12 minutes is almost double of that of the next 12 minutes. Fresh data are more important for acquiring traffic information, so that this behavior is appropriate for our objects.
disseminating own vehicle’s data within 12 minutes, most vehicles can obtain preceding information, like the head of traffic jam, effectively. As a future subject, we are studying combining two protocols to increase acquiring preceding information ratio. We are also studying algorithm to select disseminating data to diversify them. Designing this algorithm makes almost all vehicles can obtain their preceding traffic information more rapidly.
References 5. Dissemination Data Selection Algorithm From the results in Section 4.4, as the period for each vehicle’s data dissemination, 12 minutes is appropriate. We also evaluate the influence of the interval to collect own vehicle’s data within 12 minutes. There is no influence regarding geographical and time diversification even if we change the interval to every 10, 20 and 30 seconds. So we decide to fix the own vehicle’s dissemination data as every 30 seconds in 12 minutes. Totally, we use 24 own vehicle’s data as a part of a disseminating message. In Section 4.4 the algorithm to select some vehicle’s data is random. To decrease duplication frequency and to improve the geographically diversified dissemination, we are concerning to apply the following rules: (1) Discarding duplicated data: In case of the reception data being similar one with respect to time and location to one of data in the buffer, we increase the reference count of the data and leave older one to decrease duplication frequency in Fig. 5 and Fig. 6, (2) Discarding very old data: Using random selection algorithm, there may exist some data that is disseminated in the past. Since traffic situation varies in about 30 minutes, we discard data with more than 30 minutes previous time stamp. This leads more diversification than that of Fig. 5.
6. Concluding Remarks In this paper, we show the evaluation results of RMDP which changes dissemination interval by the number of reception messages (including the number of communication errors caused by collisions). Our previously proposed SDRP works well in light traffic conditions, but it poorly works in heavily jammed ones. RMDP works well even in heavy traffic jam conditions by counting communication errors as the reception of messages. In a heavy traffic jam, collisions often occur and without this improvement, vehicles can seldom communicate each other. After applying this improvement, RMDP’s collision ratio decreased dramatically and most vehicles can acquire traffic jam head information in a short time. We also evaluate the diversification of reception data with respect to geographic and time. By
[1] H. Hartenstein, B. Bochow, A. Ebner, M. Lott, M. Radimirsch and D. Vollmer : "Position-Aware Ad-hoc Wireless Networks for Inter-Vehicle Communications: the Fleetnet Project", Proc. of 2001 ACM Int. Symp. on Mobile Ad-hoc Networking & Computing (MobiHoc), pp.259-262, 2001. [2] W. R. Heinzelman, J. Kulik, H. Balakrishnan : "Adaptive protocols for information dissemination in wireless sensor networks", Proc. of 5th Annual ACM/IEEE Int. Conf. on Mobile Computing and Networking (MobiCom'99), pp.174 - 185, 1999. [3] W. Kellerer, C. Bettstetter, C. Schwingenschlogl, P. Sties, K-E Steinberg and H-J Vogel : "(Auto) Mobile Communication in a Heterogeneous and Converged World", IEEE Personal Communications, Vol. 8, No. 6, pp.41-47, 2001. [4] A. Khelil, C. Becker, J. Tian, K. Rothermel : "An epidemic model for information diffusion in MANETs", Proc. of 5th ACM Int. Workshop on Modeling Analysis and Simulation of Wireless and Mobile Systems (MSWiM'02), pp.54-60, 2002. [5] R. Miller, Q. Huang : "An Adaptive Peer-to-Peer Collision Warning System", Proc. of IEEE Vehicle Technology Conference(VTC), pp.414-418, 2002. [6] T. Nadeem, S. Dashtinezhad, C. Liao, and L. Iftode : "Traffic View: A Scalable Traffic Monitoring System", Proc. of 2004 IEEE Int. Conf. on Mobile Data Management (MDM2004), pp.13-26, 2004. [7] S. Ni, Y. Tseng, Y. Chen, and J. Sheu : "The Broadcast Storm Problem in a Mobile Ad-hoc Network", Proc. of 5th Annual ACM/IEEE Int. Conf. on Mobile Computing and Networking (MobiCom'99), pp.151-162, 1999. [8] M. Papadopouli, and H. Schulzrinne : "Effects of power conservation, wireless coverage and cooperation on data dissemination among mobile devices", Proc. of 2nd ACM Int. Symp. on Mobile Ad-hoc Networking & Computing (MobiHoc2001), pp.117-127, 2001. [9] C. Schwingenschlogl and T. Kosch : "Geocast Enhancements of AODV for Vehicular Networks", ACM SIGMOBILE Mobile Computing and Communications Review, Vol.6, No.3, pp.96-97, 2002. [10] B. Xu, A. Ouksel and O. Wolfson : “Opportunistic Resource Exchange in Inter-vehicle Ad-hoc Networks”, Proc. of 2004 IEEE Int. Conf. on Mobile Data Management (MDM2004), pp.4-12, 2004. [11] A. S. Tanenbaum : "Computer Networks Forth Edition", Pearson Education Inc., 2003. [12] Theodre Rappaport: ”Wireless Communications: Principles and Practice, Second Edition”, Prentice Hall, 2001. [13] E. Teramoto, M. Baba, H. Mori, H. Kitaoka, I.Tanahashi, Y. Nishimura, et. al. : "Prediction of Traffic Conditions for the Nagano Olympic Winter Games Using Traffic Simulator : NETSTREAM", Proc. of 5th World Congress on Intelligent Transport Systems, Vol.4, pp.1801-1806, 1998. [14] M.Saito, M. Funai, T. Umedu and T. Higashino: “Inter-vehicle Ad-hoc Communication Protocol for Acquiring Local Traffic Information”, 11th World Congress on ITS, 2004. [15] IEEE802.11 Standard : "Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Specifications", ISO/IEC 8802-11:1999, 1999. [16] Vehicle Information and Communication System Center: “VICS”, http://www.vics.or.jp/eng/. [17] Toyota Motor Corporation: “G-BOOK”, http://www.toyota.co.jp/g-book/. [18] Honda Motor Co., Ltd: “InterNavi Premium Club”, http://premium-club.jp/.