Chapter 1 Effective Information Sharing based on Mass User Support for Reduction of Traffic Congestion Tomohisa Yamashita, Kiyoshi Izumi, Koichi Kurumatani CARC, AIST, Japan tomohisa@carc.aist.go.jp kiyoshi@ni.aist.go.jp k.kurumatani@aist.go.jp In this research, our aim is to increase the utility of the drivers by reducing traffic congestion. To attain our purpose, we propose a simple route guidance mechanism based on route information sharing (RIS). The drivers using the RIS notify their route to the route information server, and then it returns accumulative information based on their routes to them. As a result of multiagent simulation, we conform that the travel time of the drivers using the RIS and others decrease as the drivers using the RIS increased. Recently, ubiquitous computing environments in road transportation systems have developed rapidly; car navigation systems have spread widely, the accuracy of GPS is constantly advancing, and the range over which the vehicle information and communication system (VICS) can provide congestion information has been 2 Information Sharing for Reduction of Traffic Congestion extended [9]. 1 These advances were made possible by the development of invehicle communication devices, sensors, and processors. One of most important information services in road transportation systems is navigation from an origin to a destination. Based on such developments, many researchers have been trying to develop navigation systems, and to examine what kind of traffic information is variable [1, 5]. However, previous research has revealed that the inprovement of traffic efficiency by simply providing traffic information is difficult [4, 6, 8]. Usually, navigation systems recommend routes that have shorter distance and are not congested based on map information and the current traffic state. To decrease travel time, a driver with a navigation system chooses one recommended route. However, if other drivers simultaneously choose the same recommended route, traffic is concentrated on that route. As a result, traffic congestion is caused on the recommended route. Therefore, traveling the recommended route often takes longer even though navigation systems recommended it for decreasing travel time. To make matters worse, traffic congestion increases even more exponentially as more drivers choose the recommended route. As car navigation systems have been spreading rapidly, this is a serious problem. In our work, we introduce the concept of mass user support [2, 3] to avoid the unintentional traffic congestion caused by use of navigation systems. Mass user support provides information services to users, or groups, or mass that go beyond conventional personal services. Although conventional services only considers individual utility, mass user support considers not only the amount of individual utility but also interactions among users. The purpose of mass user support is to increase utility for both individuals and society, but to increase the benefit to society by sacrificing benefits to individuals. Mass user support requires some kind of social coordination that leads to mutual concessions among users [2]. Moreover, a service based on mass user support is required to provide not only an increase of utility but also fairness among users, stability of the service, transparency of the mechanism, and robustness against disturbance because mass user support deals with many users at the same time. In order to increase individual’s utility and social benefits by reducing traffic congestion, we introduce information sharing of the route that vehicles will pass in the future through the route information server. This mechanism is expected to lead the situation that the drivers using the RIS know the tendency of congestion, and then, without concentrating at one route, they voluntarily and gradually change their routes to avoid area that will be congested. With multiagent simulation, we examine the effectiveness of our proposed mechanism from the following two points, i) as the drivers using our proposed mechanism increases, the efficiency, e.g., travel time from their origins to their destinations, should increase, ii) the efficiency of the drivers using our proposed mechanism should always be better than that of the drivers using other mechanisms. 1 The VICS Center [9] collects, processes and edits information about traffic congestion, road control, and other traffic information in real time, and then provides road traffic information in words and graphics through communications and broadcast media (FM multiplex broadcasting and beacons). Drivers can receive traffic information via navigation devices installed in vehicles. Information Sharing for Reduction of Traffic Congestion 3 Figure 1.1: Direction of movement of vehicles and revision of blocks 1.1 1.1.1 Multiagent Modeling Traffic Model In our traffic flow model, a road between intersections is called a link, and is divided into several blocks [6, 7]. The length of a block is equal to the distance that a vehicle runs at free flow speed Vf of the link during one simulation step. After division of a link, an order is assigned to each block from downstream to upstream. Concerning the block assigned to be the i-th, we define Ki as the density of block i, Li as the length of block i, Ni as the number of the vehicles in block i, and Vi as the feasible speed of vehicles in block i. In block i, Vi is revised based on Greenshield’s V-K relationship, described as Vi = Vf (1 − Ki Kjam ), (1.1) where Kjam is jam density, which means the minimum density prevents vehicles in a traffic jam from moving. The process of the flow calculation between the two neighboring blocks i and i + 1 is as follows. At every step, the speed of vehicles in each block is revised according to the V-K relationship. Based on Vi , vehicle j can move forward from downstream to upstream as shown in Fig. 1.1. If Ki is exceeds jam density Kjam , no vehicles can move into block i from block i + 1. At the next step in block i, vehicles can accelerate or slowdown to revised Vi immediately regardless of the speed in the previous last step. 1.1.2 Route Choice Behavior In our simulation, three types of drivers in a route choice are prepared. Shortest Distance Route The driver using the shortest distance route (SD) simply chooses the route that provides the shortest distance from the origin to their destination. It searches the route based on only map information. 4 Information Sharing for Reduction of Traffic Congestion Shortest Time Route The driver using the shortest time route (ST) chooses the route that has the shortest travel time from an origin to a destination, and revises the route whenever it approaches an intersection. This type of the driver searches a route based on current traffic information of the entire network that a traffic information center (e.g., VICS Center) provides. It considers the expected travel time of a link as calculated based on the current density of each block. First, speed Vi on block i is calculated based on the V-K relationship with density Ki . Next, the passage time of block i is calculated based on length Li of block i and speed Vi on block i. Finally, the passage time of a link is calculated by summing these of all blocks on the link. We define expected travel time ET Tl as the summation of the passage times of all blocks on link l. The drivers using the ST search for the shortest route based on the expected travel time. Shortest Time Route with Route Information Sharing The driver using the shortest time route with route information sharing (RIS) chooses a route based on current congestion status and accumulative information about the routes of drivers using the RIS, and revises the route whenever it approaches an intersection. The route information server shares route information of the drivers using the RIS. In the RIS mechanism, at first, the driver using the RIS searches the shortest route from its origin to its destination in the expected travel time. The route information server collects the routes of all drivers using the RIS, and assigns passage assurance of a driver to a link, which the degree of assurance that a driver will pass through a link in the future. Passage assurance P Al,j of driver j to link l is calculated as follows. If a route passes through p links from driver’s current position to a destination, the route information server assigns a order to each link from the destination to driver’s current position on the route. Next, the route information server divides the order by p and regards it as the weight of a link. Based on passage assurance of each link, the route information server calculates the total passage assurance of each link, which means the accumulation of passage assurance of all drivers using the RIS, and provides it to them. Total passage assurance T P Al is defined as the sum of the passage assurances of all drivers to link l. The route information server calculates T P Al as, k (1.2) T P Al = k∈RIS P Al,k , where RIS is the set of the drivers using the RIS. The route information server provides the total passage assurance of all links for the drivers using the RIS. Finally, the prospective traffic volume P T Vl of link l is defined as the product of ET Tl and (T P Al + 1.0). The driver using the RIS searches the shortest route in the prospective traffic volume from its current position to its destination. Information Sharing for Reduction of Traffic Congestion 5 Figure 1.2: (a) Lattice network and (b) radial and ring network 1.2 1.2.1 Computer Simulation Simulation Settings In order to evaluate our proposed route guidance mechanism, we performed simulation in the ratio of three types of route choice that the ratio of the drivers using the SD is fixed at 0.2, and that of the drivers using the ST and RIS are altered from ST:RIS = 0.8:0 to ST:RIS = 0:0.8. As the structure of road networks, we use two kinds of road networks, the lattice network and the radial and ring network in Fig. 1.2. In these road networks, all links had the same capacity. The origin and destination of a driver were assigned randomly to any block on any link. After reaching its destination, the vehicle is taken away from the network. At every simulation step, vehicles were generated until the amount of vehicles reaches to 25,000. In order to examine the effectiveness under the situation that roads are not vacant but a deadlock is not caused, the number of vehicles generated at one step Ngen is set at 45 in the lattice network, and at 35 in the radial and ring network. The travel time of each driver was normalized by the ideal travel time to compare our simulation results of the different road networks and the different sets of origins and destinations of vehicles. The ideal travel time was the time required from an origin to a destination when a driver passes through the shortest route for the distance at free flow speed. The travel time was defined as the ratio of the actual time taken to travel from an origin to a destination and the ideal travel time. 6 Information Sharing for Reduction of Traffic Congestion Figure 1.3: Average travel time in the lattice network 1.2.2 Simulation Results In estimating the results of simulations, we especially focused on a comparison of the average travel time of the drivers using the SD, ST, and RIS. The results of our simulation based on the average of 10 trials are shown in Figs. 1.3 and 1.4. In these graphs, the horizontal axis is the ratio of the drivers using the RIS, and the vertical axis is the average travel time. The ratio of the drivers using the RIS in all drivers is described as RRIS . The average travel time of the drivers using the SD, ST, and RIS is described as T SD , T ST , and T RIS . Fig. 1.3 shows the average travel time in the lattice network. From RRIS = 0.1 to RRIS = 0.3, T SD , T ST , and T RIS slightly increased. After that, all three decreased as RRIS increased. the average travel time of three types were always ranked in ascending order as T SD , T ST , and T RIS . There was always only marginal difference among them. Fig. 1.4 show the average travel time in the radial and ring network. T SD , T ST , and T RIS decreased as RRIS increased. The average travel time of three types were always ranked in ascending order as T SD , T ST , and T RIS . T RIS was always marginally better than T ST . The difference between T RIS and T SD was from 0.35 to 0.62. From these results, we observed that, in both networks, i) as the drivers using the RIS increased, the travel time of the drivers using the RIS and the other drivers substantially decreased, ii) the travel time of the drivers using the RIS was better than that of the drivers using other mechanisms. Therefore, we confirmed that the RIS mechanism was effective on improving the traffic efficiency of drivers. Here, we discuss the effectiveness of the RIS based o the difference of two networks. In the radial and ring network a driver had only one or two shortest distance routes. Because these shortest ones statistically tended to go through the innermost ring, the drivers using the SD tended to concentrate at the inner- Information Sharing for Reduction of Traffic Congestion 7 Figure 1.4: Average travel time in the radial and ring network most ring. The drivers using the ST and RIS could avoid traffic congestion in the innermost ring. However, the drivers using the ST often caused traffic congestion by avoiding the innermost ring and concentrating vacant links in the second and third innermost ring. On the other hand, the drivers using the RIS could prevent causing traffic congestion in the second and third innermost rings by route information sharing. Compared with doing in the radial and ring network, the drivers using the SD did not concentrate seriously at central links in the lattice network because the drivers using the SD had some shortest distance routes Because traffic congestion tended to be caused also at links other than central links, the drivers using the ST and RIS had more difficultly to preliminarily avoid congested links. Accordingly, however the drivers using the RIS increased, the travel time of the drivers using the RIS did not decrease monotonously, and the difference of the travel time among the SD, ST, and RIS was small. In our simulation, the lattice network we use in our simulations was very symmetrical and ideal, i.e., all links had the same capacity and length, and there were no one-way streets and traffic regulation. Although we can easily find the lattice network in New York and Kyoto, there are not so many shortest distance routes from an origin to a destination, and the links where the number of passing vehicles is more than traffic capacity, i.e., the bottlenecks, exist certainly. Therefore, lattice networks in practice may has also the characteristics similar to the radial and ring network that traffic congestion tend to be caused in certain bottlenecks. In such a network, it is important to prevent two types of traffic congestion. One is caused in the bottlenecks, and the other is caused in neighboring links by drivers avoiding first type of congestion and concentrating vacant alternative routes. Because, in the radial and ring network, route information sharing is regarded as the effective mechanism about preventing two types of traffic congestion, the RIS mechanism will be effective on reducing traffic congestion in actual traffic flow. 8 1.3 Information Sharing for Reduction of Traffic Congestion Conclusion In this paper, we proposed a route guidance mechanism with route information sharing. Three types of route choice behavior were prepared: the Shortest Distance (SD), the Shortest Time (ST), and the Shortest Time with Route Information Sharing (RIS). The effectiveness of the RIS mechanism was examined through multiagent simulation in the lattice network and the radial and ring network. From the results of simulation, we confirmed that the RIS mechanism was effective on improving the traffic efficiency of drivers. Bibliography [1] Klugl, F., Bazzan, A.L.C., Wahle, J.: Selection of information types based on personal utility: a testbed for traffic information markets. In Proceedings of the second International Joint Conference on Autonomous Agents and Multiagent systems (2003) 377-384 [2] Kurumatani, K.: Mass User Support by Social Coordination among Citizens in a Real Environment. In Multiagent for Mass User Support, LNAI 3012, Springer (2004), 1-19 [3] Kurumatani, K.: Social Coordination with Architecture for Ubiquitous Agents: CONSORTS. In Proceedings of International Conference on Intelligent Agents, Web Technologies and Internet Commerce 2003 (CD-ROM) (2003) [4] Mahmassani, H. S., Jayakrishnan, R.: System Performance and User Response under Real-Time Information in a Congested Traffic Corridor. Transportation Research 25A(5) (1991) 293-307 [5] Shiose, T., Onitsuka, T., Taura, T.: Effective Information Provision for Relieving Traffic Congestion. In Proceedings of 4th International Conference on Intelligence and Multimedia Applications (2001) 138-142 [6] Tanahashi, I., Kitaoka, H., Baba, M., H. Mori, H., Terada, S., Teramoto, E.: NETSTREAM, a Traffic Simulator for Large-scale Road Networks, R & D Review of Toyota CRDL, 37(2) (2002) 47-53 (in Japanese) [7] Teramoto,E., Baba, M., Mori, H., Asano, Y., Morita,H.: NETSTREAM: Traffic Simulator for Evaluating Traffic Information Systems. In Proceedings of IEEE International Conference on Intelligent Transportation Systems ’97 (CD-ROM) (1997) [8] Yoshii, T., Akahane, H., Kuwahara, M.: Impacts of the Accuracy of Traffic Information in Dynamic Route Guidance Systems. In Proceedings of The 3rd Annual World Congress on Intelligent Transport Systems (CD-ROM) (1996) [9] http://www.vics.or.jp