Combined Models of a Atomic Route-choice Model and Cell-based Traffic Flow Simulation

Combined Models of a Atomic Route-choice Model and Cell-based Traffic Flow Simulation


Toshihiko Miyagi, Gifu University, Genaro Peque Jr., Tohoku University


We will present a comprehensive, sophisticated traffic simulation procedure, in which each driver’s route choice and cell-based traffic flow simulation is interactively connected at micro level.


One of the key features of intelligent transportation systems (ITS) is to allow an interactive communication between drivers and a Traffic Management Center (TMC): Each vehicle’s location sequences is transmitted to the TMC and each driver can get the information on the congestion level of each segment in road networks. In Japan, most automobile companies individually provide such services to their customers. The Ministry of Land use, Infrastructure, Transport and Tourism (MLITT) started a project for building effective transportation systems by using such big data of travel information that are managed separately among the automobile companies.
This paper is motivated with the needs of comprehensive and sophisticated traffic simulation procedures, in which drivers’ decision on route choice and traffic flow simulation is interactively connected. While traffic flows can be described by cell-based simulation tools, however, the existing simulator assumes inflow distribution to the network is given. On the other hand, the atomic models for route choice that has been developed by the authors are atomic models based on game theory and allow treating each trip-maker as the ones with different travel information. Therefore, it is not a surprising thing to consider a combination of the two streams.
In this paper, we provide two schemes for combining the atomic route-choice models and the cell based traffic simulator. In the first scheme, we assume that each trip maker can get travel time information on each section of road networks from the TMC. In addition, each vehicle is assumed to be equipped with a processor that can compute travel times between each driver’s origin and destination and predict the travel time when changing the current route. This implies that each driver knows the delay functions of each link. A driver can recognize his experienced travel-time on the current day: Each driver’s trajectory is traced by a cell automaton. Thus, there is one-to-one correspondence between a cell automaton and a trip-maker. The scheme may be called a synchronized simulation. In decision-making of today’s route he will use the computed travel times based on information provided by the TMC together with his realized travel time. However, those data are in hindsight, so a driver will be required to take forward-looking behavior when selecting today’s route choice. In addition, the experienced travel time of each driver at a certain period is possibly different from each other. Thus, we have to examine whether this scheme can provide a stable traffic flow or not. Even if it can successfully lead a stable traffic system, we must ask whether the resultant equilibrium is in accordance to the known equilibrium concept.
We also propose another scheme, in which we assume that drivers have no knowledge about the delay functions and that their decisions rely on travel information provided by the TMC. While at a first glance, this scheme looks like almost the same as the first scheme, however, the conditions of route-choice are much different from the previous scheme for his most reliable data source is the sequence of the realized travel times obtained through his own travel experiences. His decision-making will also be affected by his realized travel times and the announced travel times from the TMC. Since drivers don’t know the objective information on travel times, the resultant equilibrium is possibly different from the ones achieved by the first scheme.
In this paper, we will first outline our atomic models in relation with the interaction of travel information and route-choice, and then examine the validity of our models through applications of the two schemes to a simple road network with a single origin-destination pair.


Association for European Transport