Introducing Driving Behavior Model into Large-scale Multi-agent Traffic Simulation

i 2 1 2 – 2 1. 1 ii 2. 2 OD 1. 1 2 2 MATSim-T MATSim-T 2. Abstract Agent-based simulation is recognized as a useful technology for understanding social systems and social phenomena. In order to make a social simulation have more realism, it is effective to assign a real human behavior to each agent which makes up the virtual society. In addition, it is desirable to conduct a simulation on as a large scale as the social phenomenon to reproduce. Thus large-scale multi-agent simulation is a promising technology which modern society demands. There are so many changes in Japan's road traffic such as aging of the drivers or the widespread use of new traffic systems. New technologies are needed for predicting the problems in the future road traffic or consideration of a road traffic policy to solve the problems. Unlike pilot programs in real world, simulations are useful in examining the effect of a road traffic policy as many times as the policymaker wants. There are two major approaches to a traffic simulation. One approach is to reproduce the behavior of each vehicle driven by the microscopic model. Each vehicle has the model which can determine its driving speed, lane change, turning left or right, street parking, and how to avoid obstacles. Another approach is to reproduce traffic jams or routing of drivers on a large-scale road network. Each road link has the model which can calculate the position of vehicles on the link according to the relationship between traffic volume and density. Road traffic is a social phenomenon in which a number of individual drivers participate. Thus both a macroscopic point of view and a microscopic point of view are needed when the policymaker examine the policy. The challenges of this study are as follows. 1. Development of large-scale traffic simulation platform based on driving behavior model The first challenge is to develop a simulation platform which allows users to see urban traffic from a microscopic point of view and a macroscopic iv point of view. The platform must be able to reproduce large-scale traffic at the micro level. Each driver agent in the platform has its own driving behavior model to recognize the environment and drive its vehicle. 2. Platform verification by example simulations The second challenge is to show the potential of the platform by examples. The configuration of an example simulation contains a …