Single-lane traffic simulation with multi-agent system

Agent-based traffic simulation has emerged as an efficient tool to investigate traffic phenomenon. However, the main problem is how to reproduce realistic patterns of traffic flow at both macroscopic and microscopic levels with restricted computational resources. In this paper, we present a multi-agent framework of single-lane traffic simulation. We focus on the driver-vehicle agents, which is the most important ones in the framework. The three-layer architecture of driver-vehicle agents is proposed, the decision tree of tactical longitudinal movement is put forward, and the simplified formulas of acceleration and deceleration rate are derived. With parallel update mode of agent states, a one-dimension traffic simulation model is developed. To validate it at the macroscopic level, we reproduced the realistic flow-density and speed-density relation of traffic flow with periodic boundary condition. To validate it at the microscopic level, a platoon of 10 vehicles was simulated. Given four kinds of typical speed profiles of the leader, each follower can follow its leader safely and stably. The result shows that this model can reproduce realistic macroscopic and microscopic characteristics of single-lane traffic flow.