基于Swarm和机器学习的城市道路交通系统多主体仿真研究

Multi-Agent Simulation of Urban Traffic System Based on Swarm and Machine Learning

摘要: 城市道路交通系统仿真是分析和优化现有城市交通系统的重要方法之一。本文设计了基于Swarm和DQN机器学习算法的多主体仿真模型。首先，基于复杂适应系统理论，建立了多主体仿真模型。对于系统中主体的个体决策行为采用机器学习中的DQN算法来模拟，并基于Swarm软件包实现了所设计的仿真模型。最后，使用滴滴出行盖亚数据开放计划中成都市二环线上的数据对所建立的模型进行了仿真和分析，结果表明了方法的有效性。

Abstract: Urban road traffic system simulation is one of the important methods to analyze and optimize the existing urban traffic system. In this paper, a multi-agent simulation model based on Swarm and DQN machine learning algorithm is designed. Firstly, based on complex adaptive system theory, a multi-agent simulation model is established. Then the DQN machine learning algorithm is used to simulate the individual decision-making behavior of the agent in the system. The simulation model is implemented based on Swarm software package. Finally, the model is simulated and analyzed by using the data from the second ring line of Chengdu City in the Gaia Data Open Plan. The result illustrates the effectiveness of this method.