【摘要】 与传统的集计评估模型不同，智能体仿真模型能够捕捉个体之间的行为交互，在交通政策精细化评估中具有优势。但智能体仿真模型对数据要求较高，在实践中面临活动计划生成、参数标定和结果验证等难点。基于MATSim建立了在有限数据条件下面向应用研究的可操作、可复现的智能体仿真模型，并且以上海市拥堵收费评估场景为例详细介绍了模型的建模流程和关键技术。重点介绍基于手机轨迹数据的活动计划生成方法以及模型的参数标定和结果验证细节。结果表明，智能体仿真模型能够满足交通政策评估需求，基准场景的上海市域出行方式划分平均相对误差为8%、快速路交通量平均相对误差为17%。进而从集计和个体两个层面对仿真结果进行分析，展示了智能体仿真模型的应用潜力。更多还原

【Abstract】 Unlike traditional aggregate evaluation models, agent-based simulation models can capture the behavioral interactions among individuals, offering advantages in the refined evaluation of transportation policies. However, agent-based simulation models require high-quality data and face challenges in practice, such as generating daily activity schedule, calibrating parameters, and validating results. This paper establishes an operational and reproducible agent-based simulation model based on MATSim, designed for application research with limited data. Using the congestion charging evaluation scenario in Shanghai as an example, the modeling process and key techniques are elaborated. The focus is on daily activity schedule generation methods based on mobile phone trajectory data and parameter calibration and results validation details of the model. The results show that the agent-based simulation model is capable of meeting the transportation policies evaluation needs, with an average relative error of 8% for travel mode classification and 17% for estimation of traffic volume on expressways in the baseline scenario of Shanghai. The simulation results are analyzed at both aggregate and individual levels, demonstrating the application potential of the agent-based simulation model.更多还原