融合多准则考量的路网信控鲁棒仿真优化方法

With growing attention paid to the traffic safety, vehicular emissions and equity of right of way, the management and control target for signalized intersections gradually changes from “Smooth” to “Smooth, Safe, Green, Harmonious”. This project studies the multi-criteria oriented robust optimization method for road network signal control, based on the simulation-based optimization framework. Firstly, the multi-fidelity traffic simulation platform is built, and then calibrated and validated from various aspects with multi-source heterogeneous traffic data. This aims to guarantee the reliability and generalization of this traffic simulation platform. Later, with the reasonable choices of evaluation methods for traffic efficiency, safety, emissions and equity, the multi-objectives oriented road network signal control optimization model is constructed under uncertainties, along with which the multi-fidelity traffic simulation platform is used to set up the simulation-based optimization system. After that, with the introduction of switch mechanism between macro- and micro- traffic simulation models, surrogate models and their updating management mechanism, the low-cost multi-objective robust optimization algorithm is designed to obtain the Pareto optimality with respective to multi-criteria. Finally, the signal control parameters and consequents under various scenarios are compared and analyzed. These methods and results will provide theoretical basis and technical foundation for robust optimization of signal control that satisfies the diversified management and control objectives.

随着交通安全、尾气排放及路权公平性问题日益获得重视，信号交叉口的管控目标逐渐从“畅通”过渡到“畅通、安全、绿色、和谐”。本课题基于仿真优化框架，研究融合多准则考量的路网信控鲁棒优化方法。首先，搭建多精度交通仿真平台，并采用多源异构交通数据从多角度对其进行标定和验证，保证交通仿真平台的可靠性与推广性。然后，选取合理的效率、安全、排放与公平评价方法，建立不确定环境下多目标路网信控优化模型，并与多精度交通仿真平台构成仿真优化系统。之后，引入宏、微观交通仿真模型的切换机制、替代模型及其更新管理机制，设计低成本的多目标鲁棒优化算法，获取关于多准则考量的帕累托最优解，并比较分析不同场景下信号控制参数及控制效果的差异，为满足多样化管控目标的信控鲁棒优化提供理论依据和技术支撑。

随着交通安全、尾气排放及路权公平性问题日益获得重视，信号交叉口的管控目标逐渐从“畅通”过渡到“畅通、安全、绿色、和谐”。本课题基于仿真优化框架，研究融合多准则考量的路网信控鲁棒优化方法。项目主要在四个方面取得了重要进展：1）综合考虑微观交通仿真模型的高计算精度与宏观交通仿真模型的低计算成本，搭建多精度交通仿真平台；利用COM接口技术实现宏、微观交通仿真模型与数值计算软件之间的数据通信，完成多精度交通仿真优化平台的搭建；基于多源异构的交通数据对宏、微观交通仿真模型从多角度进行标定与验证，保证交通仿真评价的可靠性。2）兼顾阿特金森系数或基尼系数、平均旅行时间、车辆尾气排放和代理安全评价因子等考量，以干线/路网的信控方案（包含相序、信号周期、绿灯时间以及相位差）为决策变量，在干线与路网层面建立融合多准则考量的信控随机/鲁棒仿真优化模型，为满足多样化管控目标的信号控制方案设计提供理论模型依据。3）设计了一系列基于代理模型的多目标仿真优化算法，包括确定环境下的多目标仿真优化算法与不确定环境下的多目标随机/鲁棒仿真优化算法。通过引入代理模型的更新管理机制，使得算法能够平衡交通仿真的计算成本与代理模型的预测评价精度，为解决多目标信控仿真优化问题提供有效可行的求解算法，也可移植到其他工程领域类似优化问题的求解。4）建立以多目标仿真优化模型和多精度交通仿真平台为上、下层，以低成本的多目标仿真优化算法为纽带的交通仿真优化系统。该系统可以有效解决不确定环境下融合多准则考量的信控仿真优化问题，为城市道路交通系统高效、安全、低碳和公平地运行提供科学合理的信控方案。