基于网联车辆的按需出行智能交通系统运行优化

This project investigates novel scheduling modes and methods of autonomous mobility-on-demand (MoD) transportation and regional logistics systems for use in metropolises based on connected vehicles. The emphasis is focused on some fundamental common scientific problems in modeling, operational optimization, performance and scale evaluation of autonomous mobility-on-demand intelligent transportation systems. The main contents of the project are as follows. (1) The mechanism and mathematical modeling of mobility-on-demand systems: seeking for system architecture and modeling methods (based on methods like network flow, queuing theory and distributed state space model) that are convenient for analysis and optimization of MoD transportation systems. (2) Methods for mobility demand forecasting: presenting dynamic mobility forecasting schemes for metropolises using methods such as frequency statistics, deep reinforcement learning based on historical dada. (3) Methods for dynamic rebalance scheduling and routing of vehicles: finding out high efficiency low cost methodologies for optimal operational scheduling, vehicle cooperative control, system scale analysis and performance evaluation, by constructing optimization cost function on the basis of the general cost, service quality and traffic efficiency of MoD transportation systems. (4) Experimental platform of autonomous mobility-on-demand transportation and logistics systems: developing technical index system and verification testing platform of shared mobility-on-demand transportation systems.

本项目旨在探讨基于网联车辆的按需出行城市交通及区域物流系统调度新模式及新方法，重点研究自主按需出行智能交通系统的建模、运行优化、性能及规模评价等方面的重要共性科学问题。主要研究内容包括：(1) 按需出行系统的机理和数学建模。拟探讨用网络流、队列理论和分布式状态空间法等方法，建立便于按需出行交通系统分析和优化的系统架构和建模方法。(2) 出行需求的预测方法。拟基于历史数据，寻求用频率统计、深度强化学习等方法获得城市出行需求的动态预测方法。(3) 动态再平衡车辆调度和路径规划方法。根据移动出行交通系统的总体成本、服务质量和交通效率构造优化指标，研究交通系统高效低成本运行优化调度、车辆协调控制、系统规模分析及性能评价方法体系。(4) 自主按需出行交通与物流系统实验平台。构建共享移动出行交通系统技术指标体系和验证测试平台。

本项目探讨了基于网联车的城市交通物流系统的新模式及新方法，重点对车辆共享按需出行系统的建模、运行优化、性能及规模评价等重要科学问题进行了深入研究。主要研究进展包括：(1) 按需出行系统的机理和数学建模。基于流动模型、排队理论和分布式状态空间方程，建立了按需出行系统的基本架构和机理模型，作为分析和优化设计的基础。(2) 城市交通流与出行需求的时空分布预测。基于历史交通出行数据和外部数据的统计分析，构建了短期交通流和出行需求预测的峰值敏感深度强化学习架构。(3) 系统动态再平衡调度与车辆路径规划。以按需出行系统的成本和 乘客等待时间（决定服务质量和出行效率）为指标，提出高效低成本的运行优化调度、充电调度、车辆协调控制、系统规模与性能评价方法体系。(4) 按需出行交通物流系统实验平台。构建了支持共享出行系统的验证测试平台和相关技术指标体系。研究成果发表IEEE Transacctions、Control Engineering Practice、Transportation Research等国际顶级期刊论文19篇，ESI高被引论文8篇，ESI热点论文1篇，授权发明专利9项，获河北省自然科学一等奖和中国自动化学会自然科学二等奖。