ABSTRACT

This study aims to evaluate and quantify the impact of demand forecast spatial resolution on the operational performance of a shared-use automated vehicle (AV) mobility service (SAMS) fleet. To perform the evaluation, this study employs an agent-based modeling framework that includes user requests, AVs, and an SAMS fleet controller. In the simulation, an SAMS fleet controller dynamically assigns AVs to on-demand user requests and repositions empty AVs throughout the service region to serve expected future demand requests. The fleet controller uses an offline demand forecast model and an online optimization model that jointly assigns AVs to users and repositioning trips. Results indicate that despite demand forecast quality decreasing at higher spatial resolutions, the operational efficiency of the SAMS fleet increases with higher spatial resolution forecasts (i.e. smaller subareas). Results also indicate that there is a significant operational value associated with improving short-term demand forecasts at high spatial resolutions.

摘要

这项研究旨在评估和量化需求预测空间分辨率对共享自动驾驶汽车（AV）机动性服务（SAMS）车队的运营性能的影响。为了进行评估，本研究采用了基于代理的建模框架，该框架包括用户请求，AV和SAMS车队控制器。在模拟中，SAMS车队控制器会动态地将AV分配给按需用户请求，并在整个服务区域中重新定位空的AV以服务预期的未来需求请求。车队控制器使用离线需求预测模型和在线优化模型，共同将AV分配给用户并重新安排行程。结果表明，尽管在较高的空间分辨率下需求预测质量下降，随着更高的空间分辨率预报（即较小的分区），SAMS机队的运营效率将提高。结果还表明，在高空间分辨率下改善短期需求预测具有重要的运营价值。