Real-time Management of Large Fleets of Self-Driving Vehicles Using Virtual Cyber Tracks

使用虚拟网络轨道实时管理大型自动驾驶车队

• 批准号: 1663657
• 负责人: Xuesong Zhou
• 金额: $ 40万
• 依托单位: Arizona State University
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2017
• 资助国家: 美国
• 起止时间: 2017-09-01 至 2020-08-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1663657&HistoricalAwards=false
• 关键词: Real; time; Management; Large; Fleets

It is only a matter of time before the transportation infrastructure of freeways, roads, and traffic control systems must accommodate self-driving vehicles (SDVs) at the same time as manually driven vehicles (MDVs). In large scale systems that exist in nearly all metropolitan areas, the question is how can we efficiently, reliably and safely accomplish this? This project aims to generate fundamental knowledge needed to design such a system. More specifically, the project will study and develop decision models and algorithms, and attendant decision-support systems to manage, in real time, large fleets of SDVs and MDVs on the current infrastructure, without the need to construct special roads or guideways. This project will assume that SDVs are cyber-connected and that, through cyber mechanisms of computing and communication, it is possible to guide these SDVs, both individually and in platoons, on our transportation infrastructure efficiently, without sacrificing comfort, safety and efficiency in mobility. The fundamental concepts in the decision-support architecture are (a) controlling directions, speeds and stops to individual SDVs in real-time, (b) grouping SDVs in platoons, (c) moving SDVs in platoons, with short and uniform headways, using the concept of cyber-enabled virtual tracks on the roads, and (d) providing traffic signals on the roads and blocking control (platooning sizing and dispatching) on the virtual tracks to maximize throughput and other desirable traffic performance measures. In addition to the tools developed for operating SDVs in real time, this project will help transportation agencies to efficiently satisfy increasing transportation demand with limited road infrastructure expansion and constrained road capacity. Finally, the research and tools will be integrated into the current and new courses and laboratories for computer science, operations research, and transportation engineering courses.This project will address fundamental knowledge in networking self-driving agents at extremely large scales to meet temporally and spatially distributed traveler demand. The goal is to develop a set of new models for integrated traveler mobility optimization and multi-agent-based control under the new environment of shared SDV networks. It will investigate a novel cyber-track based concept and methods that optimally provide real-time guidance to meet temporally and spatially distributed traveler demand for SDVs (from origins, to intermediate platoons, to destinations), possibly leading to new large scale nonlinear optimization methods that include vehicular dynamics and safety/comfort consideration. By taking full advantage of distributed computing power associated with connected SDVs, the dispatching and operating system for SDVs will simultaneously route and control individual SDVs and platoons on existing highways and streets. Based on a space-time cyber track network modeling framework for representing physical transportation system with constraints, the project will also develop real-time algorithms for proactive control of traffic supply infrastructure (e.g., traffic signals, ramp meters, and traffic information and congestion pricing) that optimizes delays and other performance metrics for both SDVs and MDVs. The project will integrate parallel computing and hierarchical system control, as well as a wide range of real-time vehicle routing/scheduling algorithms, including vehicle routing for platooning, block control and timetabling, to ensure the safety, efficiency and reliability of SDV operations. The project will also study the computational tractability of large scale deployment of SDVs using tools of cloud computing, big data management and parallel computation. The project will develop protocols of collecting steaming data from connected SDVs and managing, distributed computing, and effective logistics for SDV fleets.

高速公路、道路和交通控制系统等交通基础设施必须同时容纳自动驾驶车辆（SDV）和手动驾驶车辆（MDV），这只是时间问题。在几乎所有大都市地区都存在的大规模系统中，问题是我们如何有效，可靠和安全地实现这一目标？该项目旨在产生设计这样一个系统所需的基本知识。更具体地说，该项目将研究和开发决策模型和算法，以及随之而来的决策支持系统，以真实的时间管理现有基础设施上的大型SDV和MDV车队，而无需建造特殊的道路或导轨。该项目将假设SDV是网络连接的，并且通过计算和通信的网络机制，可以在我们的交通基础设施上有效地引导这些SDV，无论是单独的还是成排的，而不会牺牲舒适性，安全性和流动性的效率。决策支持架构中的基本概念是（a）实时控制各个SDV的方向、速度和停止，（B）将SDV分组为队列，（c）使用道路上的网络启用虚拟轨道的概念，以短且均匀的车头时距按队列移动SDV，以及（d）在道路上提供交通信号，并在虚拟轨道上提供阻塞控制（排队大小和调度），以最大化吞吐量和其他期望的交通性能测量。除了为真实的运行SDV开发的工具外，该项目还将帮助运输机构在道路基础设施扩建有限和道路容量有限的情况下有效满足日益增长的运输需求。最后，研究和工具将被整合到计算机科学，运筹学和交通工程课程的现有和新课程和实验室中。该项目将解决超大规模网络自动驾驶代理的基础知识，以满足时间和空间分布的旅行者需求。目标是在共享SDV网络的新环境下，开发一套新的模型，用于集成的旅行者移动性优化和基于多智能体的控制。它将研究一种新的基于网络轨道的概念和方法，最佳地提供实时指导，以满足时间和空间分布的旅行者对SDV的需求（从起点，到中间排，到目的地），可能导致新的大规模非线性优化方法，包括车辆动力学和安全性/舒适性的考虑。通过充分利用与连接的SDV相关的分布式计算能力，SDV的调度和操作系统将同时路由和控制现有高速公路和街道上的单个SDV和车队。基于用于表示具有约束的物理交通系统的时空网络跟踪网络建模框架，该项目还将开发用于交通供应基础设施（例如，交通信号、匝道仪表、交通信息和拥堵定价），优化SDV和MDV的延迟和其他性能指标。该项目将集成并行计算和分层系统控制，以及广泛的实时车辆路线/调度算法，包括队列车辆路线，块控制和调度，以确保SDV操作的安全性，效率和可靠性。该项目还将研究使用云计算，大数据管理和并行计算工具大规模部署SDV的计算可处理性。该项目将开发从连接的SDV收集流数据的协议，并为SDV车队管理，分布式计算和有效的物流。

项目成果

Dynamic programming-based multi-vehicle longitudinal trajectory optimization with simplified car following models

• DOI: 10.1016/j.trb.2017.10.012
• 发表时间: 2017-12
• 期刊: Transportation Research Part B-methodological
• 影响因子: 6.8
• 作者: Yuguang Wei;C. Avci;C. Avci;C. Avci;Jiangtao Liu;Baloka Belezamo;Baloka Belezamo;N. Aydin;P. Li;Xuesong Zhou

Observability quantification of public transportation systems with heterogeneous data sources: An information-space projection approach based on discretized space-time network flow models

• DOI: 10.1016/j.trb.2019.08.011
• 发表时间: 2019-10
• 期刊: Transportation Research Part B: Methodological
• 作者: Jiangtao Liu;Xuesong Zhou

Open-source VRPLite Package for Vehicle Routing with Pickup and Delivery: A Path Finding Engine for Scheduled Transportation Systems

• DOI: 10.1007/s40864-018-0083-7
• 发表时间: 2018-06
• 期刊: Urban Rail Transit
• 影响因子: 1.5
• 作者: Xuesong Zhou;L. Tong;M. Mahmoudi;Lijuan Zhuge;Yu Yao;Yongxiang Zhang;Pan Shang;Jiangtao Liu;Tie Shi

A stepwise interpretable machine learning framework using linear regression (LR) and long short-term memory (LSTM): City-wide demand-side prediction of yellow taxi and for-hire vehicle (FHV) service

• DOI: 10.1016/j.trc.2020.102786
• 发表时间: 2020-11-01
• 期刊: TRANSPORTATION RESEARCH PART C-EMERGING TECHNOLOGIES
• 影响因子: 8.3
• 作者: Kim, Taehooie;Sharda, Shivam;Pendyala, Ram M.
• 通讯作者: Pendyala, Ram M.

Integrated vehicle assignment and routing for system-optimal shared mobility planning with endogenous road congestion

• DOI: 10.1016/j.trc.2020.102675
• 发表时间: 2020-08
• 期刊: Transportation Research Part C-emerging Technologies
• 影响因子: 8.3
• 作者: Jiangtao Liu;P. Mirchandani;Xuesong Zhou