拥塞环境下的交通流量估计、预测及诱导关键技术研究

With the explosive growth of the number of vehicles in China, the urban traffic congestion problem is becoming more and more serious in recent years. Traffic congestion consumes enormous amounts of energy and causes vastly increased journey time, thus it has aroused great attention from the researchers, so it is urgent to design judicious and effective traffic control and induction strategies to address this contradiction. It is obvious that accurate traffic volume estimation and prediction is the premise of reasonable traffic control and guidance. However, the location and trajectory privacy of vehicles will be breached if we directly use the ID information of vehicles in the analysis. The existing studies on this problem, unfortunately, have many limitations such as: the privacy preservation problem is not considered; the level of privacy preservation is too weak; or the main impact factors on the traffic flow are not fully taken into account. Moreover, although the challenge of precisely measuring the traffic volume under a stringent privacy-preservation requirement has opened the door to a new world for the researchers, none of the existing studies can fully protect the vehicle privacy and estimate (or predict) the traffic volume accurately. To address the issues described above, this project will first study the problem of how to estimate the traffic volume accurately while achieving differential privacy at the same time. Then, we will focus on the traffic volume estimation method to infer the traffic volume on the roads without any traffic information (e.g., some roads with no cameras and RSUs installation) based on the traffic volume of related intersections, while taking the main impact factors on the traffic flow into consideration. After that, we will study the short-term traffic flow prediction problem based on the historical traffic information, and finally propose a traffic guidance mechanism to balance the load of the transportation network and the personal utilities of vehicles. Through the study of this project, we will provide a comprehensive set of traffic control and guidance techniques, which will be verified and optimized through the simulations based on real traffic traces.

近年来我国汽车保有量飞速增长，城市交通拥塞问题日益严重，亟需通过高效的交通流量疏导策略来缓解该矛盾。准确的交通流量估计与预测是进行合理交通疏导的前提，但直接利用路口采集得到的车辆信息进行流量分析会暴露车辆的位置及轨迹隐私。现有的交通流量估计与预测机制大都未考虑车辆的隐私保护问题或提供的隐私保护力度较弱，且对影响交通流量的主要因素考虑不全，导致提出的交通流量估计与预测机制无法很好地保护车辆隐私且精度较低。针对上述问题，本课题拟首先研究如何在保证差分隐私的基础上实现交通流量信息的精准采集与估计，然后利用估计得到的流量信息，结合影响交通流量的主要因素，对信息缺失路口交通流量进行补全，并利用估计得到的历史数据进行短时流量预测，最终根据估计和预测得到的流量信息，在均衡交通网络和车辆个体利益的基础上进行合理流量诱导，设计出一套完整的交通疏导解决方案，在真实交通数据集的基础上不断验证和优化。

本项目针对日益严峻的交通拥塞问题，研究了如何实现智能交通系统中的交通流量实时测量、预测，并基于流量信息给出合理、高效的诱导决策，从而实现交通资源的优化配置，从而缓解交通压力。因此，本项目首先针对交通流量实时测量问题，设计了一系列机制旨在保护车辆位置隐私和轨迹隐私的基础上，仅通过安装在车辆上的RFID标签和路口的阅读器交互，即可高精度地估计出交通单点流量、同时经过多路口的车流量及多周期单点及多点的持续车流量等，并通过理论分析证明所设计的机制可以为车辆提供差分隐私级的敏感信息保护；其次，结合多路口间车流量的时空相关性，设计出基于稀疏流量信息的交通网络数据补全技术，从而估计出未安装流量收集设备路口的车流量信息；此外，本研究还利用已有的出租车数据，结合天气、空气质量和路面维护建设等多种因素，针对性设计了高效的短时流量预测机制，为交通疏导提供可靠的依据；最后，针对交通诱导问题，结合不同类型车辆的移动路线特点、道路的承载能力以及交通流量的变化预测，设计了能够实现负载均衡的交通流量分配方式，尽可能地将车辆从拥堵路段引导至流量相对较轻的路段，从而达到交通网络负载均衡的目的。通过本项目的研究，所产生的相关研究成果共发表论文55篇，其中SCI 期刊论文19 篇，EI收录的文章28篇，申请发明专利5项，其中还包括了19篇CCF A类期刊会议论文。最后所形成的一套完整的交通流量估计、预测及诱导技术为解决严峻的城市交通拥塞问题提供了强有力的理论基础和实际指导意义。依托本项目取得的科研成果，主持人获批了一项国家级青年人才计划支持。