基于张量模型的交通数据重建方法研究

Missing and corrupted data in traffic information collection system has a great impact on the performance of Intelligent Transportation System (ITS). Based on the inherent multi-correlation property of traffic data, this project plans to conduct research on reconstruction of unreliable (missing and corrupted) traffic data from the point of employing the redundant information. First, this project analyzes the inner properties of traffic data to establish a tensor model for characterizing (and/or incorporating) its multi-correlation property, and then explores the methods of multi-mode weight distribution for subsequent reconstruction.Second, we consider three types of data, namely, single point, single link, and road network. For each type, this project aims to build reconstruction models through combining the traffic theory and tensor reconstruction theory. The reconstruction models can be classified into three categories, tensor completion model for missing traffic data, tensor recovery model for corrupted traffic data, and tensor reconstruction model for the hybrid case(i.e., both missing and corrupted data exist simultaneously). Third, the project studies fast and efficient solutions for the reconstruction models to meet the requirement of traffic information processing.To summarize, this project aims to provide a systematic reconstruction method for unreliable traffic data. The resulting technology is expected to serve as a building block for intelligent traffic information processing and analyzing. In addition, it can be applied to other research areas as a theoretic reference to analyze and/or process unreliable data.

交通信息采集系统中存在的数据丢失和污染问题严重影响智能交通系统的性能。基于交通数据固有的多关联性特征，本项目从信息利用的角度出发，对不可靠（丢失和污染）交通数据的重建问题展开研究。分析交通数据特性，构建表征交通数据多关联性的张量模型，探索交通数据多模式权重分配方法；针对单点、路段和路网三种交通数据形式，融合交通理论和张量重建理论，建立丢失交通数据张量填充模型、污染交通数据张量恢复模型、以及同时存在丢失和污染情况下的交通数据张量重建模型；研究上述重建模型的快速求解算法。本项目研究旨在形成较为系统的不可靠交通数据重建技术与方法，为智能交通信息处理与分析提供基础，同时为其他领域的不可靠数据处理和分析提供借鉴和参照。

交通信息采集系统中存在的数据丢失和污染问题严重影响智能交通系统的性能。基于交通数据固有的多关联性特征，本项目从信息利用的角度出发，对不可靠（丢失和污染）交通数据的重建问题展开研究。分析交通数据特性，构建表征交通数据多关联性的张量模型，探索交通数据多模式权重分配方法；针对单点、路段和路网三种交通数据形式，融合交通理论和张量重建理论，建立丢失交通数据张量填充模型、污染交通数据张量恢复模型、以及同时存在丢失和污染情况下的交通数据张量重建模型；研究上述重建模型的快速求解算法。本项目研究旨在形成较为系统的不可靠交通数据重建技术与方法，为智能交通信息处理与分析提供基础，同时为其他领域的不可靠数据处理和分析提供借鉴和参照。