面向智能交通的车联网时空数据流异常分析研究

Spatio-temporal data streams in Internet of Vehicles (IoV) including a great amount of information, this proposal aims to find out the traffic changes which have big spatial deviations from the normal modes in real-time by modeling and analysis. This not only discovers the anomalous event patterns hidden in big data streams, but also provides analysis for Intelligent Transportation System (ITS) and beneficial reference for the application of similar situation. This proposal plans to research on the traffic anomalous conditions real-time data minning and analysis under the environment of spatio-temporal data streams. We will build series of outlier models based on hybrid minning with the IoV data streams by analyzing the network of roads spatio-temporal relation for the short coverage and the poor real-time traffic monitoring. We will research to reduce the spatiotemporal complexity of the algorithm, and design an efficient algorithm to the anomalous detection and analysis over spatio-temporal data streams for classic algorithms are difficult to meet the real-time analysis. We will deploy cloud computing platform based on the present IoV platform to assess the proposed outlier models and distributed detection algorithms by experiment.

车联网时空数据流中包含丰富的交通信息，本课题拟通过建模分析从海量数据流中发现影响交通效率的异常事件，为智能交通的预测与控制提供分析基础。同时，为时空大数据流处理和分析提供重要的理论方法，为其他同类环境下的应用研究提供有益参考。因此，本课题对时空大数据流环境下道路交通的实时异常状态挖掘分析展开深入研究。针对现有道路交通状态监测中遇到的覆盖范围小、实时性低等问题，本课题拟基于车联网时空数据流，建立混合挖掘的异常状态模型，从数据角度揭示路网的时空关系，提供检测分析的特征模型；针对海量异构时空数据流环境下，经典方法难以满足实时分析需求的问题，研究降低分析方法的时空复杂度，设计面向智能交通的高效异常状态检测分析算法；利用现有车联网平台，构建基于云计算的实验与验证平台，通过实验评估上述模型和算法的性能和可靠性。

研究工作的主要内容包括车联网络的建设，数据的采集、融合和预处理，交通数据的挖掘分析三大部分。本项目以广州的城市交通作为研究对象，抽取具有代表性的珠江新城CBD交通数据进行分析。车联网络的建设由前期团队支撑和协同项目共同完成了相关线路的公交车联网络的建设和数据采集，并且通过其他途径获得大量的交通相关数据，进行融合分析。