基于城市多源数据的海量轨迹分析关键技术研究

With the rapid advance in mobile sensing, positioning technology and mobile internet, large-scale trajectory data has been one important kind of big data which have stable sources and are easily accessible. Trajectory data contain rich and valuable information about individuals, crowds and urban dynamics, and thus analytics on trajectory data is of great importance to a wide spectrum of real-world stunning applications. Current analytic techniques of big trajectory data are subject to several limitations, such as failure to exploiting multi-source urban data, inefficiency in dealing with heterogeneous trajectory data, limited parallel query processing, and limited ability in recognizing semantics of trajectories. In response to these limitations, this proposal aims to carry out research on the following problems, including 1) a unified storage and querying in-memory system for heterogeneous trajectory data, 2) multi-source urban data based mobility modelling and trajectory prediction, 3) multi-source urban data based semantics labelling of trajectory data, and 4) heterogeneous trajectory data based map updating and its system design. The project team has accumulated rich experience in trajectory data collection, analytics and applications. Being part of the Shanghai Big Data Technology and Innovation Center, we strive to deliver four key technologies upon the completion of this project. Through partnership with several government agencies and enterprises (e.g., Shanghai Transport and Port Research Center), our research outcomes will serve important sections such as transportation, insurance and smart cities.

随着移动感知技术、定位技术及移动互联网技术的不断发展和广泛应用，轨迹数据已经成为来源稳定且较易获取的一类重要大数据。轨迹数据蕴含移动个体、群体及城市动态的丰富信息，轨迹数据的分析具有重要的实际应用价值。现有的工作在多源数据融合能力、异构轨迹数据的统一存储、轨迹查询的并行化水平、轨迹语义获取方法等方面还存在较大不足。针对以上不足，本项目拟开展四个方面的研究：1）基于内存的异构轨迹数据统一存储及并行查询框架与系统，2）基于城市多源数据的运动建模及轨迹预测方法，3）基于城市多源数据的轨迹语义标注方法，4）基于异构轨迹数据流的地图更新方法与系统设计。申请人在轨迹数据获取、分析及应用的相关领域有良好的研究基础，依托上海市大数据技术与应用创新中心等研究基地，项目的成功开展将形成四项融合城市多源数据的轨迹数据分析关键技术，通过上海市港航发展研究中心等政府和企业平台，服务交通、保险及智慧城市等应用领域。

轨迹大数据蕴含移动个体、群体及城市动态的丰富信息，轨迹数据的分析具有重要的实际应用价值，本项目重点研究基于城市多源数据的轨迹存储、分析和建模关键技术。项目的研究团队完成了任务书中的研究任务，达成了所有的研究目标。代表性的研究成果包括：设计了基于GPU的轨迹统一存储及并行查询关键技术，提出了融合城市多源数据的轨迹个性化深度预测模型，开发了基于城市多源数据的轨迹语义标注及模式挖掘关键技术，建立了融合群智数据和轨迹数据的时空数据预测模型，提出了基于用户交互轨迹的用户建模及预测技术。在学术论文发表方面，项目组共发表致谢项目资助的国际会议论文15篇和国际期刊论文20篇，包括IEEE Transactions on Knowledge and Data Engineering、ACM Transactions on Information Systems等顶级国际期刊，以及ACM KDD、IJCAI等顶级国际学术会议，其中一篇论文入选ESI高被引论文。在发明专利方面，申请发明专利10项，其中4项获得授权。在人才培养方面，培养3名研究生获得博士学位、8名研究生获得硕士学位。基于本项目的部分科研成果，项目负责人入选中组部青年拔尖人才，荣获上海市科技进步二等奖（排名第五）。本项目研发的关键技术对城市交通管理、个性化精准推荐等应用领域重要的意义。