A Pattern-Based Predictive Indexing Method for Distributed Trajectory Databases

Recently, it has become possible to collect large amounts of trajectory data of moving objects by using sensor networks. To manage such trajectory data, we have developed a distributed trajectory database composed of a server and many sensor nodes deployed over wide areas. The server manages the trajectory data of each moving object by using indices. However, since each sensor node cannot send trajectory data to the server all the time, the server does not always manage indices for the current trajectory data. In other words, the server is delayed in answering queries for current data because it has to forward each query to the sensor nodes to answer them. This is defined as a delay problem. To avoid this problem, we propose a pattern-based predictive indexing method for the database to answer queries in real time. This method uses past motion patterns of moving objects to predict the future locations of moving objects. In this paper, we describe the method and evaluate it with practical trajectory data. We conclude that the technique can predict the future locations of moving objects well enough in real time and show optimal parameters for prediction.