Predictive Indexing for Position Data of Moving Objects in the Real World

This paper describes a spatial-temporal indexing method for moving objects with a technique to predict future motion positions of moving objects. To build efficient index structure, we had an experiment to analyze practical moving objects, such as people walking in a hall. As the result, we found that any moving objects can be classified to just three types of motion characteristics; 1) staying, 2) straight moving, and 3) random walking. Indexing systems can predict accurate future positions of each object based on our found characteristics, moreover, the index structure can reduce the cost to update MBRs in spatial-temporal data structure. To show an advantage of our prediction method to previous works, we had an experiment to evaluate performance of each prediction method.