关键词: 通行时间估计, 轨迹数据, 时空索引, Cassandra

Abstract:

The travel time of a path implies the traffic condition in period of time. How to calculate this time accurately is of great significance in traffic monitoring and route planning. The existing studies usually use the taxi historical trajectories to estimate the travel time of a path. However, when the path is long, it's difficute to find a historical trajectory pass through the specified path completely, so that we can't estimate the travel time accurately. On the other hand, massive historical trajectories generate huge data management and computational pressure on path travel time estimation. For this reason, we introduce the geographic grid(Cell). First, we construct a unified spatio-temporal index, divide the trajectory data into a series of trajectory segments that fall in the geographic grid (Cell) and split the road network into a series of road patterns that fall in Cell. Then, the travel time of vehicle in current road pattern is represented as the residence time of a series of frequent shared trajectories in Cell. Finaly, we estimate the travel time of vehicles on longer routes using the passage time of a set of similar road patterns in historical periods. At the end of this paper, we verify the effectiveness of our method in dealing with massive historical trajectories and verify the superiority in estimating vehicle travel time on longer paths based on the GPS trajectories of 10 000 taxicabs over a period of one week.

Key words: travel time estimation, trajectory data, spatio-temporal index, Cassandra