A Collaborative Car Auto-Navigation Framework Based on Intelligent Trajectory Mining

Abstract

Vehicle trajectories are widely used in varies of applications. As to car navigation applications, typical usages of trajectories include correcting the digital maps, finding out the traffic jams, looking for the locations of cars. In this paper, it is assumed that trajectory data of expert drivers implies “best practice” routes, which will be helpful to naïve drivers during driving. Therefore we propose a new idea called “collaborative navigation” in which the best practice routes are mined from trajectories and are send to naïve drivers when needed. The major difficulties of this approach are that the trajectory data may be too large and the data processing time may be too long. To address the problem, we design a framework which covers the whole life cycle of trajectories processing, including original data collection, “point to route” conversion, “best route” mining, and route query. Corresponding algorithms, data structures and data indexes are devised for each step in the life cycle. Experiments show “collaborative navigation” can be used to enhance routing selection with small footprint and quick response time using our framework.