Driving Behavior Identification of Operating Vehicles Based on Real-Time Trajectory Data

Trajectory data of operating vehicles contains abundant spatiotemporal information about driving behaviors. With the development of big data processing and mode identification techniques, these methods can be used to mine the patterns of driving behaviors from the trajectory data, and compensate the disadvantages of visual image analysis results. In this research, the trajectory data is partitioned by reactive greedy randomized adaptive search procedure into small segments, and driving behavior on the segments is identified by gated recurrent unit neural network model. The proposed method is verified by real-time trajectory data from Chongqing and Shanghai with over 370000 trajectory points, and the result shows that the accuracy of identifying operating vehicles’ driving behavior has reached 91.57%, which indicates that the proposed method improves performances in dealing with large scale trajectory data. This could be used to enhance the supervision on operating vehicles’ safety for transportation companies as well as transportation management departments, and improve operating vehicles transportation management strategies.