Title: Vehicle lane change trajectory learning and prediction model considering vehicle interactions and driving styles in highway scene
Authors: Chenyu Song; Zhizhong Ding; Wanli Xu
Addresses: College of Engineering, Northeastern University, 360 Huntington Ave, Boston, MA 02115, USA ' School of Computer and Information, Hefei University of Technology, Hefei, 230009, China ' School of Information Science and Technology, University of Science and Technology of China, Hefei, 230026, China
Abstract: Accurate lane change trajectory prediction is of great significance for improving driving safety in the highway scene. However, current methods ignore the fact that the driver's subjective manipulation fundamentally affects the trajectory. By analysing our real driving data on highways, we found that the vehicle interaction and the driver's driving style affected driver's operation heavily. Therefore, the innovation of this paper is to comprehensively consider the influence of vehicle interaction and driving style on the lane change trajectory to improve the accuracy of lane change trajectory prediction. The vehicle interaction and the driver's driving style are modelled jointly as a BiLSTM-based module and a cluster-based module to reduce prediction uncertainty. The HighD dataset is used to test the model, which shows that the mean absolute error of displacement prediction is reduced by nearly 24%, which achieves better results than some other existing models.
Keywords: intelligent transportation system; lane change trajectory prediction? advanced driving assistance system; self-organising mapping; SOM? bidirectional long short-term memory; BiLSTM? self-attention mechanism.
DOI: 10.1504/IJSNET.2023.134906
International Journal of Sensor Networks, 2023 Vol.43 No.3, pp.172 - 183
Received: 03 Apr 2023
Accepted: 08 Jun 2023
Published online: 16 Nov 2023 *