Trajectory representation learning with multilevel attention for driver identification

Mengyuan Li, Yuanyuan Zhang, Yaya Zhao, Yalei Du, Xiaoling Lu

Abstract: Highlights•Providing a trajectory representation method for driver identification.•Using a multilevel attention mechanism to extract multilevel features.•Integrating all available motion, spatial and temporal features together.•Applying a multi-loss to optimize the model.•Providing better performance on empirical analysis than benchmarks.