Driver Lane Change Intention Recognition Based on Attention Enhanced Residual-MBi-LSTM Network

Accurate identification of lane-changing intention can effectively assist intelligent driving vehicles in terms decision-making and trajectory planning, which plays a significant role enhancing safety by reducing traffic accidents caused lane-changing. Based on the characteristics vehicle interaction information, an attention-enhanced bidirectional multi-layer residual long-short term memory neural network (Attention Enhanced Residual-MBi-LSTM) model is proposed for lane change recognition this paper. Firstly, EWMA filter employed to smooth noisy data collected from vehicle. Then four-layer LSTM (Residual-MBi-LSTM) structure used extract features historical trajectories ego-vehicle information. Besides, attention mechanism added adjust weight different time frames. After that, current probability calculated output Softmax function. Finally, firstly trained then verified HighD dataset. According HIL experiment, has ability identify driver average 2.07 seconds advance.