Lane Change Trajectory Prediction for Intelligent Vehicles Based on an Improved Transformer

Lane change trajectory prediction serves as a critical component in intelligent driving systems, acting as a bridge between perception and decision-making, and plays a vital role in enhancing driving safety. While existing studies predominantly emphasize predictive accuracy, computational efficiency is often overlooked. To address this issue, an improved Transformer-based model is proposed for lane change trajectory prediction in intelligent vehicles. The model incorporates the ProbSparse self-attention mechanism to replace the conventional attention structure and integrates attention distillation to selectively retain dominant attention scores from the encoder. This design significantly reduces computational complexity when handling long input sequences and improves overall inference efficiency. Experimental results demonstrate that the proposed model achieves a favorable balance between accuracy and computational performance. Comparative evaluations against several classical time series prediction models further highlight the superiority of the proposed approach in the context of lane change trajectory forecasting.