Vehicle trajectory prediction combining bidirectional LSTM and attention mechanism
To improve management and service quality in intelligent transportation,autonomous driving and other systems,we propose a bidirectional long short-term memory network combined with self-attention model for vehicle trajectory predic-tion.Our approach adopts Douglas-Pook compression algorithm to compress and pre-process trajectory data,reducing data redundancy and optimizing data processing.The encoder uses a bidirectional long and short-term memory network to fully capture temporal correlation features,while the self-attention mechanism extracts global spatial correlation features from sur-rounding vehicles.The future position of the vehicle is obtained through the fully connected layer of the decoder,and the complete predicted trajectory route is obtained through model iteration.Experimental results show that the prediction per-formance of the proposed model is better than that of the comparison model.In addition,the results of ablation experiments show that the introduction of the trajectory compression algorithm and the improved long short-term memory network com-bined with the attention mechanism have a positive contribution to prediction accuracy.
国家科技期刊平台
NSTL
万方数据
