基于Attention-BiLSTM网络的车辆换道意图识别

Vehicle lane change intention recognition based on attention-bilstm network

针对换道意图识别方法仅考虑车辆历史状态信息,未充分利用车辆连续性和时序性特征的问题,提出了一种基于Attention-BiLSTM网络的换道意图识别方法。首先,分析行驶车辆之间的交互行为,采用双向长短期记忆网络学习换道意图特征编码信息;其次,通过引入模拟人脑推理行为的注意力机制进行网络权重自适应分配,提高网络捕捉重要状态信息能力;最后,利用HighD车辆轨迹数据集对模型进行训练和评估。试验结果表明:所提出的Attention-BiLSTM模型与LSTM模型相比,其准确率和F1分数分别提高了13.2%和10.5%,有效提升网络对周围车辆换道意图的识别性能。

Aiming at the problem that the lane-changing intention recognition method only considers the historical state information of the vehicle and does not utilize completely the vehicle characteristics with respect to the continuity and timing,a lane-changing intention recognition method based on Attention-BiLSTM network is proposed.Firstly,the interaction behavior between driving vehicles is analyzed and the bidirectional long short-term memory network is adopted to learn the coded information about the intention to change lanes.Secondly,improving the network's ability to capture important state information by introducing an attention mechanism that simulates the reasoning behavior of the human brain for adaptive distribution of network weights.Finally,the HighD vehicle trajectory dataset is employed to train and evaluate the models.The experimental results show that the accuracy and F 1 score of the proposed Attention-BiLSTM model are increased by 13.2%and 10.5%,respectively,which effectively improves the network's recognition performance of the lane-changing intentions of surrounding vehicles,compared with the LSTM model.