【摘要】 文章基于循环神经网络RNN(Recurrent Neural Network)和长短时记忆网络LSTM(Long-Short Term Memory)的理论研究,提出了一种基于LSTM的智能车变道行为预测模型。首先,搭建LSTM网络模型框架;然后根据人类驾驶场景对真实数据集NGSIM(Next Generation Simulation)进行特征选择与数据提取。最后使用长短时记忆网络(LSTM)模型进行训练,测试车辆变道预测结果,并将结果与利用RNN模型预测的结果进行比较,验证了本文方法的有效性。更多还原

【Abstract】 Based on the theoretical research of Recurrent Neural Network(RNN) and Long-Short Term Memory Network(LSTM),this paper proposes a prediction model of lane change behavior of intelligent vehicles based on LSTM.Firstly, theframeworkof LSTM model is built. Then, according to the human driving scene, feature selection and data extraction of the real data set NGSIM(Next Generation Simulation) were carried out.Finally, the LSTM model was used for training to test the prediction results of vehicle lane change,and the results were compared with those predicted by the RNN model to verify the effectiveness of the proposed method.更多还原