The invention discloses a pedestrian behavior prediction method for a pedestrian-vehicle game at a no-signal right-turn intersection, and the method comprises the steps: S1, obtaining the historical data of pedestrians and vehicles at the no-signal right-turn intersection as sample data, and building a training sample data set; s2, analyzing human-vehicle game influence factors of the non-signal right-turn intersection, and establishing a game utility matrix according to two influence factors of safety utility and timeliness utility; s3, introducing the game utility matrix into a maximum entropy inverse reinforcement learning model, and constructing a pedestrian-vehicle game pedestrian behavior prediction network model based on an inverse reinforcement learning model network architecture; s4, using the training data set to train a pedestrian-vehicle game pedestrian behavior prediction network model; and S5, using the trained pedestrian-vehicle game pedestrian behavior prediction network model to perform real-time prediction on pedestrian behaviors at the non-signal right-turn intersection. According to the method, the prediction accuracy of the pedestrian track at the no-signal right-turn intersection can be ensured, and the pedestrian behavior virtual simulation can be more accurately carried out.
本发明公开了一种无信号右转交叉口人车博弈的行人行为预测方法,包括:S1、获取行人及车辆在无信号右转交叉口的历史数据作为样本数据,建立训练样本数据集;S2、分析无信号右转交叉口的人车博弈影响因素,根据安全性效用和时效性效用两个影响因素建立博弈效用矩阵;S3、将所述博弈效用矩阵引入到最大熵逆强化学习模型,构建基于逆强化学习模型网络构架的人车博弈行人行为预测网络模型;S4、使用所述训练数据集对人车博弈行人行为预测网络模型进行训练;S5、使用训练后的人车博弈行人行为预测网络模型对无信号右转交叉口的行人行为进行实时预测。本发明方法能后保证对无信号右转交叉口的行人轨迹的预测准确性,更准确的进行行人行为虚拟模拟。
Pedestrian behavior prediction method for pedestrian-vehicle game at no-signal right-turn intersection
一种无信号右转交叉口人车博弈的行人行为预测方法
2024-05-10
Patent
Electronic Resource
Chinese
No-signal right-turn intersection pedestrian trajectory prediction method based on game theory
European Patent Office | 2023
|European Patent Office | 2024
|