The invention discloses an automatic driving behavior integrated decision-making method based on deep reinforcement learning, and the method comprises the steps: starting from the priority and logic of two basic driving behaviors, i.e., lane changing and car following behaviors; firstly, an upper-layer deep reinforcement learning model and a lower-layer deep reinforcement learning model are adopted to process a large-scale mixed state space, composite action output of a lane changing decision and a car following decision is ensured, and a D3QN algorithm is used in an upper-layer model to distinguish potential value of an environment and value of the lane changing decision; and adopting a DDPG algorithm to process a car-following decision in the lower-layer model so as to output continuous vehicle speed control. After a mixed standard driving condition is selected to carry out coupling training on the two algorithms, the model is tested in a highly complex driving scene reconstructed by using NGSIM data. According to the integrated driving decision model provided by the invention, the collaboration of car following and lane changing decisions is improved, and the lane dominant function and a dangerous action shielding mechanism are integrated into the decision model, so that the driving speed is remarkably improved, and the integrated driving decision model is superior to other intelligent driving decision models.
本发明公开了一种基于深度强化学习的自动驾驶行为一体化决策方法,从两种基本驾驶行为,即换道和跟驰行为的优先级和逻辑出发,首先采用上下两层深度强化学习模型处理大规模混合状态空间并保证换道决策和跟驰决策的复合动作输出,在上层模型中使用D3QN算法来区分环境的潜在价值和换道决策的价值,在下层模型采用DDPG算法处理跟驰决策从而输出连续的车速控制。选取混合标准驾驶工况对两种算法进行耦合训练之后,在利用NGSIM数据重建的高度复杂的驾驶场景中对模型进行测试。本发明提出的集成驾驶决策模型提高了跟驰与换道决策的协同性,并将车道优势函数与危险动作屏蔽机制整合到决策模型中,显著提升了行驶速度,优于其他的智能驾驶决策模型。
Automatic driving behavior integrated decision-making method based on deep reinforcement learning
一种基于深度强化学习的自动驾驶行为一体化决策方法
2022-11-11
Patent
Electronic Resource
Chinese
IPC: | B60W CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION , Gemeinsame Steuerung oder Regelung von Fahrzeug-Unteraggregaten verschiedenen Typs oder verschiedener Funktion / G06N COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS , Rechnersysteme, basierend auf spezifischen Rechenmodellen |
Automatic driving behavior decision-making method based on deep reinforcement learning
European Patent Office | 2020
|Automatic driving behavior decision-making method based on deep reinforcement learning
European Patent Office | 2024
|Automatic driving decision-making method based on deep reinforcement learning
European Patent Office | 2024
|Automatic driving vehicle overtaking decision-making method based on deep reinforcement learning
European Patent Office | 2022
|Automatic driving decision-making system based on deep reinforcement learning
European Patent Office | 2024
|