The invention provides a reinforcement learning-based autonomous vehicle decision planning method in an urban traffic scene, and the method comprises the steps: enabling an autonomous vehicle to obtain the observation information, state information, motion information and the like of an own vehicle and an adjacent vehicle in an environment through a visual sensor and a laser radar; and inputting the acquired environment information into a DQN algorithm based on an attention mechanism, and outputting a corresponding automatic driving decision planning strategy by the algorithm. And outputting the reward value of the strategy according to the reward function in the urban traffic scene, and storing the corresponding state information, action information and reward value into an experience pool. And processing and extracting data in the experience pool by adopting a method based on improved priority playback experience, calculating a loss value, and updating parameters of the main network and the target network. According to the method, vehicles in an urban traffic scene can be observed emphatically by using an improved attention mechanism, and training can be performed more quickly and better performance can be obtained.
本发明提供了一种基于强化学习的城市交通场景下自动驾驶车辆决策规划方法,自动驾驶车辆通过视觉传感器和激光雷达获取环境中的自身车辆和相邻车辆的观测信息、状态信息和动作信息等,接着将获取到的环境信息输入基于注意力机制的DQN算法中,算法输出相应的自动驾驶决策规划策略。根据城市交通场景下的奖励函数输出该策略的奖励值,并将相应的状态信息、动作信息和奖励值存入经验池中。采用基于改进优先回放经验的方法对经验池中的数据进行处理和提取,计算损失值,并更新主网络和目标网络的参数。该方法利用改进的注意力机制能够有侧重的对城市交通场景中的车辆进行观测,并能够更快速的训练和获得更好的性能。
Decision planning method for automatic driving vehicle in urban traffic scene based on reinforcement learning
基于强化学习的城市交通场景下自动驾驶车辆决策规划方法
2024-07-26
Patent
Electronic Resource
Chinese
IPC: | G06N COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS , Rechnersysteme, basierend auf spezifischen Rechenmodellen / 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 |
European Patent Office | 2023
|Intelligent automobile reinforcement learning cooperative driving method for urban traffic scene
European Patent Office | 2024
|Automatic driving solution under multi-target complex traffic scene based on reinforcement learning
European Patent Office | 2022
|Automatic driving vehicle microscopic decision-making method based on reinforcement learning
European Patent Office | 2020
|Automatic driving decision planning method based on deep reinforcement learning and deep learning
European Patent Office | 2024
|