The invention relates to the technical field of automatic driving, and discloses a driving decision and control method for an automatic lane changing behavior based on deep reinforcement learning, and the method comprises the following steps: S1, carrying out the lane changing command of an upper layer; s2, a decision layer determines when to change a lane, and uses DQN to determine when to execute a lane change operation; and S3, the adjustment layer judges how to perform operation properly, a Q-learning method and a quadratic Q function are adopted to process longitudinal continuous control actions, and in the task, the task is divided into two steps, namely an adjustment step and an execution step. The invention not only can provide an architecture based on hierarchical reinforcement learning for decision making and control under the condition of lane changing so as to solve the problem of complex decision making tasks faced by an automatic driving vehicle, but also combines the advantages of a classical control method and a machine learning method, realizes efficient and steady execution of the automatic driving vehicle, and improves the driving reliability of the automatic driving vehicle. And training and testing can be carried out based on a simulation environment.
本发明涉及自动驾驶技术领域,且公开了基于深度强化学习的自动变道行为的驾驶决策与控制的方法,括以下步骤:S1:上层的变道命令;S2:决策层决定何时变道,使用DQN来确定何时执行变道操作;S3:调整层判断如何适当地进行操作,采用Q‑学习方法和二次Q函数来处理纵向连续控制行动,在这项任务中,我们将其分为两个步骤,即调整步骤和执行步骤。本发明不仅能够提出一种基于分层强化学习的架构,用于变道情况下的决策和控制,以解决自动驾驶汽车面临的复杂决策任务问题,同时结合了经典控制方法和机器学习方法的优势,实现自动驾驶车辆的高效、稳健执行,还能够基于仿真环境进行训练和测试。
Driving decision and control method for automatic lane changing behavior based on deep reinforcement learning
基于深度强化学习的自动变道行为的驾驶决策与控制的方法
2024-05-10
Patent
Elektronische Ressource
Chinesisch
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 |
Europäisches Patentamt | 2022
|Automatic driving lane changing decision control method based on rule fusion reinforcement learning
Europäisches Patentamt | 2022
|Automatic driving vehicle lane changing method based on double-layer deep reinforcement learning
Europäisches Patentamt | 2023
|Automatic driving behavior decision-making method based on deep reinforcement learning
Europäisches Patentamt | 2020
|