The invention provides an automatic parking method based on deep reinforcement learning. The basic idea of the invention is that the current state of the automobile is sensed by using the strong perception capability of deep learning, namely, the relative position relation between the automobile and the parking space is sensed. The state of the automobile is input into a depth deterministic strategy gradient model; an action needing to be executed in the state is obtained by utilizing the decision-making capability of reinforcement learning; and an action to be executed is input into the environment model to obtain a next state and a reward function value, then the quality of the action which is just executed is evaluated through a reward mechanism, the automobile is guided to run towardsthe direction of the parking point, and the previous operation in the next state is repeatedly executed until the model converges, i.e., the automobile is accurately parked to the parking space. By means of the design, the problems that an existing automatic parking system cannot achieve real-time interaction with the environment, and parking cannot be started at any place and any angle are solved.
本发明提供了一种基于深度强化学习的自动泊车方法,本发明的基本思想是使用深度学习强大的感知能力来感知汽车的当前状态,即感知汽车和停车位的相对位置关系,然后把汽车的状态输入进深度确定性策略梯度模型,利用强化学习的决策能力得到在该状态下需要执行的动作,把需要执行的动作输入环境模型得到下一状态和奖励函数值,然后通过一个奖励机制去评价刚刚执行的动作的好坏,并引导汽车朝着停车点的方向行驶,在下一个状态重复执行前面的操作,直到模型收敛,即汽车准确泊达车位。本发明通过以上设计,解决了现在的自动泊车系统无法做到和环境的实时交互,以及无法实现在任意地点和任意角度开始泊车的问题。
Automatic parking method based on deep reinforcement learning
一种基于深度强化学习的自动泊车方法
2020-09-11
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 |
Full-automatic parking lot scheduling method based on deep reinforcement learning
Europäisches Patentamt | 2020
|Deep reinforcement learning method suitable for automatic parking of high-speed train
Europäisches Patentamt | 2022
|Train accurate parking control method based on deep reinforcement learning
Europäisches Patentamt | 2022
|Train temporary parking strategy optimization method based on deep reinforcement learning
Europäisches Patentamt | 2023
|