The invention is suitable for the technical field of automatic parking, and provides an intelligent driving vehicle autonomous parking system based on deep reinforcement learning, and the system comprises an MDP frame which is used for defining a state space S, defining an action space A, and defining a reward function R; in the strategy network, an Actor-Critic framework is used, the strategy network pi outputs an action a, CNN is adopted to extract features, and a full connection layer represents a strategy; the Q network is used for evaluating a state-action function, updating a strategy network and decomposing the strategy network into a state value and a dominant function by adopting Dueling DQN; the deep reinforcement learning algorithm module is used for iteratively updating the network theta by using a PPO or DDPG algorithm and learning a parking strategy; the multi-sensor module is used for extracting the edge of the parking space and the contour of the marker as features; a parking marker is detected and recognized, and a scene image is provided. According to the invention, real unmanned parking is realized, the system works completely and autonomously, and no manual intervention is needed to participate in the parking process.
本发明适用于自动泊车技术领域,提供了基于深度强化学习的智能驾驶车辆自主泊车系统,包括:MDP框架,用于定义状态空间S,定义动作空间A,以及定义奖励函数R;策略网络,使用Actor‑Critic框架,策略网络π输出动作a,采用CNN提取特征,全连接层表示策略;Q网络,用于评估状态‑动作函数,更新策略网络,采用Dueling DQN分解为状态价值和优势函数;深度强化学习算法模块,使用PPO或者DDPG算法迭代更新网络θ,学习泊车策略;多传感器模块,用于提取泊车位边缘和标识物轮廓作为特征;检测并识别泊车标识物,提供场景图像。本发明实现真正意义的无人泊车,系统完全自主作业,无需任何人工干预参与泊车过程。
Intelligent driving vehicle autonomous parking system based on deep reinforcement learning
基于深度强化学习的智能驾驶车辆自主泊车系统
2024-03-15
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 |
Reinforcement Learning Based Parking Space Egress for Autonomous Driving
SAE Technical Papers | 2024
|Europäisches Patentamt | 2023
|Europäisches Patentamt | 2021
|Autonomous Driving with Deep Reinforcement Learning
SLUB | 2023
|