The invention provides an automatic driving method and device based on hierarchical reinforcement learning, and the method comprises the steps: obtaining driving data which comprises the background data of a simulation driving platform; training an encoder according to the driving data to obtain a multi-dimensional feature vector; unsupervised training is carried out on the multi-dimensional feature vector to obtain a plurality of initial skills, and the unsupervised training comprises training added with information entropy constraint; hierarchical reinforcement training is carried out on the multi-dimensional feature vector to obtain a controller, and the hierarchical reinforcement training comprises the step of adjusting the loss of the deep learning network; the plurality of initial skills are combined by the controller to perform an automated driving action. According to the method disclosed by the invention, a plurality of general skills can be obtained through training, the skill training is objective and efficient, the skills are combined through the controller to realize automatic driving, the interpretability is high, and the automatic driving can cope with a complex environment.
本公开提供一种基于分层强化学习的自动驾驶方法及装置,方法包括:获取驾驶数据,驾驶数据包括模拟驾驶平台的后台数据;根据驾驶数据训练编码器,得到多维特征向量;对多维特征向量进行无监督训练,得到多个初始技能,无监督训练包括加入信息熵约束的训练;对多维特征向量进行分层强化训练,得到控制器,分层强化训练包括调整深度学习网络的损耗;通过控制器对多个初始技能进行组合,以执行自动驾驶动作。本公开的方法可以训练得到多个通用技能,技能的训练客观高效,通过控制器组合技能实现自动驾驶,可解释性强,使得自动驾驶可以应对复杂的环境。
Automatic driving method and device based on hierarchical reinforcement learning
基于分层强化学习的自动驾驶方法及装置
2023-05-05
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 / G06N COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS , Rechnersysteme, basierend auf spezifischen Rechenmodellen |
Automatic driving decision-making method and device based on hierarchical reinforcement learning
Europäisches Patentamt | 2022
|Europäisches Patentamt | 2020
|Unmanned driving hierarchical motion decision control method based on deep reinforcement learning
Europäisches Patentamt | 2021
|Automatic driving decision-making method based on deep reinforcement learning
Europäisches Patentamt | 2024
|