The invention provides a reinforcement learning parking path planning method and system for a non-structural scene, and the method comprises the steps: coding an unstructured obstacle map into a potential feature vector through a variational automatic encoder, obtaining the hidden features in the obstacle map, representing the state information of a vehicle and the information of a parking space through vectors, and carrying out the recognition of the vehicle. The three are connected to form a state vector, a mixed A * algorithm is used as an expert to search an initial parking path, and the state vector and path information are input into an imitation learning network, so that the automatic driving vehicle learns to plan a basic parking path; secondly, optimizing an imitation learning planning strategy by utilizing reinforcement learning, so that a basic path is optimized, and a parking scene which does not exist in imitation learning can be adapted; the reinforcement learning is converted from image processing to potential vector processing, so that the model training time is shortened, and the strategy is easier to popularize; under the condition of a parking strategy, interaction with the environment is continued, and the problems of insufficient imitation learning data set and low learning efficiency are solved.
本发明提出了一种非结构场景的强化学习泊车路径规划方法及系统,通过变分自动编码器将非结构化障碍物地图编码成潜在特征向量,获取障碍物地图中隐藏的特征,然后将车辆自身状态信息和泊车位信息用向量表示,将三者连接构成状态向量,以混合A*算法作为专家搜索出初始泊车路径,将状态向量和路径信息输入模仿学习网络,使自动驾驶车辆学会规划处一条基本的泊车路径;再利用强化学习优化模仿学习规划策略,使其不仅优化基本的路径,还能适应模仿学习中没有的泊车场景;使强化学习从对图像的处理转换为对潜在向量的处理,减少模型训练时间,使策略更容易推广;在已经泊车策略的情况下继续与环境交互,解决模仿学习数据集不足和学习效率低下的问题。
Reinforcement learning parking path planning method and system for non-structural scene
一种非结构场景的强化学习泊车路径规划方法及系统
2023-12-15
Patent
Electronic Resource
Chinese
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 |
Parking path planning method based on reinforcement learning
European Patent Office | 2020
|Parking path planning and parking method for unstructured scene autonomous vehicle
European Patent Office | 2023
|European Patent Office | 2015
|OBSTACLE AVOIDANCE METHOD IN AUTOMATIC PARKING PATH PLANNING AND PARKING PATH PLANNING SYSTEM
European Patent Office | 2021
|