Automatic parking method and system based on geometric programming and reinforced learning

Abstract

The invention relates to an automatic parking method and system based on geometric programming and reinforced learning. According to the automatic parking method, at the initial state of automatic parking, by determining a parking state, the parking trajectory is determined through geometric programming, and then is controlled through trajectory tracking and chassis control, through the stage, a vehicle can be adjusted to the pose for entering a garage at a time, and at the moment, controlling is further conducted through reinforced learning. Compared with the prior art, according to the automatic parking method, the error of trajectory planning, trajectory tracking and chassis control can be eliminated, the more ideal parking pose is achieved, and the automatic parking method can be suitable for the narrow parking environment in cities, and is high in adaptability to the environment.

Description

technical field [0001] The invention relates to the technical field of automatic parking planning for intelligent vehicles, in particular to an automatic parking method and system based on geometric planning and reinforcement learning. Background technique [0002] The existing automatic parking technology is mainly realized through the following methods: rule-based decision-making planning method, that is, through a fixed parking process, using a finite state machine to control the state of the vehicle, and planning a driving trajectory through a planning method. The output of this parking control method is predictable and stable, but it is not intelligent and cannot effectively deal with the complex scene of real parking. At the same time, the traditional architecture of trajectory planning-trajectory tracking-chassis control cannot eliminate the error of trajectory tracking and chassis control, resulting in inconsistencies between the planned trajectory and the actual tra...

AI Technical Summary

Problems solved by technology

The output of this parking control method is predictable and stable, but it is not intelligent and cannot effectively deal with the complex scene of real parking

At the same time, the traditional architecture of trajectory planning-trajectory tracking-chassis control cannot eliminate the error of trajectory tracking and chassis control, resulting in inconsistencies between the planned trajectory and the actual trajectory, and cannot adapt to the parking scene with stricter pose requirements.

Images