Robot map-free navigation method based on deep safety reinforcement learning

Abstract

The invention particularly relates to a robot map-free navigation method based on deep safety reinforcement learning, which specifically comprises the following steps: initializing a training environment, and designing a mobile robot reward function and a safety risk cost function; combining image information and laser radar information detected by a sensor with target information and motion information of the mobile robot, processing state information and then outputting decision actions to the robot through an Actor network, so the robot executes the actions output by the Actor network, and new state observation and reward information at the next moment is obtained from the environment; storing experience obtained by interaction between the robot and the environment in an experience pool, and updating network parameters regularly; and judging whether the training is finished or not, and applying the trained model to a real mobile robot for navigation. According to the method, based on deep safety reinforcement learning of an actor-critic-safety (ACS) framework, by introducing a constraint strategy optimization (CPO) algorithm, the safety of reinforcement learning for a map-free navigation task is improved.

Description

technical field [0001] The invention relates to the field of map-free navigation of robots, in particular to a map-free navigation method for robots based on deep safety reinforcement learning. Background technique [0002] Robot navigation refers to the technology that the robot reaches the target position from the current initial position without colliding with other static or dynamic obstacles during the process. In recent years, robot navigation technology has been widely used in the fields of sweeping robots, service robots, logistics robots, special rescue robots and Mars exploration robots. Safety in navigation is mainly reflected in the ability to avoid collisions. [0003] Comparing with maps and without maps: the current relatively mature navigation technology is basically based on SLAM mapping with map navigation. However, when special robots such as field search and rescue are performing tasks, the environment is often unknown, and maps cannot be built at this t...

AI Technical Summary

Problems solved by technology

[0005]Although reinforcement learning has been successfully applied in the fields of games and control, most of the reinforcement learning work is developed based on simulation platforms, and it is more used in real mobile robots. The reason is that the strategies learned by reinforcement learning cannot be guaranteed to be safe, which may cause serious consequences

Images