Robot obstacle avoidance behavior learning and target search method based on deep belief network

Abstract

The invention discloses a robot obstacle avoidance behavior learning and target search method based on a deep belief network. Firstly, the robot is controlled to avoid obstacles in the environment, and the color and depth image data, as well as the linear velocity and angular velocity corresponding to the moving substrate of the robot are collected at the same time, and then based on these data, a network model for automatic obstacle avoidance behavior learning of the robot is constructed; the robot is in the process of automatic target search In the process, the target is randomly searched in the environment through the automatic obstacle avoidance function; once the target is searched, it will directly approach the target; if an obstacle appears during the approaching process, the robot can avoid the obstacle and re-plan the path; The target is lost, and the robot searches randomly again; the above process is repeated until the robot reaches the target position. The present invention only uses a single RGB-D camera to realize path planning and target search with automatic obstacle avoidance functions, and has high feasibility and practicability in terms of cost and application.

Description

technical field [0001] The invention relates to the technical fields of machine learning and pattern recognition, in particular to a method for automatic obstacle avoidance behavior learning and target search of a robot based on a Deep Belief Network (DBN) in an unstructured environment. Background technique [0002] Automatic object search is a necessary skill for all kinds of robots in the current human environment, such as social robots, service robots, search and rescue robots, etc. to function well in unstructured environments. This skill requires the robot to have many other basic capabilities, including object recognition, automatic obstacle avoidance, path planning and navigation. Among them, it is the key to realize the automatic obstacle avoidance learning ability of the robot and the path planning ability from the starting point to the destination in an unknown environment. Although the above functional modules are very active in the research fields of computer v...

AI Technical Summary

Problems solved by technology

Although the above functional modules are very active in the research fields of computer vision and robotics, there are relatively few studies on the complete realization of target recognition, search and approaching on robotic systems through the learning of automatic obstacle avoidance, and only use a single RGB- There are relatively few studies on the automatic obstacle avoidance ability learning of robots with 3D cameras.

Images