Research Of Mobile Robot Scene Cognition Based On Brain-mind Developmental Network

Traditional robot control and planning methods are usually used for the single task,learning process is offline and modeling process is complex,which leads robot difficult to adapt to the unknown environment.This paper presents the research of mobile robot scene cognition based on brain-mind developmental network,to enable a robot to learn and develop like human beings,so that the robot can gradually adapt to complex and changeable environments.The main research work is as follows:Firstly,the working mechanism of human brain and the scene cognition methods based on the biological vision mechanism are presented.The object detection approaches based on spectral residual,bootstrap learning and cellular automata under two different scenes are compared,and the advantages and disadvantages are analyzed.Secondly,a face detection experiment based on the photometric stereo technique has been designed.Compared with traditional face detection methods,it could overcome the influence of light and improve detection accuracy.Based on this method,the face recognition experiment is designed combing with the IHDR algorithm.The experimental result reveals that the combination of the IHDR algorithm and 3D face construction using the photometric stereo technique have better effect on face recognition.Finally,the visual based navigation for robot using the autonomous learning framework is proposed.By guiding the robot to learn the planned path,the mapping relationship between scene images and robot velocity is established using the IHDR autonomous learning framework.When the robot learns online,it can achieve autonomous navigation by retrieving data from the knowledge library.However,the experiment could only be carried out in static scenes.In order that the robot can adapt to more complex scenes,an autonomous planning method for robot navigation has been proposed,which combines the learning algorithm and environmentalmodeling to ensure that the robot can move within the safe zone.In the autonomous planning experiment where obstacles exist,the feasibility of combining the autonomous learning framework and the avoidance approach has been justified.This method allows a robot to adapt to more complex scenarios.