Simultaneous Localization And Mapping Of Brain Teleoperation-based Mobile Robot

Recently,in order to improve the robot's abilities of unknown environment exploring and information collecting,the human-machine interface based mobile robot has become a major direction of the robot application development.And the combination of brain teleoperation and human-machine interface based mobile robot is significant in improving the robot controlling technology,which enables the operator to have a deeper understanding of the unknown environment,and expands the application scope in the people with physical disfunction,such as the elder and the disable.Meanwhile,to enhance the operator's understanding of the surrounding environment and the correct rate of the operator's decision,a simultaneous localization and mapping(SLAM)technique is presented,which achieves the robot's selflocalization,self-acquisition of walking path map and environmental mapping.Aiming at the unknown corridor environment exploration task,this paper develops a Motor Imagery(MI)based brain teleoperation system on a mobile robot.Besides,in order to help the operator understand the surrounding environment in time,the system combines a SLAM system which is optimized for the corridor environment which achieves automatic information collecting,analyzing and 3-D dense mapping.The major content of this paper is as follow.1)This paper proposes a motor imagery based mobile robot brain teleoperation method.This method utilizes a Common Spatial Pattern(CSP)based Support Vector Machine(SVM)classification algorithm to transform the EEG signals into the specific digital signals after analyzing the features of the EEG signals,and develop a polar polynomial trajectory strategy to build up the relationship between the EEG signals and the robot control commands,which achieves the brain teleoperation.2)An improved simultaneous localization and mapping system is developed for the corridor environment.The system combins the optical flow and the feature matching strategy to promote the speed of the system and guarantee the accuracy when there are not enough features.Besides,to improve the quality of the image features and reduce the errors,a deep learning-based object recognition is adopted to exclude the invalid feature points on the moving obstacles such as walking people and finally achieves the accurate real-time localizing of the mobile robot and 3-D mapping of the robot's surrounding environment.The systems developed in this paper have all completed relevant experiments,the operator utilizes the brain teleoperation and SLAM system to achieve the navigation control of the mobile robot through complex paths in corridor environment without collision.The research results have a good guiding role in expanding the mobile robot's use of the unknown environment information collection function under the diversified manipulation mode such as brain teleoperation.