Improved SLAM and Motor Imagery Based Navigation Control of a Mobile Robot

Abstract

A brain-robot interface (BRI) based control system combined with the simultaneous localization and mapping (SLAM) has been developed to achieve the navigation and control of a mobile robot in unknown environments. The BRI system is based on motor imagery(MI), which analyze the human electroencephalograph (EEG) signals through the CSP-based SVM classification algorithm. The SLAM system presented in this paper is improved by combining the RGB-D SLAM, optical flow and deep leaning based Recognition. The optical flow method is used to accurately track the displacement of ORB feature points which consume much less time than calculating the ORB characteristics in each image, while the deep learning based recognition reduce the error caused by detected moving objects, and the error of the localization is minimized through global graph-based nonlinear error optimization. Experiments have been taken by 3 operators and successfully certify the feasibility and efficiency of the whole system.