Localization And Navigation Technology For Brain-controlled Wheelchair Under Indoor Environment

As a service robot in medical care, the brain-controlled wheelchair is a perfect combination of brain-computer interface technology and intelligent robotics. With the increasing trend of global aging and people’s urgent desire for wisdom medical services, the brain-controlled wheelchair will gradually get people’s attention.To put the brain-controlled wheelchair into practical application, autonomous navigation capability is the key. The autonomous navigation capability of the brain-controlled wheelchair is achieved by localization and navigation technology. At present, the research on the mobile robot localization and navigation technology is full, but the brain-controlled wheelchair and its applications have their own characteristics, and there is not a mature solution in indoor environments. In this paper we study the key technology and difficulties of the brain-controlled wheelchair localization and navigation technology, mainly doing the following work:Firstly, the brain-controlled wheelchair simultaneous localization and mapping(SLAM) problem is studied in this paper. The traditional SLAM method based on Rao-Blackwellised particle filter requires a large number of particles, thus we propose an improved algorithm. As the error of odometer reading is large, this paper employs the iterative closest point(ICP) algorithm to estimate the relative position, and then uses the registration result to replace the odometer reading. In the improved SLAM algorithm, motion model based on the ICP registration result is taken as proposal distribution. According to the proposal distribution for each particle, sampling for several times to select high-quality particles, and resampling according to the number of effective particles. Simulations on the brain-controlled wheelchair data sets illustrate the superior performance of our approach.Secondly, the brain-controlled wheelchair self-localization technology is studied. Traditional Monte Carlo localization(MCL) algorithm is effective and has wide applicability, however, its initial localization depends on a large number of particles and the convergence process is slow. In this paper, we propose an effective self-localization algorithm according to the characteristics of brain-controlled wheelchair and the deficiency of MCL algorithm. According to the initial scan data and map information obtained by the laser range finder, using ICP algorithm to correct the initial set of particles, and then using improved MCL algorithm to locate the wheelchair. The improved self-localization algorithm can significantly reduce the number of required particles and improve the convergence speed, and the success rate of localization is greatly improved in case that the number of particles is small.And then, we propose a contour-based localization algorithm for the static obstacles in the indoor environment. Obtain images from relative installation of cameras, complete the localization of static obstacles through correction, obstacles foreground extraction, contour extraction and fitting, image plane to ground plane and contour intersection, and then detect and process the contour over fitting problem in contour fitting process. Computational results validate the effectiveness of the proposed contour-based obstacle localization algorithm.Finally, brain-controlled wheelchair path planning and path tracking control are studied. After a variety study of efficient path planning algorithms, Voronoi diagram is selected to plan the safest path according to the principle of “safety first”. As the planned path is a piecewise linear segments, we propose the piecewise path tracking strategy, then use classical PID control algorithm to complete the tracking of each piecewise linear path. Brain-controlled wheelchair navigation experimental results show the effectiveness of the algorithm.