Real-time Stereo Visual Odometry and its application to Semi-Autonomous Wheelchair

Abstract

The ability to localize is amongst the most important traits of any mobile robot. In outdoor environment, GPS is usually used for localization, but to correctly localize in indoor environments, is a challenging task for a mobile robot. Stereo camera provides the depth information of the near environ- ment. This depth information is used to provide ego-motion without using any other sensory information. The odometry computed from intra-frame motion is fed to Sigma Point Kalman Filter (SPKF) to provide a better estimate of robots position and velocity. The proposed algorithm is imple- mented on robotic wheelchair developed in RISE Lab to provide real-time visual odometry of wheelchair, which is further used by Laser based GMap- ping algorithm to generate wheelchair's environment map.