Indoor navigation of skid steering mobile robot using ceiling landmarks

This paper presents an indoor navigation scheme for a skid-steering mobile robot using a low-cost gyro sensor and a monocular camera. Based on the kinematic model under nonholonomic constraint, a motion controller is designed, and so as to account for some uncertainties from the wheel-ground friction and slip, the rotational velocity measurement from a gyro is incorporated into the model. This control scheme based on dead-reckoning localization basically exhibits the aspect of cumulative position error. Hence, the position of the robot is estimated from ceiling landmark images, and is combined with odometric information by an extended Kalman filter (EKF). Consequently, experimental results are given to validate this estimation and control methodology.