Vision-based obstacle avoidance of wheeled robots using fast estimation

A baseline path-following controller with an online obstacle-avoidance algorithm such that the robot tracks a prespecified path and avoids isolated obstacles using only visual information collected from an onboard camera. The visual information collected by the camera is processed by an estimator providing the inputs that includes estimation of the obstacle's position and size to the obstacle-avoidance algorithm. The obstacle-avoidance algorithm uses a potential function that yields larger values for paths that are close to obstacles and lower values for paths that are far from obstacles. The obstacle-avoidance method deforms the original path on a kinematic level in real time to move away from obstacles and ensures that the new path is collision free. In the vision-based range identification problem relative range is estimated between the vehicle and an obstacle with the use of a fast-estimation scheme. The fast estimator allows for fast adaptation of the unknown parameter via a large adaptation gain and low-pass filter.