Robot Navigation Controller: a Non-Parametric Regression Approach

This paper presents a kernel regression approach towards the navigation control of a robotic agent in static as well as dynamic environments where robot's field of operation may comprise number of obstacles. Two different scenarios namely, off-line and online navigation controls, have been examined while dealing with static environment. In off-line case, a fully observable environment, with regards to position information of the obstacles, has been assumed. Whereas in case of online navigation (both static as well as dynamic obstacles), the controller modifies the robot's path at run-time to avoid potential collision(s) as per robot-obstacle vicinity.