OmniNav: Obstacle Avoidance for Large, Non-circular, Omnidirectional Mobile Robots

This paper describes an obstacle avoidance method designed specifically for large, non-circular robots with omnidirectional motion capabilities. The method, called OmniNav, is reflexive, meaning that it is able to make navigation decisions without having to discern details of the environment, such as edges. The OmniNav method combines elements of the previously developed Virtual Force Field (VFF) and Vector Field Histogram (VFH) methods to protect arbitrarily shaped mobile platforms from collisions. To this end the method computes a preliminary direction of travel based on the VFH method, which assumes the robot was point-sized. Then, highly localized potential fields are applied to "act-on points" around the robot's real perimeter. A transfer function uses these "virtual forces" to alter the preliminary direction of motion and to add rotation as necessary to avoid collisions.