Using Panoramic Images and Optical Flow to Avoid Obstacles in Mobile Robot Navigation

This paper addresses the use of panoramic images to avoid obstacles in mobile robot indoor navigation. A catadioptric sensor installed onboard a mobile robot is used to acquire panoramic views covering 180 degrees of the horizon in front of the robot. The optical flow field corresponding to two of such views is calculated, which is used to guide the robot to reactively avoid obstacles. A strategy to define how much the current robot orientation should be changed and a controller to guarantee that such a change is accomplished are here proposed. Experimental results validating such a system are also presented. The main contribution of the proposed system is that the robot becomes able to avoid wide obstacles, which would be too much difficult using a camera with a narrow field of vision. An important aspect, regarding the narrow-field-of-vision case, is that the object becomes wider as the robot gets closer to it.