Experimental validation of wide-field integration methods for autonomous navigation

Abstract

Wide-field integration (WFI) methods, which are based on the spatial decompositions of optic flow performed by motion sensitive tangential neurons within the insect visuomotor system, have been shown to be an efficient and robust way to extract visual cues for guidance and navigation. In this paper, previous analytical results on WFI-based autonomous local guidance are validated on a wheeled robotic platform. Successful corridor navigation is achieved through closed loop feedback of the first two spatial harmonics of the optic flow, which provide the lateral and rotational stiffness with respect to local obstacles required to stabilize a collision-free trajectory through the environment.