Non-Metric Navigation for Mobile Robot Using Optical Flow

In this paper, we address the problem of nonmetric navigation for mobile robot in indoor environment. With pure vision sensors, we tackle the most fundamental problem in autonomous mobile robot navigation - to achieve obstacle avoidance ability in mobile robot. We approach the problem by extracting time-to-contact information using optical flow and motion analysis. Our method is based on flow divergence which contains qualitative depth information of the environment. In addition, we presented a flexible robot heading decision making framework that is able to incorporate higher level navigation task on top of obstacle avoiding behaviour. A state-machine based control scheme is utilized for the coordination of the robot's action defined under a behaviour based design. Through virtual simulation and physical experiments, we demonstrated the effectiveness of our non-metric navigation strategy in unknown environment.