Inverse optical flow based guidance for UAV navigation through urban canyons

The paper proposes a vision based guidance strategy for an unmanned aerial vehicle (UAV) navigating through urban canyons. The UAV has a single forward facing camera and relies on optical flow computed from the sequence of images captured by the camera. The steering command is inversely proportional to the optical flow difference between the two sides of the image. This leads to high turning and safe motion while directly facing an obstacle or a wall. Analysis is carried out for infinite and finite length obstacles to highlight the obstacle avoidance properties of the proposed strategy as compared to a widely used existing strategy. Realistic simulation studies with straight, L, and T shape canyons and combination of them show the viability of the proposed strategy while navigating through urban canyons.