IJCAS

This paper presents an online obstacle localization and avoidance scheme for aerial vehicles over the order of seconds for general aviation. The scheme contains three stages. In the first stage, stereo visual odometry is adopted to locate the local positions of the obstacles, and the local positions are fused with global positioning system (GPS) or differential GPS to obtain global map information. In the second stage, we search the safety areas on the basis of depth image obtained by semi-global Matching (SGM) algorithm and computing safety points of every frame. In the third stage, the smooth trajectory is planned in real-time based on the map and those safety points through receding horizon control based mixed integer linear programming (MILP). Simulations and experiments verified the feasibility of the proposed scheme."

Keywords: General aviation, online obstacles localization and avoidance, stereo-vision, trajectory planning.