Vision-based Reactive Collision Avoidance Algorithm for Unmanned Aerial Vehicle

Abstract

A real-time midair collision avoidance algorithm using a single vision sensor is proposed. To deal with moving aircraft, the closest point of approach (CPA) is adopted for conflict detection. The characteristics related to the CPA and relative motions are analyzed to overcome the dimension problem of a single vision sensor, and the required heading changes for collision avoidance are derived. And, the avoidance time criterion is also considered for activation of the algorithm. For real-time applications, the algorithm is designed to be operated reactively. To evaluate the proposed algorithm, numerical simulations are performed for two illustrative scenarios of head-on collision and broadside collision.