Pursuit-Evasion Game approach to Tree-Based Path Planning for Airborne Dynamic Obstacle Avoidance
Vahram Stepanyan, Anjan Chakrabarty, Corey A. Ippolito · AIAA Scitech 2021 Forum · 2021
View Video Presentation: https://doi.org/10.2514/6.2021-1575.vid The paper presents an approach to avoid non-cooperative airborne dynamic obstacles for unmanned aerial systems using pursuit-evasion differential game framework and tree-based path planning. The goal is to generate safe trajectories inside a predefined volume, which is approved for the aerospace operations. It is assumed that the evading vehicle can determine its own state in real time, and estimate the intruder's largest size and center of mass position using on-board sensors. In addition, it is assumed that the intruder's information is available at a much slower rate or not available at all for a some time intervals. The evading vehicle has to travel between given way points, while avoiding the collision and satisfying given constraints. In the meantime, it has to stay as close to the given path as possible during the avoidance maneuvers. Sufficient conditions are derived based on the aerial vehicle's position and velocity, and intruder's position at the detection/estimation time instance, assuming that the intruder's maximum possible velocity is known. The presented approach relies on an analytical solution of the differential game problem formulated for two point masses with constraints imposed on the velocities, and generates a safe trajectory on a plane defined by the UAS and non-cooperative airborne dynamic obstacle positions at the measurement instance, and the target way point. This trajectory separates the approved volume into safe and unsafe flight regions along the normal direction to its plane. Then the tree based path planning is applied to generate safe path inside the safe region taking into account UAV dynamics. The process is repeated as soon as a new information on the intruder's position becomes available.