UAV Collision Avoidance using A* Algorithm

Abstract

Collision avoidance is the essential requirement for unmanned aerial vehicles (UAVs) to become fully autonomous. Several algorithms have been proposed to do the path planning in a simulated environment, but only few can make them effectively survive in a dynamic environment. This issue keeps UAVs from commercial and other applications because when the UAVs fly autonomously, the inability to reliably sense and avoid other aircraft in the air can cause serious hazards. In this thesis, we review several approaches including A* algorithm, total field sensing approach, and Markov decision process. Then, a modification of A* algorithm is proposed. Typically, A* algorithm is implemented in a mobile robot system for the path planning in a static environment. We introduce some approaches to allow us using A* algorithm in a dynamic environment. The evaluation of this algorithm is based on the simulation of different scenarios, and the comparison between two heuristic functions will be detailed. We discuss the performance of our approach, the suitable condition for it to work reliably, and what issues could affect its performance. We also investigate the limitation of our approach in the extreme scenarios to provide useful suggestions for improvement.