Multi-MAV trajectory planning is essentially a high dimensional path and motion planning problem. The main challenge of this problem beyond individual MAV trajectory planning is collision avoidance for dynamic obstacles, which is usually achieved by spatial planning or spatial and temporal combined planning. In this paper, however, we address this problem by planning only in time domain, which can naturally decouple the static and dynamic obstacle collision avoidance. By introducing the time scheduling graph and constructing timeirrelevant collision range pair as high dimensional time domain static obstacles, we formulate this problem as a convex programming problem. Starting from arbitrary static object collision free trajectories, we propose a novel approach to achieve near optimal,...[ Read more ]

Multi-MAV trajectory planning is essentially a high dimensional path and motion planning problem. The main challenge of this problem beyond individual MAV trajectory planning is collision avoidance for dynamic obstacles, which is usually achieved by spatial planning or spatial and temporal combined planning. In this paper, however, we address this problem by planning only in time domain, which can naturally decouple the static and dynamic obstacle collision avoidance. By introducing the time scheduling graph and constructing timeirrelevant collision range pair as high dimensional time domain static obstacles, we formulate this problem as a convex programming problem. Starting from arbitrary static object collision free trajectories, we propose a novel approach to achieve near optimal, kinodynamical feasible, dynamic collision-free re-scheduled trajectory planning. We also propose a sufficient condition to guarantee the existence of solutions.