A Parallel Mapping and Planning Strategy for Micro Aerial Vehicle

Abstract

Autonomous flight is a fundamental problem for various applications of micro aerial vehicles (MAVs). Thanks to the development of Guidance, Navigation and Control (GNC) technologies, the research on this problem is also becoming mature. However, safe flight in unknown, cluttered environments remains an open question, especially with real-time requirement on onboard computer. This paper proposes a framework including parallelly mapping and planning and implements it on Graphics Processing Unit (GPU). First, a spherical coordinate projection is used in the occupancy grid map to avoid memory conflicts. After that, in the planning phase, a method based on lattice state space sampling is applied to obtain multiple trajectories parallelly. Then, we design a series of soft constraints to ensure that the MAV is in a safe known space with optimal dynamics. By solving the cost for each trajectory and comparison, the optimal trajectory can be generated. The effectiveness of the proposed strategy is demonstrated through simulation tests.