Sense-and-avoid system development on an FPGA

In this paper, the first steps of development towards a collision avoidance system for UAV are introduced. The system is based on camera sensors controlled by an FPGA SoC, running an image processing algorithm, controlling a quadcopter to avoid mid-air collisions with other aircraft. The algorithm was already tested in a real mid-air close encounter scenario between two UAVs, but in that scenario, the implementation platform was a GPU-based SoC which had serious limitations concerning processing frame rate and power consumption. Our aim is to realize the same algorithm with higher framerate and lower power consumption. The paper introduces fundamental considerations for the selection of tools used during this process.