Millimeter-wave radar onboard dynamic obstacle avoidance system proposed

Dnyandeep Mandaokar and Bernhard Rinner propose an onboard perception-and-control dynamic obstacle avoidance system based on millimeter-wave radar for small quadrotor drones, and claim it is the first such system. The system combines a lightweight interacting multiple model tracker with a control barrier function controller that directly outputs avoidance accelerations, and derives sufficient obstacle avoidance conditions relating sensing range, relative velocity and control delay. The authors implement an onboard version on a Raspberry Pi 4B, claiming an end-to-end radar processing latency of about 14 milliseconds per frame and a system latency of 75 milliseconds from first detection to issuing an avoidance command. Experiments claim the drone avoided obstacles in 20 tethered indoor flights and 39 untethered outdoor avoidance maneuvers, covering three sphere sizes, in hover and forward flight up to 5 meters per second, with measured closing speeds up to 14.6 meters per second. The paper is a preprint, has been submitted and is under review, and the above performance figures are reported by the authors and have not yet been confirmed by peer review.

AI 根据报道生成 · 2 days ago更新