Feasibility of Onboard Vision-Based Detect-and-Avoid for Small UAS Under Size, Weight, and Power Constraints

Abstract

Small unmanned aircraft systems increasingly share low-altitude airspace with crewed aircraft, rotorcraft, and other unmanned platforms whose trajectories are only partially observable and weakly cooperative. In this environment, onboard detect-and-avoid capabilities are necessary to support beyond visual line of sight operations when reliance on ground-based surveillance or cooperative transponders is not assured. Vision-based sensing is attractive for small platforms because of its geometric richness and the availability of compact low-cost cameras, but its feasibility is constrained by severe limits on mass, power, computation, and thermal dissipation. This paper examines the feasibility of onboard vision-based detect-and-avoid for small unmanned aircraft systems subject to realistic size, weight, and power constraints by integrating sensor modeling, algorithmic complexity analysis, and closed-loop encounter-level performance. The discussion focuses on monocular and stereo visible-band configurations mounted on multirotor and fixed-wing platforms with maximum take-off mass below 25 kg and continuous electrical power budgets below 80 W. A dynamical engagement framework is used qualitatively to relate detection range distributions, track continuity, and decision latency to miss-distance statistics without presupposing a particular regulatory standard of performance. The analysis highlights the coupled role of optics, pixel-level signal-to-noise, embedded inference latency, and maneuver authority in shaping achievable detect-and-avoid envelopes, while acknowledging that environmental variability and non-cooperative traffic behavior introduce uncertainties that limit deterministic guarantees. The results collectively indicate conditions under which onboard vision sensing constitutes a technically viable component of layered separation assurance architectures for small unmanned aircraft.