Robust Localization of Maneuvering UAV Swarms Using Lateral Doppler Resolution

A radar with a limited aperture may lack sufficient cross-range resolution to separate closely spaced members of an unmanned aerial vehicle (UAV) swarm. This paper proposes a localization method that combines lateral-Doppler scaling, data-driven estimation of the equivalent cross-range reference velocity, and acceleration-hypothesis backprojection (BP). The reference velocity is estimated by maximizing a regularized spatial-concentration criterion, and target-dependent motion is accommodated by pixel-wise maximum fusion over an acceleration grid. For a fixed 15-target formation with acceleration scales from 1.8 to 9 m/s², the method achieved conditional localization RMSE values of 0.81-1.15 m and single-scene detection rates of 0.87-1.00. The differences relative to range-Doppler (RD) and second- and third-order Taylor-Chirplet processing were most pronounced at the larger acceleration scales. Monte Carlo experiments with randomized noise and Swerling-I scattering further showed that maximum fusion yielded higher matched-detection probabilities and lower miss-penalized RMS errors than mean fusion and a zero-acceleration baseline over per-pulse SNRs from -18 to -6 dB. Acceleration-grid, PRF, and reference-velocity sweeps characterized the associated computational trade-offs and operating range.