Digital twin of a mobile radar for small unmanned aerial vehicle detection

Abstract

The purpose of the study is to substantiate and experimentally verify the architecture of a digital twin of a mobile radar station intended for automatic search, detection, tracking and identification of small multirotor unmanned aerial vehicles operating within an active protection complex. The model reproduces a 360° azimuth surveillance sector, upper-hemisphere coverage over the elevation range of 0…60°, slant ranges from 10 to 500 m, target altitudes up to 500 m and target speeds up to 80 m/s. A four-layer architecture is proposed, in which the scene and motion layer, the radar channel layer, the signal and track processing layer and the verification layer are separated, and the adequacy of the model is assessed by a quantitative criterion of relative divergence from reference measure-ments of a hardware-in-the-loop test bench. The digital twin receives synthetic radar returns from targets with a radar cross section from 0.01 m² and produces estimates of range, azimuth and eleva-tion angle; the probability of detection for a stationary and a moving carrier, the information update rate of 0.01–0.2 s, the resolution and the root-mean-square error of coordinate estimation are examined. The experimental part covers six trajectory scenarios — hovering, manoeuvring, diving, approach at extremely low altitudes, repeated approach and group flight — with 1200 randomised runs each, giving 7200 runs for each of the two carrier modes (14,400 in total). Statistical processing includes Wilson confidence intervals, a variance-based sensitivity analysis using Sobol indices and a comparison with the reference measurements of the test bench. For a target with a radar cross section of 0.01 m² at a range of 300 m, the obtained probability of detection is 0.940 for the stationary carrier and 0.891 for the moving carrier; the root-mean-square error of range estimation is 2.80 m and 4.10 m, respectively, and the classification accuracy for five classes of air objects reaches 0.910. The maximum relative divergence between the digital twin and the test bench does not exceed 11.4% against an accepted tolerance of 15%, which allows the model to be regarded as adequate within the declared operating envelope. The sensitivity analysis shows that the radar cross section, the signal-to-noise ratio and the aspect angle jointly account for about 74 % of the output variance. The engineering result consists in confirming the compliance of the algorithms with the target characterristics before a physical prototype is manufactured, in reducing the amount of field trials and in forming a verified parameter set for subsequent prototyping.

Keywords: digital twin, radar station, small unmanned aerial vehicle, radar cross section, probability of detection, micro-Doppler signature, target tracking, model verification.