Robust Source Localization Exploiting Collaborative UAV Network

Abstract

In this paper, we propose a robust strategy to localize multiple ground sources exploiting a distributed unmanned aerial vehicle (UAV) network in the presence of impulse noise. We achieve robust source localization by using ℓ1-principal component analysis (ℓ1-PCA) based signal subspace estimation at each individual UAV. This approach significantly reduces the signal subspace perturbation compared to the conventional ℓ2-PCA based counterpart. The obtained robust signal subspace estimate is exploited to provide an improved estimate of the noise subspace, which is in turn utilized by the MUSIC algorithm to render coarse source localization at each individual UAV. The source localization information obtained at multiple UAVs is then fused by exploiting group sparsity using the re-weighted ℓ1 minimization. Simulation results demonstrate the effectiveness of the proposed approach.