Distortion-Aware UAV Placement for Aerial Semantic Relay Communications: An Analytical Approach

Abstract

Aerial semantic relay communications (SRC) employs an unmanned aerial vehicle (UAV) equipped with a semantic encoder as a relay, which not only extends the data acquisition coverage of the base station (BS) from resource-limited sensing device (SD) but also enhances communication efficiency through semantic feature transmission over the UAV-BS link. Existing works mainly focus on sum-rate maximization, overlooking the end-to-end reconstruction distortion of sensory data in UAV-assisted SRC systems. Optimizing the UAV placement is crucial for minimizing the end-to-end reconstruction distortion, as it fundamentally trades off the input perturbation at the UAV-side encoder against that at the BS-side decoder through the two-hop wireless channel conditions. In this paper, we propose an interpretable and efficient UAV placement policy by minimizing end-to-end reconstruction distortion in aerial SRC. This is a challenging task since the black-box nature of the DNN-based codecs and the intricate coupling between the heterogeneous codec sensitivities, along with two-hop channel impairments, render the end-to-end distortion analytically intractable to characterize. We first derive an analytical expression of the end-to-end distortion, explicitly revealing the impact of cross-hop perturbation coupling, wireless channel and radio resource on the reconstruction error. Based on that, we develop a closed-form UAV placement strategy with fast adaptability across various aerial SRC system configurations. Numerical results demonstrate that the proposed distortion-aware UAV deployment closely tracks the empirical exhaustive-search optimum, while achieving lower distortion compared to representative capacity-based and curve-fitting benchmarks.