| 摘要: | 在這篇論文中，我們考慮了由載波發射器 (CEs) 供電的多個反向散射裝置 (BDs)，這些反向散射裝置可將收集到的數據鑲嵌至由載波發射器發送之連續的弦波信號以上傳至多台全雙通無人機 (UAVs)，並同時透過無人機和基地台 (BS) 之間的通訊執行聯邦式學習 (FL)。在此情境下，我們聯合設計了無人機飛行軌跡、無人機用戶關聯決策，以確保在聯邦式學習模型收斂的情況下最小化目標函數即總任務時長。由於訊號從載波發射器發射途經反向散射裝置，再由反向散射裝置反射訊號至無人機，中間經過兩段傳輸，代表訊號受到了兩次通道的影響，通常被假設為雙重雷利衰落通道，而這些影響疊加在一起是難以分析的，並且在我們考慮的系統模型中由於同時收發的全雙通無人機自干擾問題以及雙重雷利衰落通道造成訊號快速衰退等問題使得通訊環境十分苛刻，在如此苛刻的環境中設計無人機的飛行軌跡和用戶關聯決策顯得尤為重要。為了克服這些挑戰，我們首先評估了該場景的可行性以確保研究內容在實際應用中能夠實現，並嘗試近似雙重雷利衰落通道之解析解，使用該近似解針對所提出的問題進行凸最佳化求解，該問題被拆分為兩個凸子問題，利用逐次凸近似（SCA）和交替優化（AO）來解決。;In this paper, we considered multiple backscatter devices (BDs) powered by carrier emitters (CEs), these BDs can embed the collected data into continuous sinusoidal signals transmitted by CEs for uplink to multiple full-duplex unmanned aerial vehicles (UAVs), while simultaneously executing federated learning (FL) through communication between the UAVs and the base station (BS). In this context, we jointly design UAV flight trajectories and UAV-node user association to minimize task completion time as the objective function, while ensuring the convergence of the FL model. Since the signal is transmitted from the CE, passes through the BD, and is then reflected by the BD to the UAV, it undergoes two transmissions, indicating that the signal encounters channel effects twice, it is typically assumed to be a double Rayleigh fading channel, and the combined effects of these factors are complex and challenging to analyze. Furthermore, in the system model under consideration, the communication environment is highly challenging due to the rapid signal attenuation caused by the double Rayleigh fading channel, coupled with the self-interference issue of full-duplex UAVs operating in simultaneous transmission and reception, in such a challenging environment, designing UAV flight trajectories and user association decisions becomes particularly crucial. To overcome these challenges, we first evaluate the feasibility of the scenario to ensure that the research can be practically implemented, and attempt to approximate a closed-form expression for the double Rayleigh fading channel, and based on this approximation, we apply convex optimization to solve the proposed problem. The problem is decomposed into two convex subproblems, which are solved using successive convex approximation (SCA) and alternating optimization (AO). |
