A Novel Federated Learning-Based Smart Power and 3D Trajectory Control for Fairness Optimization in Secure UAV-Assisted MEC Services
Raja Karmakar, Georges Kaddoum, Ouassima Akhrif
Abstract
Unmanned aerial vehicles (UAVs)-aided mobile-edge computing (MEC) systems face several challenges that hinder their practical implementation. First, the broadcast nature of wireless communications can cause security issues. Second, UAVs have constrained onboard power. Finally, the UAV should be able to serve a maximum number of ground users (GUs). It is also crucial to maintain fairness such that all GUs get equal opportunities to securely offload tasks to UAVs. We seek to address the aforementioned challenges by designing an intelligent mechanism, <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">FairLearn</i> , which maximizes the fairness in secure MEC services by controlling the UAV 3D trajectory, transmission power, and scheduling time for task offloading by mobile GUs. To this end, we formulate a maximization problem and solve it using a <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">deep neural network (DNN)</i> -based model, where the UAVs collaboratively learn the model by utilizing a <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">federated learning (FL)</i> approach. Each UAV uses a <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">reinforcement learning (RL)</i> -based approach to individually generate the training dataset, making the training data span different network scenarios. Our model is based on UAV pairs, where one UAV executes the GUs' offloaded tasks, while the other is a jammer that suppresses eavesdroppers. The simulation evaluation of FairLearn shows that it significantly improves the performance of UAV-enabled MEC systems.