Federated Learning-Based Collaborative Wideband Spectrum Sensing and Scheduling for UAVs in UTM Systems

In this paper, we propose a data-driven framework for collaborative wideband spectrum sensing and scheduling for networked unmanned aerial vehicles (UAVs), which act as secondary users (SUs) to opportunistically utilize detected “spectrum holes”. Our overall framework consists of three main stages. Firstly, in the model training stage, we explore dataset generation in a multi-cell environment and train a machine learning (ML) model using the federated learning (FL) architecture. Unlike the existing studies on FL for wireless that presume datasets are readily available for training, we propose an end-to-end architecture that directly integrates wireless dataset generation, which involves capturing I/Q samples from over-the-air signals in a multi-cell environment, into the FL training process. To this purpose, we propose a multi-label classification problem for wideband spectrum sensing to detect multiple spectrum holes simultaneously based on the I/Q samples collected locally by the UAVs. In the traditional FL that employs federated averaging (FedAvg) as the aggregating method, each UAV is assigned an equal weight during model aggregation. However, due to the differences in wireless channels observed at each UAV in a multi-cell environment, the received signal powers and collected datasets at different UAV locations could be significantly different, which could degrade the FL performance using equal weights. To address this issue, we propose a proportional weighted federated averaging method (pwFedAvg) in which the aggregating weights are proportional to the received signal powers at each UAV, thereby integrating the intrinsic properties of wireless channels into the FL algorithm. Secondly, in the collaborative spectrum inference stage, we propose a collaborative spectrum fusion strategy that is compatible with the unmanned aircraft system traffic management (UTM) ecosystem. In particular, we improve the accuracy of spectrum sensing results by combining the multi-label classification results from the individual UAVs by performing spectrum fusion at a central server. Finally, in the spectrum scheduling stage, we leverage reinforcement learning (RL) solutions to dynamically allocate the detected spectrum holes to the secondary users. To evaluate the proposed methods, we establish a comprehensive simulation framework that generates a near-realistic synthetic dataset using MATLAB LTE toolbox by incorporating base station (BS) locations in a chosen area of interest, performing ray-tracing, and emulating the primary user’s channel usage in terms of I/Q samples. This evaluation methodology provides a flexible framework to generate large spectrum datasets that could be used for developing ML/AI-based spectrum management solutions for aerial devices.