Federated Learning-Based Resource Allocation Optimisation in Smart Farm Edge Intelligent Networks

Smart sensors embedded in plants (SPs) dynamically generate massive raw data, which may raise privacy concerns during user data collection, transmission, and processing. Additionally, due to the limited communication resources of the smart farm systems and the limited computation resources at each device, seeking optimal resource allocation optimisation solutions to process latency-sensitive computational tasks efficiently still faces challenges. In this paper, we consider a federated learning (FL)-based edge intelligent smart farm network, where each SP generates delay-sensitive tasks and is able to offload their latencysensitive tasks to remote edge nodes for remote processing. We aim to minimise the total service delay of all SPs, including the transmission, queuing, and processing delays, by proposing a Federated Edge Smart farm Queuing offloading Optimisation (FESQO) algorithm. Simulation results show that our proposed optimisation algorithm outperforms the benchmarks regarding the total service delay of all SPs.