A Federated Learning Based Fault Diagnosis in UAV-Reliable Sensor Network

This paper proposes a federated learning-based framework for fault detection in an unmanned aerial vehicle (UAV)-reliable sensor network. The proposed approach is implemented in three stages. In the first stage of implementation, an experimental setup is done for real-time sensor data collection using a UAV module. In the second stage of implementation, a three-tier architecture is designed to perform a federated learning computation. In the third stage, a probabilistic neural network model is implemented in the federated learning framework for fault detection and classification. The real-time data set is used for the validation of the model in terms of fault detection rate, false alarm rate, and false positive rate. The distributed framework ensured data security and optimal resource utilization, which increased the quality of service (QoS) performance of the UAV-reliable sensor network.