Smart Resource Allocation in IoMT with UAV-Edge Federated Deep Learning

Over a period of recent years, Unmanned Aerial Vehicle (UAV) integrated into the Internet of Medical Things (IoMT) has become a viable option for providing effective medical care in remote and resource limited areas. Nevertheless, the methods currently available are characterized by the following drawbacks such as high latency, poor energy performance, inaccurate models, and low convergence speed because of poor task offloading and resource allocation algorithms. In this paper, a new Attention-based Deep Dilated Convolutional Neural Network with Planet Optimization Algorithm (ADDCNN-POA) framework will be proposed to effectively solve the above challenges by allocating the most adequate resources and offloading the computations needed by a UAV-assisted IoMT setup. Task severity and Age of Information (AoI) are used to prioritize the most critical health information, so that it is attended to in time. The ADDCNN model is capable of capturing spatial-temporal features, which are needed in making intelligent decisions, and the POA refines the hyperparameters of the model, which is needed in enhancing the accuracy and efficiency of the model as per the mimicking gravitational dynamics. Simulation outcomes demonstrate of the effectiveness of the proposed ADDCNN-POA framework, which can deliver high (98.7 %) accuracy, low latency (8.3 ms), low power utilization (19.6 J), the cost of the system (61), and convergence in only 510 episodes.