Deep Reinforcement Learning-Based Mobile Health Task Offloading Strategy for Vehicle Fog Computing

Mobile Health Services (MHS) is applied in "Intelligent Transportation + Healthcare", in which edge devices provide an efficient solution for data collection and processing of mobile devices. However, inefficiency of the computational offloading scheme may lead to high latency. This paper proposes an optimal task offloading scheme to assist the IoT-Ambulance (IA) edge computing environment by building a stable communication topology between vehicles based on fog computing, in which a car has limited arithmetic resources, and a Cloud Server (CS) is used as a centralized controller to complete the scheduling of computational resources and to build a temporary edge computing cluster to assist diagnostic and therapeutic application services. To jointly optimize user scheduling and resource allocation and minimize energy loss under delay constraints, we adopt a reinforcement learning-based PPO algorithm to balance the two objectives of "minimum system utility and maximum number of favorable service providers" and verify the optimality under different parameters. Simulation results show that the proposed algorithm outperforms the baseline in terms of convergence speed and computational complexity while ensuring the best system utility.