Double DQN Based Computing Offloading Scheme for Fog Radio Access Networks

Fog radio access network(F-RAN) is regarded as a promising network architecture for computing offloading. Due to the huge differences in computing and communication capabilities between fog nodes and user devices, it is necessary to choose an appropriate offloading policy to make full use of resources. To this end, this paper proposes a computing offloading policy in F-RAN s considering device-to-device(D2D) communication between user devices. Specifically, we formulate the offloading problem as minimizing the sum cost considering execution delay and energy consumption, which is a NP-hard problem. To obtain the optimal solution, a computing offloading algorithm based on deep reinforcement learning(DRL) is proposed. Furthermore, to reduce the complexity of the DRL algorithm, the cache-aided procedure is adopted before the training process, which enables the requesting user to obtain the task execution results from itself or nearby devices. Simulation results demonstrate that the proposed scheme can reduce the sum cost-efficiently compared with other baselines.