Deep reinforcement learning-based task offloading decision in the time varying channel

This paper proposes a dynamic task offloading decision control scheme to minimize the total delay to execute computation task taking into account the time-varying channel. Specifically, we consider the practical task offloading process, where executing computation task is carried out over multiple channel coherence times. In order to make an accurate decision on the task offloading process performed over multiple channel coherence times, we utilize the model-free reinforcement learning, since environment dynamics of the system, channel transition probabilities, is challenging to estimate. We formulate a problem of minimizing the total delay of executing computation task based on a Markov decision process (MDP). In order to solve the MDP problem, we develop a model-free reinforcement learning algorithm. Simulation results show that our proposed scheme outperforms the conventional scheme.