Cloud-Edge Collaborative Computation Offloading: A Deep Reinforcement Learning approach
Xin Pu, Song Jiang, Xuejie Zhang · 2022
Aiming at the optimization problem of computation offloading for multiple indicators in the cloud-edge collaborative environment, a cloud-edge collaborative computation offloading optimization algorithm based on deep Q-learning network (DQN) is proposed. Firstly, the cloud-edge collaborative computation offloading optimization problem is modeled as a Markov Decision Process (MDP); secondly, three key indicators of the average delay, average energy consumption, and server provider's revenue per unit time of computation offloading are jointly optimized; thirdly, DQN algorithm is used to solve it. The simulation results show that with the increase in the number of computation offloading tasks, the method in this paper achieves the optimal comprehensively on the above three key indicators compared with the other baselines.