Computation offloading and resource allocation strategy based on deep reinforcement learning

In order to expand the coverage and computing power of vehicle edge network, a computation offloading architecture was proposed for space-air-ground integrated vehicle network (SAGVN). With the consideration of the delay and energy consumption constraints of computing tasks, as well as the spectrum, computing and storage constraints in the SAGVN, the joint optimization problem of computation offloading decision and resource allocation was modeled as a mixed integer nonlinear programming problem. Based on the reinforcement learning method, the original problem was transformed into a Markov process, and a deep reinforcement learning algorithm was proposed to solve the problem. The proposed algorithm has the good convergence. The simulation results show that the proposed algorithm outperforms other algorithms in terms of task delay and success rate. © 2023 Editorial Board of Journal on Communications. All rights reserved.