Computing Offloading Based on Deep Reinforcement Learning For Virtual Reality Scene

Abstract

In virtual reality scene, computing offloading is a potential technology to improve rendering and drive applications to the ground. However, MEC servers are usually deployed in a fixed manner in base stations with different demands on computational resources, so in this paper, deep reinforcement learning and digital twin techniques are integrated into an edge computing framework to reduce computational latency and transmission latency in virtual reality scenarios so that different computational resource demands can be met. It designs an computing offloading process for virtual reality scenarios, which is solved by deep reinforcement learning algorithms. Simulation results show that the proposed method can improve the transmission latency and computation latency of multimedia data under virtual reality.