基于DDPG的MEC视频任务卸载算法

Video Task Offloading Algorithm Based on DDPG in MEC

近年来移动边缘计算(Mobile Edge Computing, MEC)的出现满足了边缘设备(Edge Device, ED)处理视频数据任务的需求,ED通过将任务卸载到MEC服务器,来缓解自身计算能力不足的缺点。然而MEC系统中网络的时变性和任务生成的动态性,为求解任务卸载问题带来了挑战。本文考虑一个多ED的MEC系统,将MEC系统的任务处理时延作为优化目标。为了最小化任务处理时延,将原问题模型转化为马尔科夫决策过程(Markov Decision Process, MDP)。考虑到任务动态性及网络时变性,本文采用一种基于深度强化学习的深度确定性策略梯度算法(Deep Deterministic Policy Gradient, DDPG)来解决MDP问题。仿真结果表明,该算法能够有效地降低任务处理时延,并优于其他基准卸载策略。

In recent years, the emergence of Mobile Edge Computing (MEC) meets the needs of Edge Device (ED) to process video data tasks. ED alleviates its lack of computing capacity by offloading tasks to MEC server. However, the time variability in MEC and the dynamics of task generation bring challenges to the task offloading problem. In this paper, a multi-EDs MEC system is considered, and the task processing latency of the MEC system is taken as the optimization objective. In order to minimize the latency of task processing, the original problem is transformed into a Markov Decision Process (MDP). Considering the dynamics of task and time variability of network, this paper exploits the Deep Deterministic Policy Gradient (DDPG) algorithm based on deep reinforcement learning to solve the MDP problem. Simulation results show that this algorithm can effectively reduce the task processing latency and is better than other baseline offloading policies.