工业物联网设备会将无法进行本地计算的任务发送至边缘服务器进行处理,但不同设备密度下的覆盖会导致不同边缘服务器的计算任务负载不均衡,进而产生计算时延过大的问题.为了解决这个问题,提出了一种基于改进的深度确定性策略梯度(modified deep deterministic policy gradient,MDDPG)的任务迁移算法,该算法具有基于深度确定性策略梯度的优先经验重放和随机权重平均机制,以寻求最佳的迁移策略,减少任务的计算时延.实验结果表明,MDDPG算法相较于传统的算法有更好的性能.
Abstract
IIoT (industrial Internet of Things) devices send tasks that cannot be computed locally to edge servers for processing. However, different device densities result in imbalanced computational workloads among various edge servers, leading to significant computation latency. To solve this problem, a task migration algorithm based on modified deep deterministic policy gradient (MDDPG) is proposed. The algorithm has a mechanism of priority empirical replay and random weight averaging based on depth deterministic strategy gradient to find the best migration strategy and reduce the computation delay of the task. Experimental results show that MDDPG algorithm has a better performance than the traditional algorithms.
边缘计算是工业物联网(industrial Internet of Things,IIoT)中解决任务量增长、减少计算时延和提高服务质量的重要技术之一.工业物联网设备会将无法本地计算的任务发送至边缘服务器进行计算,但不同设备密度下的覆盖会导致不同边缘服务器的计算任务负载不均衡,进而产生计算时延过大的问题[1-2].为了解决这个问题,学者们广泛关注边缘服务器间的协作[3-4],通过有效通信,高负载边缘服务器可以将自身需要处理的计算任务迁移到其他低负载边缘服务器,从而减轻高负载边缘服务器的负担并优化整体时延,而在此过程中需要充分考虑网络时延的影响[5],因此,合理的任务迁移算法是必要的[6].
ZhangN, FangX J, WangY,et al.Physical‑layer authentication for Internet of Things via WFRFT-based Gaussian tag embedding[J].IEEE Internet of Things Journal,2020,7(9):9001-9010.
[2]
SisinniE, SaifullahA, HanS,et al.Industrial Internet of Things:challenges,opportunities,and directions[J].IEEE Transactions on Industrial Informatics,2018,14(11):4724-4734.
[3]
AazamM, ZeadallyS, HarrasK A.Deploying fog computing in industrial Internet of Things and Industry 4.0[J].IEEE Transactions on Industrial Informatics,2018,14(10):4674-4682.
[4]
ZhangN, WuR Y, YuanS L,et al.RAV:relay aided vectorized secure transmission in physical layer security for Internet of Things under active attacks[J].IEEE Internet of Things Journal,2019,6(5):8496-8506.
[5]
WuY, NiK J, ZhangC,et al.Noma-assisted multi‑access mobile edge computing:a joint optimization of computation offloading and time allocation[J].IEEE Transactions on Vehicular Technology,2018,67(12):12244-12258.
[6]
KuangZ F, LiL F, GaoJ,et al.Partial offloading scheduling and power allocation for mobile edge computing systems[J].IEEE Internet of Things Journal,2019,6(4):6774-6785.
[7]
HeY J, YangM N, GuizaniH M.Resource allocation based on digital twin‑enabled federated learning framework in heterogeneous cellular network[J].IEEE Transactions on Vehicular Technology,2023,72(1):1149-1158.
[8]
LiuM Y, LiuY.Price‑based distributed offloading for mobile‑edge computing with computation capacity constraints[J].IEEE Wireless Communications Letters,2018,7(3):420-423.
[9]
LinH, HuJ, WangX D,et al.Towards secure data fusion in industrial IoT using transfer learning[J].IEEE Transactions on Industrial Informatics,2021,17(10):7114-7122.
[10]
ZhouZ Y, LiaoH J, GuB,et al.Resource sharing and task offloading in IoT fog computing:a contract‑learning approach[J].IEEE Transactions on Emerging Topics in Computational Intelligence,2020,4(3):227-240.
[11]
ChaiR, SongX, ChenQ B.Joint task offloading,CNN layer scheduling,and resource allocation in cooperative computing system[J].IEEE Systems Journal,2020,14(4):5350-5361.
[12]
ZhangD Y, TanL, RenJ,et al.Near‑optimal and truthful online auction for computation offloading in green edge‑computing systems[J].IEEE Transactions on Mobile Computing,2019,19(4):880-893.
[13]
AngC W, ThamC K.Analysis and optimization of service availability in a HA cluster with load‑dependent machine availability[J].IEEE Transactions on Parallel and Distributed Systems,2007,18(9):1307-1319.