Title: Combining coded computation and reinforcement learning to improve edge computing in heterogeneous clusters

Authors: Payam Mohammadi; Farzad Parvaresh; Javad Kazemitabar

Addresses: Department of Electrical Engineering, University of Isfahan, Isfahan, Darvazeh Shiraz, Iran ' Department of Electrical Engineering, University of Isfahan, Isfahan, Darvazeh Shiraz, Iran ' Department of Electrical and Computer Engineering, Babol Noshirvani University of Technology, Babol, Mazandaran, 4714873113, Iran

Abstract: The emergence of Internet of Things (IoT) technology has led to extensive connections among various devices which leads to a production of a large amount of heterogeneous data. While, cloud computing is a suitable and efficient processing model for storing and processing this large data, the demand for real-time and delay-sensitive applications is increasing rapidly. Using only cloud computing cannot address this problem properly, because the network bandwidth is limited. Therefore, edge processing as a new processing model and a cloud computing supplement is proposed which is based on a distributed processing architecture. In the proposed reinforcement learning and distributed coding computing (RLCDC) method we harness the DDPG algorithm to manage resources. Additionally, maximum distance separable (MDS) codes are also used to deal with the remaining processors, and to remove the transmitted packets, as well as increasing the network's operational capacity and throughput.

Keywords: edge computing; coded computing; reinforcement learning; DDPG; deep deterministic policy gradient; IoT; Internet of Things.

DOI: 10.1504/IJCNDS.2025.149496

International Journal of Communication Networks and Distributed Systems, 2025 Vol.31 No.6, pp.649 - 667

Received: 10 Nov 2023
Accepted: 25 Apr 2024
Published online: 04 Nov 2025 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article