A decentralized scheme for multi-user edge computing task offloading based on dynamic pricing

doi:10.1007/s12083-025-01904-1

Return

A decentralized scheme for multi-user edge computing task offloading based on dynamic pricing

2025-02-10

PRE

Zheng, Liquan

Long Tan *

DOI：10.1007/s12083-025-01904-1

Original

Original request for help

Save

Abstract

En 中文

The research on Mobile Edge Computing (MEC) has attracted considerable attention in recent years. However, previous studies often confined themselves to exploring the issue of task resource scheduling from a single perspective of mobile terminals or edge servers, or failed to take into account the growth of users, thus unable to face complex and diverse practical scenarios and effectively address various challenges in task offloading. In view of this, this paper adopts a more comprehensive and in-depth perspective, considering not only the user's performance benefits and the service provider's economic benefits, but also conducting a thorough analysis of the potential continuous growth of user numbers in actual scenarios. To effectively address the rapid increase in user numbers within large-scale networks, we employ a decentralized approach where each user independently makes decisions regarding their own offloading strategies, rather than relying on centralized decision-making by the server. Based on this, we propose an innovative edge computing task offloading scheme that is particularly suitable for large-scale network environments. Specifically, we combine the Twin Delayed Deep Deterministic Policy Gradient (TD3) algorithm and innovatively propose a decentralized task offloading scheme called DSMECO-DP. In this scheme, considering the dynamic nature of networks and user volatility, the server aims to maximize profit through coarse time scale dynamic pricing, while the mobile terminal reduces its own costs by bidding on a narrow time scale considering multiple objectives such as time delay, energy consumption, and payment cost. Simulation results demonstrate the effectiveness of the TD3 algorithm compared to other reinforcement learning algorithms, as well as the effectiveness of the dynamic pricing mechanism. Furthermore, the processing time of this algorithm is significantly reduced compared to other centralized algorithms.

Keywords:

Mobile edge computing

Dynamic pricing

Deep reinforcement learning

Computation offloading