Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Decentralized Task Offloading and Load-Balancing for Mobile Edge Computing in Dense Networks

Mariam Yahya, Alexander Conzelmann, Setareh Maghsudi

TL;DR

The paper tackles decentralized task offloading and load-balancing in dense mobile edge computing by formulating it as a mean-field multi-armed bandit problem with binary rewards influenced by shared resource contention. It proves the existence of a unique mean-field steady state and introduces a decentralized load-balancing mechanism that reshapes arm rewards to steer the population distribution toward a predetermined target $\boldsymbol{f}^{*}$. Through detailed system modeling and extensive simulations, the authors demonstrate convergence to the target load profile under realistic channel, task-size, and network-sharing uncertainties, while maintaining scalability in dense networks. The approach offers a scalable, information-efficient framework for balancing load across heterogeneous edge servers in dynamic MEC environments, with practical implications for low-latency computation offloading in 5G/6G contexts.

Abstract

We study the problem of decentralized task offloading and load-balancing in a dense network with numerous devices and a set of edge servers. Solving this problem optimally is complicated due to the unknown network information and random task sizes. The shared network resources also influence the users' decisions and resource distribution. Our solution combines the mean field multi-agent multi-armed bandit (MAB) game with a load-balancing technique that adjusts the servers' rewards to achieve a target population profile despite the distributed user decision-making. Numerical results demonstrate the efficacy of our approach and the convergence to the target load distribution.

Decentralized Task Offloading and Load-Balancing for Mobile Edge Computing in Dense Networks

TL;DR

The paper tackles decentralized task offloading and load-balancing in dense mobile edge computing by formulating it as a mean-field multi-armed bandit problem with binary rewards influenced by shared resource contention. It proves the existence of a unique mean-field steady state and introduces a decentralized load-balancing mechanism that reshapes arm rewards to steer the population distribution toward a predetermined target . Through detailed system modeling and extensive simulations, the authors demonstrate convergence to the target load profile under realistic channel, task-size, and network-sharing uncertainties, while maintaining scalability in dense networks. The approach offers a scalable, information-efficient framework for balancing load across heterogeneous edge servers in dynamic MEC environments, with practical implications for low-latency computation offloading in 5G/6G contexts.

Abstract

We study the problem of decentralized task offloading and load-balancing in a dense network with numerous devices and a set of edge servers. Solving this problem optimally is complicated due to the unknown network information and random task sizes. The shared network resources also influence the users' decisions and resource distribution. Our solution combines the mean field multi-agent multi-armed bandit (MAB) game with a load-balancing technique that adjusts the servers' rewards to achieve a target population profile despite the distributed user decision-making. Numerical results demonstrate the efficacy of our approach and the convergence to the target load distribution.
Paper Structure (8 sections, 4 theorems, 20 equations, 2 figures, 1 algorithm)

This paper contains 8 sections, 4 theorems, 20 equations, 2 figures, 1 algorithm.

Table of Contents

  1. Introduction
  2. System Model
  3. Problem Formulation
  4. The Mean Field Model
  5. Task Offloading as A Mean Field MAB Problem
  6. Load-Balancing
  7. Numerical Results
  8. Conclusion

Key Result

Theorem 1

Let $L$ be the Lipschitz continuity constant. If the following conditions hold for all $a {\in} [0,1]$ and $x, x' {\in} [0,1]$: where $\beta$ is the continuation probability, then for any policy $\sigma$, there exists a unique fixed MFSS.

Figures (2)

  • Figure 1: The fraction of agents selecting arm 1, $f_1$, for $100$ and $10^4$ agents before (blue) and after (orange) applying the load-balancing algorithm. The value of $f_1^{*}$ is $0.2$, and there are $300$ optimziation steps.
  • Figure 2: The mean distance between $f$ and $f^{*}$ over 150 optimization steps for networks with two (orange) and eight (green) servers with $100$ and $10^4$ agents in each case.

Theorems & Definitions (7)

  • Theorem 1
  • Proposition 1
  • proof
  • Proposition 2
  • proof
  • Proposition 3
  • proof