RIS-Assisted Multi-MEC Server Joint Task Offloading and Resource Allocation Strategy

0
0
0

RIS-Assisted Multi-MEC Server Joint Task Offloading and Resource Allocation Strategy

Introduction

The rapid development of the Internet of Things (IoT) has led to an explosion of connected devices, generating massive amounts of data that traditional centralized cloud computing struggles to process efficiently. Additionally, complex communication environments, such as urban areas with dense buildings or remote regions with poor connectivity, often result in transmission interruptions and degraded service quality. Mobile Edge Computing (MEC) has emerged as a promising solution by bringing computation closer to end-users, reducing latency and alleviating the burden on cloud data centers. However, MEC servers have limited computational resources, and when multiple users compete for these resources, congestion can occur, leading to performance degradation.

To address these challenges, Reconfigurable Intelligent Surfaces (RIS) have been introduced as a means to enhance wireless communication environments. RIS consists of programmable meta-surfaces that can dynamically manipulate electromagnetic waves to improve signal strength and reliability. By optimizing the phase shifts of RIS elements, the wireless propagation environment can be reconfigured to enhance communication links between users and MEC servers.

This paper proposes a novel RIS-assisted multi-MEC server system for joint task offloading and resource allocation in ultra-dense networks. The goal is to maximize system offloading utility by jointly optimizing RIS phase shifts, user transmit power, MEC server computational resources, and task offloading decisions. The proposed approach leverages alternating optimization techniques to decompose the complex mixed-integer nonlinear programming problem into manageable subproblems, each solved using specialized methods such as quasi-convex optimization and convex optimization. Additionally, an improved heuristic algorithm is introduced to efficiently solve the task offloading decision problem, overcoming the limitations of traditional greedy and exhaustive search methods.

System Model

Communication Model

The system consists of multiple MEC servers, each equipped with a base station (BS), and a RIS deployed to enhance communication between users and MEC servers. The RIS is modeled as a uniform planar array with configurable phase shifts for each element. Users can offload their computational tasks to MEC servers either directly or via RIS-assisted reflection links.

The communication channels include:

  1. Direct Link: The line-of-sight (LoS) path between a user and an MEC server.
  2. RIS-Assisted Link: The reflected path from the user to the RIS and then to the MEC server.

The RIS optimizes phase shifts to maximize the received signal power at the MEC server, thereby improving the signal-to-noise ratio (SNR) and reducing transmission latency.

Task Offloading and Computation Model

Each user generates computational tasks characterized by input data size, computational workload, and local processing capability. Users can choose to execute tasks locally or offload them to an MEC server.

• Local Computation: The task is processed on the user’s device, consuming energy proportional to the computational workload and local CPU frequency.

• Offloaded Computation: The task is transmitted to an MEC server for processing. The offloading process involves:

• Uplink Transmission: The user sends the task data to the MEC server via either the direct or RIS-assisted link.

• Edge Execution: The MEC server allocates CPU resources to process the task.

• Result Download: The computed results are sent back to the user (assumed to be negligible due to small result sizes).

Offloading Utility

The system offloading utility is defined as a weighted sum of task completion time improvement and energy consumption reduction. Each user’s utility is influenced by their preference for latency or energy efficiency, as well as their priority level in the system.

Problem Formulation

The optimization problem aims to maximize the total system offloading utility by jointly optimizing:

  1. RIS Phase Shift Matrix: Adjusts the reflection coefficients to enhance signal reception.
  2. User Transmit Power: Allocates power to minimize interference and energy consumption.
  3. MEC Server CPU Resources: Distributes computational resources among offloaded tasks.
  4. Task Offloading Decisions: Determines which tasks are offloaded to which MEC servers.

The problem is a mixed-integer nonlinear programming (MINLP) problem, which is computationally challenging to solve directly. Therefore, an alternating optimization approach is adopted, decomposing the problem into four subproblems solved iteratively.

Solution Methodology

RIS Phase Shift Optimization

The RIS phase shifts are optimized to maximize the received signal power at the MEC servers. The optimal phase shift for each RIS element is derived based on the geometric distances between the user, RIS, and MEC server.

User Transmit Power Allocation

The power allocation subproblem is formulated as a quasi-convex optimization problem. A bisection method is employed to find the optimal transmit power for each user, balancing energy consumption and transmission rate.

MEC Server CPU Resource Allocation

The CPU resource allocation subproblem is convex, and its closed-form solution is derived using Karush-Kuhn-Tucker (KKT) conditions. The optimal CPU allocation ensures that the MEC server’s total computational capacity is not exceeded while maximizing offloading utility.

Task Offloading Decision

The task offloading decision subproblem is combinatorial and NP-hard. Traditional methods like exhaustive search and greedy algorithms are either computationally infeasible or prone to local optima. To address this, an improved heuristic algorithm is proposed, incorporating:

  1. Initial Solution Generation: A feasible offloading decision is constructed based on single-user optimal associations.
  2. Neighborhood Exploration: Swap and remove operations are used to explore alternative solutions.
  3. Random Restart: Periodically resets the search to avoid local optima.

Performance Evaluation

Simulation Setup

The system is evaluated in a 3 km × 3 km area with multiple MEC servers and randomly distributed users. Key parameters include:
• RIS with 16 reflective elements.

• 3 MEC servers, each with 3 sub-bands.

• User transmit power limited to 20 dBm.

• Task sizes ranging from 800 Mbit to 1300 Mbit.

Results and Analysis

  1. Impact of RIS on Offloading Utility:
    • The proposed RIS-assisted system improves average offloading utility by approximately 24.89% compared to non-RIS scenarios.

    • The improvement is attributed to enhanced signal reception and reduced transmission latency.

  2. Effect of RIS Reflective Elements:
    • Increasing the number of RIS elements further improves offloading utility, with a 2.26% gain observed when doubling the elements.

  3. User Preferences and System Performance:
    • Higher preference for low latency (αₜ) reduces task completion time but increases energy consumption.

    • As the number of users increases, resource contention leads to higher latency and energy usage.

  4. Algorithm Comparison:
    • The proposed heuristic algorithm outperforms traditional greedy and local search methods by approximately 14.02% in offloading utility.

    • It also approaches the optimal solution obtained via exhaustive search while remaining computationally feasible for large networks.

Conclusion

This paper presents a comprehensive RIS-assisted multi-MEC server framework for joint task offloading and resource allocation. By leveraging RIS to enhance communication links and optimizing resource allocation, the system achieves significant improvements in offloading utility, particularly in ultra-dense networks. The proposed alternating optimization approach efficiently decomposes the problem into tractable subproblems, while the improved heuristic algorithm ensures scalability and near-optimal performance.

Future work may explore multi-RIS deployments, dynamic RIS positioning, and collaborative MEC server architectures to further enhance system performance.

doi.org/10.19734/j.issn.1001-3695.2024.05.0221

Was this helpful?

0 / 0