[Paper Review] Joint Task Assignment and Wireless Resource Allocation for Cooperative Mobile-Edge Computing

Hong Xing, Liang Liu+2 more|arXiv (Cornell University)|Feb 7, 2018

IoT and Edge/Fog Computing7 references3 citations

TL;DR

This paper proposes a joint task assignment and wireless resource allocation scheme for cooperative mobile-edge computing using TDMA, optimizing time and power allocation alongside task offloading to minimize computation latency under individual energy constraints. The method relaxes the mixed-integer non-linear problem into a convex form, enabling a low-complexity suboptimal solution that significantly reduces latency compared to benchmark schemes.

ABSTRACT

This paper studies a multi-user cooperative mobile-edge computing (MEC) system, in which a local mobile user can offload intensive computation tasks to multiple nearby edge devices serving as helpers for remote execution. We focus on the scenario where the local user has a number of independent tasks that can be executed in parallel but cannot be further partitioned. We consider a time division multiple access (TDMA) communication protocol, in which the local user can offload computation tasks to the helpers and download results from them over pre-scheduled time slots. Under this setup, we minimize the local user's computation latency by optimizing the task assignment jointly with the time and power allocations, subject to individual energy constraints at the local user and the helpers. However, the joint task assignment and wireless resource allocation problem is a mixed-integer non-linear program (MINLP) that is hard to solve optimally. To tackle this challenge, we first relax it into a convex problem, and then propose an efficient suboptimal solution based on the optimal solution to the relaxed convex problem. Finally, numerical results show that our proposed joint design significantly reduces the local user's computation latency, as compared against other benchmark schemes that design the task assignment separately from the offloading/downloading resource allocations and local execution.

Motivation & Objective

• Address the challenge of minimizing computation latency in multi-user cooperative mobile-edge computing systems with energy-constrained users and helpers.
• Formulate a joint optimization problem for task assignment, time, and power allocation under TDMA protocols and individual energy constraints.
• Overcome the computational difficulty of solving the mixed-integer non-linear program (MINLP) by relaxing it into a convex problem.
• Design a low-complexity suboptimal algorithm based on the relaxed solution to enable practical deployment.
• Demonstrate performance gains over schemes that optimize task assignment and resource allocation separately.

Proposed method

• Formulate the latency minimization problem as a mixed-integer non-linear program (MINLP) with constraints on individual energy budgets for the local user and all helpers.
• Relax the integer task assignment variables into continuous variables to transform the MINLP into a convex optimization problem.
• Solve the relaxed convex problem using standard convex optimization techniques to obtain a globally optimal solution.
• Construct a suboptimal task assignment for the original problem by rounding the continuous solution from the relaxed problem.
• Apply time and power allocation jointly with task assignment to minimize total latency under TDMA-based communication scheduling.
• Use a heuristic-based suboptimal design to compare performance, including channel-aware and computation-capacity-aware assignment rules.

Experimental results

Research questions

• RQ1 How can task assignment and wireless resource allocation be jointly optimized to minimize computation latency in a cooperative MEC system with energy-constrained users and helpers?
• RQ2 What is the performance gain of joint optimization compared to schemes that treat task assignment and resource allocation as separate design problems?
• RQ3 How does the proposed suboptimal algorithm based on convex relaxation compare to exhaustive search and heuristic methods in terms of latency and complexity?
• RQ4 What is the impact of energy constraints and helper CPU frequency on system latency in cooperative MEC?
• RQ5 Under what conditions does communication-aware task assignment outperform computation-capacity-aware assignment?

Key findings

• The proposed joint design significantly reduces computation latency compared to benchmark schemes that optimize task assignment and resource allocation separately.
• The proposed suboptimal solution achieves a latency performance within a negligible gap of the near-optimal random search solution under high energy constraints.
• Heuristic-2 (computation-capacity-aware assignment) outperforms random selection only when energy constraints exceed -8.5 dB, as energy imbalance between computation and communication becomes critical at low energy levels.
• As helper CPU frequencies increase, system latency eventually plateaus due to energy constraints limiting communication time, making communication the new bottleneck.
• Heuristic-1 (channel-aware assignment to the best helper) only outperforms local execution when the helper’s CPU frequency exceeds 2 GHz, and becomes infeasible beyond this due to energy constraints.
• With increasing task count, the proposed scheme maintains superior performance, especially for large-scale workloads, demonstrating scalability and robustness.