A Novel Approach for Computation Offloading Based on a Parallel Collaborative Genetic Algorithm in MEC

• Published in: Wireless personal communications Vol. 140; no. 3-4; pp. 1119 - 1146
• Main Authors: Li, Wenzao, Tang, Ran, Wang, Xiaoke, Zhang, Xiaoming, Ren, Dehao, Jiang, Hong, Wen, Zhan
• Format: Journal Article
• Language: English
• Published: Dordrecht Springer Nature B.V 01.02.2025
• Subjects: Bandwidths; Collaboration; Computation offloading; Convergence; Convex analysis; Cost function; Data processing; Edge computing; Energy consumption; Fitness; Genetic algorithms; Internet of Things; Multiple objective analysis; Optimization; Optimization techniques; Parallel processing; Solution space; User experience
• ISSN: 0929-6212; 1572-834X
• DOI: 10.1007/s11277-025-11760-0

Edge computing relocates computational resources closer to user terminals, which can effectively reduce the latency of computation offloading. The Genetic Algorithm (GA) performs well in multi-objective optimization problems. However, GA exhibits a slow convergence rate when identifying the optimal solution within extensive solution spaces. Therefore, this paper proposes a Parallel Cooperative Genetic Algorithm (PCGA) to find an optimal solution more quickly. This algorithm improves implementation efficiency by utilizing parallel computation. In our formulated computing offloading model, two processes collaboratively work to find the optimal solution, retaining the top five individuals with the highest fitness values in each iteration to update the population of GA. Furthermore, we designed a more comprehensive cost function to optimize objectives such as latency, energy consumption, and the quantity of offloaded tasks. Moreover, we considered a task priority offloading model that prioritizes the offloading of higher-priority tasks based on demand in multi-user and multi-task scenarios. The simulation results demonstrate that the proposed PCGA achieves higher fitness and faster convergence with the same fitness evaluation function. Specifically, with the same fitness evaluation function, the convergence speed of the PCGA algorithm is improved by 50% relative to that of the GA algorithm and 40% relative to that of the TOGA algorithm. In addition to this, PCGA significantly reduces the latency of the task offloading model by 19.7% and 4.4% compared to GA and TOGA, respectively.