A many-objective evolutionary algorithm for solving computation offloading problems under uncertain communication conditions

Abstract

Computation offloading is a critically important technology in the field of edge computing, enabling improved conditions for task computation. Existing studies on computation offloading largely focus on cooperative offloading and resource allocation. The impact of factors such as uncertainty during task transmission and the different task characteristics on the offloading location is ignored. To address the above issues, this paper designs a many-objective optimization computation offloading model (MaOCO) under uncertainty to cope with offloading requirements. The model formulates channel transmission level strategies and considers five objectives: latency, cost, energy consumption, load balancing, and user satisfaction. A many-objective evolutionary algorithm with deviation selection (MaOEA-DS) is designed to obtain effective offloading strategies. In the environment selection, the deviation is calculated using the concept of variance to choose solutions closer to the origin for the last layer, which further improves the convergence of the algorithm. The simulation results show that MaOEA-DS outperforms other evolutionary algorithms in IGD, GD, SP, and HV.