Online task offloading algorithm based on multi-objective optimization caching strategy

Abstract

Within the realm of Mobile Edge Computing (MEC), task offloading has consistently garnered significant attention. Within the context of intricate caching environments and multi-user scenarios, conventional solutions frequently encounter difficulties in simultaneously fulfilling the demands for latency reduction and energy consumption optimization. This paper presents a novel online task offloading algorithm that leverages a multi-objective optimization caching strategy. This algorithm addresses two challenges: the Online Task Offloading (OTO) problem and the Online Task File Caching (OTFC) problem. The OTO problem is conceptualized as a multi-user game, where Nash equilibrium is employed to effectively characterize and address it. This ensures the determination of the optimal offloading strategy in the presence of various caching scenarios. Meanwhile, the OTFC problem is transformed into a Markov decision process, and through the utilization of Deep Q-Networks, we can forecast the requirements of online tasks and subsequently determine the optimal caching vector. The incorporation of the Multi-Objective Cache Policy (MOCP) algorithm precedes the finalization of the caching vector. Rooted in multi-objective optimization, this algorithm adeptly balances various caching decisions, achieving a Pareto optimal outcome. The proposed offloading model that effectively caters to the requirements of task offloading while incorporating the demands of task file caching. Moreover, the MOCP algorithm ensures optimal caching decisions across a broad range of scenarios. Simulation tests reveal that this enhanced offloading algorithm, grounded in multi-objective optimization, outperforms traditional methods in energy conservation, boasting energy savings of up to 15%.