Parking Vehicle-Assisted Task Offloading in Edge Computing: A dynamic multi-objective evolutionary algorithm with multi-strategy fusion response

Abstract

Vehicle-edge computing, as a promising paradigm, is employed to support applications that require low latency and high computational capability. In this study, we consider the idle resources of the surrounding parked vehicles (PVs) and roadside units (RSUs) as service providers to enhance the performance of User Equipment (UE). We propose a joint offloading architecture that uses parked vehicles. Additionally, owing to the dynamic and uncertain nature of the environment, we model computation offloading as a dynamic multi-objective optimization problem to simultaneously optimize the latency and energy consumption of UE applications. In this study, we propose a dynamic multi-objective evolutionary algorithm with a multi-strategy fusion response (DMOEA/D-MSFR). Specifically, we introduce a population center positioning strategy and a learnable prediction mechanism using Long Short-Term Memory (LSTM) in DMOEA-MSFR, which divides the prediction optimization process into two stages and exhibits a rapid response to environmental changes. In the static optimization phase, an adaptive weight vector adjustment strategy is employed, which significantly aids in the distribution and diversity of the solutions. Comprehensive experiments demonstrate that our proposed framework balances the trade-off between latency and energy consumption, and the convergence, feasibility, and diversity of the non-dominated solutions obtained.