Evolutionary Multitasking for Cross-domain Task Optimization via Vehicular Edge Computing

Abstract

Efficient optimization is a key enabler for emerging intelligent applications in Internet of Vehicles (IoV). However, existing studies in IoV only focus on solving a single domain-specific optimization problem at a time, which undermines their efficiency on tackling various cross-domain optimization tasks in IoV. In this paper, we make the first effort on investigating a novel optimization framework in IoV for cross-domain tasks via vehicular edge computing. Specifically, two typical cross-domain tasks in IoV are presented, namely, the data dissemination (DD) task and the computing offloading (CO) task. Then, a cross-domain problem called DD-CO is formulated to facilitate the sharing of task features and knowledge during the solution searching. On this basis, we propose an evolutionary multitasking approach named EMA, which consists of an integer based unified representation scheme for encoding both the DD and CO tasks in a single solution, a corresponding decoding operator for task-specific solution evaluation, and a new population evolution mechanism for better adaptation to the cross-domain problem optimization. Finally, we build the simulation model and give a comprehensive performance evaluation, which demonstrate the advancement of the new optimization framework via vehicular edge computing and the effectiveness of the proposed EMA method.