Gradient Bandit Experts for Node Selection in Edge Computing

Node selection is a critical issue in Edge Computing (EC), necessitating the identification of appropriate edge nodes for effective execution of tasks. This process entails maneuvering through limitations such as restricted resources, fluctuating network conditions, and heterogeneous processing capabilities among nodes. The efficacy and overall performance of EC systems are significantly reliant on the optimization of node selection algorithms. Due to the dynamic and varied characteristics of EC environments, adaptive strategies utilizing real-time feedback are crucial. To tackle this issue, we present ExpGradBand, an innovative technique designed for node selection in EC. ExpGradBand expands the Multiarmed Bandit (MAB) framework by using gradient-based selection methods alongside historical contextual data to improve decision-making performance. We assess the usefulness of ExpGradBand by comprehensive experiments in comparison to leading algorithms. The findings emphasize the exceptional value of ExpGradBand.