Maximizing Revenue in Roadside Green Edge Server System

The autonomous vehicle industry is growing rapidly and requires substantial computational resources for tasks such as perception and decision-making. Vehicular edge computing addresses this need by offloading computations to roadside edge servers. In alignment with green cloud computing trends, these servers are increasingly powered by solar energy and are equipped with batteries to maintain continuous operation during periods of low solar availability. We refer to such infrastructure as Green Edge Server System (GESS). In this work, we maximize the revenue of GESS by efficiently scheduling offloaded computational tasks where the tasks are characterized by arrival time, deadline, resource demand, and revenue. We consider two variants of the GESS: one with an infinite battery and another with a finite battery. For each cases, we propose customized heuristic for solutions for offline scenario by exploiting predicted solar power and incoming task patterns. After this, we present a online strategy which adjusts offline schedule dynamically based on real-time deviations in solar power and incoming task pattern. Our approach takes advantage of the nonlinear behavior of the cubic power model to minimize cost through efficient task batching. Experimental evaluations on real-life datasets demonstrate that our methods outperform state-of-the-art baselines by up to 40%. Additionally, the system remains robust under uncertainty, with revenue decreasing by only 4.68% when subjected to 10% spikes in task load and power shortages.