Dynamically Adaptive RSUs in VANETs: An Approach Focused on Sustainability

Abstract

ABSTRACT Vehicular Ad Hoc Networks (VANETs) are increasingly pressured by applications that require high computational performance and energy efficiency from Roadside Units (RSUs), especially in dynamic and unpredictable traffic conditions. This work proposes an adaptive autoscaling model for RSUs, aiming to help system architects reduce energy consumption and carbon emissions without compromising service responsiveness. The model is developed using Stochastic Petri Nets (SPNs) and integrates a fault‐tolerant reinstantiation mechanism. It was experimentally validated, proving the statistical agreement between simulated and real response times. In addition, a sensitivity analysis was performed through Design of Experiments (DoE) to identify the main energy impact parameters. The strategy achieved up to 20% lower energy consumption. In the autoscaling case study, total energy stabilized at 7.5 kWh versus 9.0 kWh without autoscaling, and the carbon footprint at 0.78 vs. 0.90 CO 2 /h. Model validation showed MRT within the experimental 95% CIs across all loads ( p > 0.05; max. Absolute deviation 5.4%). DoE indicated container backlog (C_R = 0.021) and failure probability (P_F = 0.019) as the strongest effects, with the C_R Q_S interaction = 0.017. The proposed model effectively balances sustainability and performance in VANETs, providing a reliable basis for planning adaptive, resilient, and energy‐aware vehicular infrastructures.