MEDALS: A sustainable AI framework for energy-efficient routing in 5G vehicular networks

Abstract

Intelligent transportation systems require routing protocols that optimize both performance and environmental impact simultaneously in 5G-enabled Vehicular Ad Hoc Networks (VANETs). Existing solutions often treat sustainability as a secondary constraint, which limits their effectiveness in addressing climate change goals. This study presents MEDALS (Metaheuristic-Enhanced Deep Adaptive Learning System), a hybrid framework that integrates deep reinforcement learning with metaheuristic optimization to achieve both superior performance and environmental sustainability. The system introduces the Green Performance Index (GPI), the first comprehensive metric combining energy efficiency, carbon footprint, latency, and reliability. Through extensive evaluation using industry-standard simulators, MEDALS demonstrates statistically significant improvements: MEDALS achieves 96.8 % energy efficiency (+11.6 %), 0.73 ms latency (-91.6 %), 99.7 % reliability, and 42.3 % carbon reduction while scaling to 1000 + vehicles with linear computational complexity. This will allow its practical implementation in smart cities and towards fulfillment of the sustainable development goals. This complexity augmentation of 3.3x times in the network size handling is attributed to the hybrid intelligence architecture of the framework, the adaptive deep reinforcement learning with the dual metaheuristic optimisation in intelligent fusion mechanism, and the empirically quantified O(N log N) complexity.