A Novel ANFIS-Based Approach for Optimizing Energy Efficiency in Autonomous Vehicles

Autonomous vehicles (AVs) promise improved safety and sustainability, yet their sophisticated sensing, computing, and communication systems impose auxiliary power loads of 1.5–3.2 kW, risking an increase of up to 45% in global transport energy demand by 2040 if left unaddressed. Existing energy management strategies fail to jointly optimize propulsion and autonomy subsystems under real-world dynamic traffic, treat ADAS loads as static, and lack statistically rigorous validation. This paper proposes a novel Adaptive Neuro-Fuzzy Inference System (ANFIS)-PID framework that integrates (i) 5 s V2X traffic preview, (ii) online PID gain scheduling, and (iii) energy-aware rule pruning for real-time energy allocation. Validated on a real-world trajectory dataset, the approach consistently reduces fuel consumption by up to 4.4% over pure fuzzy logic, 0.05% over FL-RWOA, 1.16% over FL-GWO, and 2.39% over FL-PSO across 25–100 km segments (paired t-test, p ≤ 0.001 on 50 random segments). Additional benefits include 18% faster transient response and 18% lower inference computational load compared to metaheuristic baselines. Scaled to fleet level, the 0.51 L/100 km average saving equates to over CAD 100 million annual savings in Canada. The hybrid neuro-fuzzy architecture offers a deployable, adaptive solution for sustainable autonomous transportation.