Plug-and-play predictive energy management for fuel-efficient platooning of gasoline-powered autonomous vehicles

In this paper, we present a plug-and-play predictive energy management framework for fuel-efficient platooning of gasoline-powered autonomous vehicles. Unlike existing approaches that target electric or hybrid drivetrains, the proposed architecture addresses the nonlinear thermodynamic and aerodynamic challenges inherent to internal combustion engine (ICE) vehicles. The framework integrates an Adaptive Neuro-Fuzzy Inference System optimized via Teaching-Learning-Based Optimization (ANFIS-TLBO) with Model Predictive Control (MPC) and a Quantum-inspired Long Short-Term Memory (QLSTM) network that broadcasts a 10-second leader state prediction horizon to all follower vehicles simultaneously. A cooperative power-sharing mechanism and a predictive gap policy are superimposed as a lightweight supervisory layer that leaves the internal structure and parameters of each local controller entirely unchanged, modifying only its reference inputs (predicted leader trajectory, desired gap, and coordinated fuel-cut activation). The framework is validated across four standardized driving cycles, HWFET, US06, Artemis HW130, and Artemis HW150, in both 4-vehicle and 6-vehicle homogeneous platoon configurations, each subjected to 20 statistically independent runs incorporating stochastic wind (Weibull) and road grade (Rayleigh) disturbances. Under the HWFET cycle, the proposed method achieves an average per-vehicle fuel consumption of 6.64 L/100 km (4-vehicle platoon) and 6.61 L/100 km (6-vehicle platoon), outperforming all benchmarks including Adaptive MAB, Pure ANFIS, Pure MPC, and Pure TLBO. NOx emission reduction rates from leader to last follower reach 5.74% and 6.23% in the respective configurations, the highest recorded across all methods, reflecting a trade-off in which the leader operates at a higher absolute NOx level to sustain tighter velocity tracking, while followers benefit disproportionately from predictive power management. The plug-and-play architecture is validated on a 6-vehicle platoon with zero controller retuning, confirming scalability across platoon sizes without local-controller modification.