Multilayer Type-2 Fuzzy Q-Learning-Based Energy Management for Fuel Cell Hybrid Mobile Robot System with Varying Temperature

Abstract Fuel cell hybrid mobile robots emerge as a promising solution for enhancing manufacturing productivity while enabling eco-friendly industrial operations. Nevertheless, their dynamic power performance critically depends on coordinated management of hybrid energy systems under fluctuating environmental conditions. This investigation develops a multilayer type-2 fuzzy Q-Learning (T2-FQL) energy management strategy (EMS) to achieve autonomous power allocation optimization across diverse temperature environments. The research methodology encompasses four key aspects: First, we establish a comprehensive mathematical framework characterizing the fuel cell-battery hybrid architecture in mobile robotic systems. Subsequent the relation between temperature with critical system performance parameters is analysed. To simultaneously address multi-objective optimization requirements in economy, efficiency, and durability, we innovatively integrate fuzzy logic systems (FLS) with Markov decision processes (MDP) through a multilayer T2-FQL architecture. This synergistic approach effectively enhances Q-learning generalization capabilities while maintaining operational scenario adaptability. Finally, hardware-in-the-loop (HIL) validation demonstrates the proposed EMS’s superior optimization performance, particularly in varying temperature conditions.