Fuzzy Q-learning-based energy management for fuel cell mobile robot considering multi-scenario parameter

Fuel cell hybrid mobile robots present significant potential for enhancing manufacturing while promoting environmentally friendly transportation. However, the power performance of mobile robots largely depends on the effective coordination of hybrid energies in the multi-scenario parameters environment. To solve the above problem, a three-dimensional fuzzy Q-Learning (3D-FQL) method is proposed for energy management that autonomously distributes the energy power in different parameter scenarios. Firstly, the topology framework of the hybrid mobile robot with a fuel cell and battery is established. Based on the model, the relation between multi-scenario parameters and system performance metrics is analyzed. Secondly, to improve the economy, efficiency and durability performance, a 3D-FQL algorithm scheme is introduced into the energy management strategy (EMS) process. Moreover, the Markov decision process (MDP) combined with the fuzzy logic system (FLS) and multi-scenario concept is designed to improve the generalization of the Q-learning (QL) algorithm. Finally, different simulation and hardware-in-loop (HIL) tests are selected to verify the optimization performance of the proposed method. Compared with the other EMS algorithms, the proposed method has improved the overall performance in case 1 by 8.93% and 12.21%, in case 2 by 12.02% and 11.29%, in case 3 by 14.45% and 13.12%, respectively. • The system topology of fuel cell hybrid mobile robot is presented. • A reinforcement learning integrated with the multi-scenario parameters is proposed. • The generalization performance of the Q-learning method is improved. • Simulation and experiment are offered to validate the proposed method. • Proposed method has improved performance about 23.5% compared with the PSO method.