Fuzzy Q-Learning with Fuzzified Bellman Equation for Unmanned Ground Vehicle Navigation

This paper presents a novel Fuzzy Q-Learning with Fuzzified Bellman Equation (FQL-FBE) that is utilized to advance autonomous navigation in dynamic and unknown environments. A modified Fuzzified Bellman equation is introduced to integrate fuzzy logic into reinforcement learning by updating the Q-table using Fuzzy temporal differences. The proposed approach serves as a state-action value estimator, addressing challenges associated with continuous state-action spaces by introducing discretization for enhanced learning efficiency. Key advantages include low computational complexity, high adaptability to changing environmental conditions, effective handling of complex scenarios, smooth path generation, interpretability, and a fast learning process. The results of simulations and comparisons validate the efficiency of the proposed method, demonstrating enhanced obstacle avoidance, improved path efficiency, and reduced computational cost. This work marks a step forward in the development of robust and efficient solutions for real-time robotic navigation.