Reinforcement learning and tuning for neural network based fuzzy logic controller

Proposes a reinforcement neural network based fuzzy logic controller (RNNFLC) for solving reinforcement learning and tuning problems. A simplified and effective reinforcement based learning algorithm is proposed to generate fuzzy rules automatically. A reinforcement-tuning algorithm is introduced to tune the membership function. The RNNFLC is applied to a cart-pole balancing system and shows significant improvements.