Reinforcement Learning Hierarchical Neuro-Fuzzy Model for Autonomous Robots

This work presents the application of a new hybrid neuro-fuzzy model, called RL-HNFP (Reinforcement Learning – Neuro Fuzzy Hierarchical Politree), in automatic learning of autonomous robots. This model provides an agent with intelligence, making it capable, by interacting with its environment, to acquire and retain knowledge for reasoning (infer an action). Promising results have already been obtained in tests with a Khepera robot simulator in a simplified environment (low complexity). The objective of this work is to evaluate the RLHNFP model performance in a bigger and/or more complex environment, carrying out the necessary modifications in the original model to achieve good performance in this new environment. The adapted system was evaluated again with Khepera and the obtained results show the potential of this modified model to efficiently interact in more elaborated situations.