Neuro-fuzzy controller modeling for mobile robot navigation using a rule base simplification strategy

The neuro-fuzzy approach is a highly effective hybrid solution for modeling control systems used in several fields. In robotics, several recent studies have adopted thearchitecture of the adaptive network-based fuzzy inference system (ANFIS) to design controllers for the autonomous navigation of mobile robots. However, most of the resulting models face problems of complexity and interpretability regarding the fuzzy rule base. In this work, we have proposed a combination of methods that considers the precision/complexity trade-off to design controllers capable of performing reactive, efficient and smooth navigation in previously unknown environments. This approach focuses on simplifying the fuzzy rule base based on data-driven modeling, merging the most similar fuzzy sets and reducing redundant rules. The effectiveness of the proposed approach is demonstrated through simulation experiments and comparisons with other state-of-the-art methods.