Reasonable capture: Community detection based fuzzy rule discovery

The fitting accuracy of fuzzy logic systems is fundamentally constrained by the quality of fuzzy rule bases, particularly in high-dimensional state spaces where manual tuning leads to excessive workload, heavy reliance on designers' experience, and lack of systematic methodological guidance. Based on the optimization of graph theory modularity, namely community detection, we propose a method for discovering the fuzzy rule base and conduct a theoretical analysis of the proposed method using mathematical tools. The theoretical analysis shows that, based on sampled data and optimizing modularity to tune the fuzzy rule base, the action intervals of fuzzy rules can better fit the contour lines of the function, reducing the inherent error of each fuzzy rule and improving the fitting accuracy. Simulation experiments demonstrate that the proposed method could significantly enhance fitting accuracy compared with manual tuning and other community detection algorithm-based methods. Additionally, a path planning experiment, avoiding high-temperature areas, is conducted, illustrating the potential application of the proposed method.