Learning a hierarchical fuzzy system with autonomous navigation as an example

In this paper, a hierarchical fuzzy system for high-dimensional dataset is proposed. The sequential least-squares method is introduced to estimate Takagi-Sugeno rules. A hierarchical clustering takes place in the product space of input and output, and each path from the root to a leaf corresponds to a fuzzy if-then rule. Only a subset of the rules is considered, based on the location of the input query data. At each level of the hierarchy, a discriminating subspace is generated automatically from the high-dimensional input space for a good generalization capability. Both a synthetic data set and a real robot autonomous navigation experiment are considered to illustrate how effective the system is.