Soft computing system using fuzzy clustering and on-line learning

An attractive self-organizing fuzzy system (SOFS) with on-line learning ability is proposed in this paper. The SOFS possesses the ability of self-organization and on-line learning for its knowledge base in application while the plant is viewed as a "black box" for the SOFS. A novel concept of pseudo-errors is utilized to capture the characteristic of plant behavior, and the information of pseudo-errors is considered as the behavior information of a plant. With statistical regression, observed data pairs of pseudo-errors are collected and analyzed. Input space formed by pseudo-errors is partitioned into several regions in cluster form. In cluster-based fuzzy reasoning, the partitioned region in the input space is viewed as the antecedent of a fuzzy if-then rule, in which Gaussian membership functions are used. Based on random optimization algorithm, an on-line learning method is proposed in the paper. It enables the SOFS to learn on tracking a reference curve directly. The learning iteration is divided into several sub-parts to make the on-line learning feasible by which the system parameters are updated in a real-time style to search for the optimal fuzzy sets of if-then rules. The T-S fuzzy model is utilized, where the consequent parts are in the form of linear function. Inferred result does not require defuzzification because the consequents of the fuzzy rules are crisp. Motion control of an auto-warehousing crane system is used to illustrate the feasibility of the proposed approach.