Skip to Main Content
This paper proposes the design of a Takagi-Sugeno-Kang (TSK)-type Recurrent Fuzzy Network (TRFN) using ant colony optimization in real space (ACOR). The TRFN contains feedback loops in each rule. When the TRFN is applied to control a dynamic plant, no a priori knowledge of the plant order is necessary. Only the current state(s) and desired output(s) are fed as TRFN inputs. All of the free parameters in each recurrent rule are optimized using ACOR. The ACOR stores solutions in an archive and updates solutions using selection and Gaussian random sampling processes. The ACOR-designed TRFN is applied to control a dynamic plant for performance verification. Comparisons with other optimization algorithms verify the advantage of ACOR.