Identification of nonlinear dynamic systems-a fuzzy logic approach and experimental demonstrations

A newly developed dynamic fuzzy logic system and corresponding stable identification algorithm are described and applied to identify the dynamics of a single link mechanical manipulator. The manipulator is characterized by a very flexible link, significant nonlinear joint friction, and an unknown internal actuator control loop, and serves as a challenging testbed for our identification algorithm. Experimental results are presented which demonstrate good convergence of the identification process for various types of inputs. The trained identifiers are tested by producing the response of the manipulator for complex command inputs. The test results are quite satisfactory, indicating that the proposed identification approach is very effective in dealing with complex, highly nonlinear systems, and can be useful in practical applications