Identification methods for adaptive manipulator control

Models for industrial robots are characterized by highly nonlinear equations with nonlinear coupling between the variables of motion. This paper examines possible model structures suitable for identification of manipulator dynamics. Attention is focussed on single input/single output models with a view to the implementation of decentralized control. Both linear and simple nonlinear models are considered. The nonlinear models are based on the Hammerstein and Volterra structures for which modified linear control methods have already been developed. Implementation results from these methods are presented for a PUMA 560 robot to demonstrate the applicability of each method. Some implementation details are also discussed