Model parameter on-line identification with nonlinear parametrization – manipulator model

This paper presents an example of solving the parameter identification problem in the case of a robot with two degrees of freedom. In this study, a weighted recursive least squares algorithm was generalised to a case of nonlinear parameterisation in which the identified parameters did not satisfy the linear model. The generalisation involved linearising the model in the neighbourhood of current values of the parameter estimates. It was assumed that the estimates were updated every N steps of signal sampling. This method of identification can be applied whenever the parameters concerning a model need to be determined at the time of measurement. This is particularly useful in adaptive control when the plant parameters vary over time.