Improving Recursive Dynamic Parameter Estimation of Manipulators by knowing Robot's Model integrated in the Controller

Abstract

By identifying the manipulator's model that is integrated in some industrial model-based controllers, recursive parameters’ estimation algorithms can be enhanced to have a better performance in online applications. In this paper, two improvements on recursive estimation of robot's dynamic parameter estimation are addressed. Firstly, the internal model can serve to initialize the parameters in recursive estimation algorithms, as the Recursive Least-Squares (RLS) and the Recursive Instrumental Variables (RIV). Secondly, the commanded position, which is used by the controller as a reference trajectory, can replace the external simulation of the dynamic model needed for recursive algorithms as the RIV. These two improvements make recursive algorithms more suitable for online application, specially RIV, where no data filtering nor external simulation needs to be done. Offline experimental validation on the KUKA LBR iiwa R820 is carried out, showing its feasibility for online application.