Online Identification of Robot Manipulator Dynamics Using Recursive Least Squares with Nonlinear Functions

A robot manipulator, also known as a robotic arm, is a nonlinear, multi-input and multi-output (MIMO) system. Deriving the equations for nonlinear system dynamics is challenging due to the complex nature of nonlinear systems, which are not easily modeled with linear equations. The Recursive Least Squares (RLS) method is traditionally used for modeling systems with streaming data that are continuously updated. Initially, RLS was employed for identifying linear systems, however, it can be adapted to identify nonlinear systems by modifying the input variables to produce equations that describe nonlinear system dynamics. This study proposes a Recursive Least Squares with Nonlinear function (N-RLS) using Dynamic Expression Nonlinearization (DEx-N), an extension of Linear RLS, to model robot manipulators. The DEx-N Algorithm transforms input variables into nonlinear forms, enabling RLS to be effectively used for nonlinear systems. The results obtained using N-RLS demonstrated good accuracy on test data, with values of 92.38 %, 80.01 %, and 90.30 %, and RMSE values of 0.82, 1.09, and 0.91 for joints 1, 2, and 3, respectively. Moreover, the N-RLS method is capable of producing accurate nonlinear system equations..