Learning-based gravity estimation for robot manipulator using KRR and SVR

In this paper, a learning-based method for estimating the parameters of the gravity term of a manipulator with the kernel trick approach is presented. This method extracts the mapping equation from the analytical form of the dynamic equation. Based only on the configuration and sampling data of the robotic arm, Kernel ridge regression (KRR) and Support vector regression (SVR) algorithms are introduced to estimate the position parameters and provide a comparison between different learning regression techniques. The novelty of this work is the time-efficient estimation of robot gravity through randomly located joint sampling data using the kernel trick. The optimal solution to the optimal trade-off curve is proposed and discussed. Theoretical analysis shows that the joint angle and driving torque can be used to estimate the relationship between the center of gravity of the manipulator links and the mass of the connecting rod to obtain an accurate dynamic gravity model.