A dynamic identification method for general serial manipulators from an analytical perspective based on Lie theory

The dynamic model describes the relationship between the motions of robots and the forces or torques causing these motions. This work describes a novel dynamic identification method for general serial manipulators from an analytical perspective based on Lie theory. Unlike the existing work, the proposed method does not rely on the Kronecker product to obtain the linear identification model from the closed-form dynamic equation. We derived the dynamic equation of a single rigid body into a linear model using symbolic computation. Then using the Newton–Euler method, we obtained the equivalent linear identification model for the manipulator. By analyzing the expression of the identification model, we discussed the impact of different joint types on the linear independent columns of the coefficient matrix in the identification model. Several criteria were proposed for directly extracting the linear independent columns based on the robot’s joint screw axes without the need for experimental data. Furthermore, a recursive method was proposed to obtain the complete expression of the base parameters (BPS). The optimal excitation trajectory and physical feasible constraints are also incorporated into the workflow of the proposed identification method. Compared to the existing dynamic identification method, the method significantly reduces the computational complexity and does not limited to prismatic and revolute joints. The method was demonstrated using a RS010N manipulator. Through a series of experiments, the BPS has been obtained. The differences between the actual torques and the calculated torques from the BPS is quiet small, demonstrating the effectiveness of the identification method. Finally, a set of physical feasible dynamic parameters was searched from the BPS using linear matrix inequalities.