On the fast robot dynamic parameters learning

A computationally efficient solution to the problem of identifying the dynamic parameters of a robot manipulator is presented. The identification procedure is based on a simplified form of the dynamics. The approach has three important characteristics. First, being based on the Lagrangian representation, the equations are linear in the dynamic parameters, which makes possible the application of linear identification techniques. Second, the dynamic parameters are easily recognized, extracted, and grouped. Third, the equations are amenable to the implementation of parallel processing schemes. For the identification a recursive least squares algorithm is used. The algorithm is distributed over a network of transputers. Real-time results have been produced to demonstrate the speedup and efficiency of the proposed technique. A case study is given for the first three links of the Stanford Arm positioning system