Improved numerical technique for industrial robots model reduction and identification

This paper addresses the robot dynamic model derivation and reduction based on a numerical technique. The aim is to obtain a final driven model landing itself to the identification procedure without any symbolic calculation. Besides, the real robot characteristics and limits are accounted for in the model reduction for significant simplification. The proposed model simplification is basically performed using a numerical categorization and elimination of the dynamic parameters. This technique, unlike the existing numerical ones, performs an approximation yielding more consistent parameters values and reduces the model complexity according to the geometric limits and kinematic constraints. The simplification is based on tow criteria that have been designed to automatically assess the relevance of including a given dynamic parameter in the final model. A simulation is carried out as a walk-through for this method and to show its efficiency in the case of pseudo-ideal data acquisition. In the simulation study, the algorithm setting are discussed as well as the obtained results. The simulation results showed the new procedure ease-to-use, its scalability as well as a satisfactory accuracy/simplicity trade-off.