Simulation-based Parameter Identification Framework for the Calibration of Rigid Body Simulation Models

Abstract

Typically, the parameter identification process for robotic systems requires a symbolic mathematical description of the inverse dynamics equation. The manual derivation of the inverse dynamics often is very time consuming and error-prone. Fortunately, modern simulation systems provide high-level interfaces for the calculation of the inverse dynamics, constituting user-friendly access to the inverse dynamics. The key idea of our research is to directly use the abstract interface of a simulation system for the parameter identification process to foster a flexible, comprehensive, application-independent parameter identification process. Applying this novel approach, the complex derivation of the inverse dynamics equation is superfluous. Instead, the inverse dynamics is described by a CAD-based simulation model and is computed by a unifying simulation algorithm. In this paper, we present the design and realization of our innovative simulation-based parameter identification framework and demonstrate the capacity of the framework by identifying the rigid body properties of the KUKA LWR4 robot.