A Geometric Approach for Deriving Multicontact Dynamic Identification Models

Dynamic identification models allow for system parameters such as mass and inertia to be computed by combining knowledge of the system’s configuration, and the forces applied to the system. These methods are often used to calibrate individual dynamic models in robots, and have also been used to estimate human parameters. While commonly used, the Denavit-Hartenberg approach can lead to modelling errors due to limitations in modelling rotations about arbitrary axes. This paper introduces a geometric framework for dynamic identification modelling. Geometric models rotation axes to be written explicitly, and offer a concise representation for common kinematic and dynamic operations. This Geometric approach is compared to the DH form, highlighting the ease in modelling complex rotations and the reduction in error due to the removal of oversimplification. Multiple points of contact can also be modelled easily using the geometric approach, allowing for under-actuated systems to be identified.