Modeling, Identification, and Control of a Mechanically Coupled 2-DoF Torque Controlled Lightweight Joint

Kurzfassung

Manipulators for safe human-robot interaction should have a lightweight structure and compliant behaviour while being powerful enough to perform complex tasks. The DLR LWR-III has a load-to-weight ratio of about 1:2, which is quite limited for industrial assembly tasks. The SARA:IV concept increases the load-to-weight ratio significantly by using a coupled base joint. However, this induces strong elastic coupling effects. Instead of distinct joint torque sensors, the coupled joint is equipped with a 6-DoF base force/torque sensor. Both, the coupled elasticity and the measured base wrench have to be considered in the dynamics model to allow joint torque estimation and the design of appropriate control structures. Standard identification methods and control structures do not cover the entire elastic coupling effects. In this thesis it is approached in complex identification processes with model-based non-linear optimizations and a decoupling MIMO full state-feedback controller is used to provide appropriate decoupling performance. To include the base force/torque sensing in the model, a 6-DoF floating base is assumed. Various approaches for joint torque estimation are implemented. The MIMO controller significantly increases control performance for coupled oscillations compared to a SISO controller. Additionally, an admittance controller based on the measured base wrench is presented and tested in simulation. With the developed methods, full operation of the mechanical prototype has been achieved. The gathered identification data will be fundamental for the further prototype development.