Experimental evaluation of identification methods as a contribution to adaptive force control in industrial robotics

Inhalt:
Force control offers new and advanced applications for industrial robots, such as force sensitive assembly, machining or hand-guided path teaching and hand-guided operation. For all these applications the control plant characteristics will regularly change during operation due to variations within a certain force control situation and also due to changes between different force control scenarios. The paper deals with the overall stiffness of the mechanical system of a standard industrial robot – mainly represented by the manipulator arm, the tool and the contact environment – as one of the main factors responsible for a significant change of the control process. Stiffness variations are typically caused by varying stiffness of the contact environment or by tool changes (e.g. a change from a compliant to a rigid tool). In order to avoid instability and with the goal to achieve a desired control behaviour, the application of an adaptive force controller is suggested. As a contribution to adaptive force control approaches, the paper examines the feasibility of three common identification methods: firstly model-based identification, secondly identification of the process parameter overall stiffness and thirdly the identification of a performance index. All these methods have been implemented and tested under real-world conditions on a standard industrial robot.