Data Driven Thermal Drift Compensation for Industrial Robots

• Applicant: Professor Dr.-Ing. Alexander Verl
• Subject Area: Automation, Mechatronics, Control Systems, Intelligent Technical Systems, Robotics
• Term: since 2025
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 562367049

Industrial robots (IR) are cheaper than conventional machine tools and have a large working area. Due to their serial kinematic chains and low drivetrain stiffness, they suffer in terms of positioning accuracy. To increase this, kinematic calibration or load calibration methods can be applied. However, current scientific literature indicates that thermal changes can lead to model errors of a millimeter or more. To minimize these effects, long warm-up periods or even climate control production lines are required, which is not economical. There is still no software compensation method because thermal processes and IR link expansion drift effects are difficult to formulate mathematically. To overcome this problem, a data-driven method is proposed by mounting many temperature sensors in the robot cell. During a calibration phase, a suitable model (e.g., autoregressive exogenous, ARX) is identified using a laser tracker, and a subset of these sensors is located at sensitive spots. These generic models can be identified without the need to derive complicated physical models. If process or the ambient temperature changes, an adaption algorithm is used to update the model with inline measurements and another set of temperature sensors. Finally, the data-driven models are benchmarked against conventional linear models with long, continuous experiments.

• DFG Programme: Research Grants
• Co-Investigator: Dr.-Ing. Armin Lechler