Static, cross-workspace online calibration for articulated robots with inertial sensors (RCAL-IMS)

• Applicant: Professor Dr.-Ing. Alexander Verl
• Subject Area: Automation, Mechatronics, Control Systems, Intelligent Technical Systems, Robotics
• Term: since 2022
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 497511198

The use of jointed-arm robots in production is steadily increasing. For tasks with high absolute accuracy, however, serial kinematics is a challenge: Elastic deformations in joints and links lead to absolute errors in the range of a few millimeters that vary over the workspace. Previous mechanical calibration methods, for example by probing dimensional scales, are location-bound and limit the calibrated working space that can be used for the application to a partial area. In addition, they allow calibration only in non-productive time, and the actual task must be interrupted for this purpose. Optical methods, such as laser trackers and camera systems, offer high workspace coverage, but are significantly more expensive to purchase than the robot itself and require line-of-sight contact.To compensate for these disadvantages, this proposal investigates a novel calibration method that uses inertial sensors attached to the robot structure that can be used throughout the workspace without an external reference. The inertial sensors are not used at the position level, which has previously limited the achievable accuracy to the millimeter range due to integration drift. Instead, the tilt of the gravitational acceleration vector is measured during static operating phases. This is used to calibrate the parameters of an elastokinematic robot model to improve absolute accuracy and reduce trajectory errors via model-based inverse kinematics. In addition, non-modeled residual errors are measured during operation and iteratively compensated to further increase absolute accuracy. The maximum achievable accuracy increase at the end effector is validated simulatively and experimentally for both parameter identification and online compensation of residual errors.

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