From: 2024-10-30 11:00 to 12:00
Place: Seminar Room M 3170-73 in the M-building, LTH
Contact: bjorn [dot] olofsson [at] control [dot] lth [dot] se

Date & Time: October 30th, 11:00 - 12:00
Location: Large conference room 2485-88 at Dept. of Automatic Control, LTH 
Author: Teodor Westholm
Title: Elasto-kinematic modelling and calibration of an industrial manipulator with a coupled wrist gearbox
Supervisor: Björn Olofsson, Dept. Automatic Control, LTH & Andreas Stolt, Cognibotics AB
Examiner: Yiannis Karayiannidis, Dept. Automatic Control, LTH

Abstract: This thesis examines how elastic deformations affect the kinematics and stationary positional and orientational accuracy of an industrial manipulator. The analyzed robot had a wrist design which introduced a coupling between the two wrist joints, i.e., the outermost two joints responsible for orienting the end-effector. This meant that one of the wrist motors had an effect on both joints, and that external loads could cause the joints to deform in unexpected ways, which may lead to kinematic inaccuracies if not accounted for. Friction in the gearbox could also lead to complications when identifying the elasticities, as it creates uncertainty in the torque measurements.Several elasto-kinematic models, which accounted for the coupling in different ways, were designed, and calibration algorithms were developed to identify the model parameters. The results were compared to a simple flexible joint model, which did not include the coupling, to investigate if the inclusion had a positive effect on the results. Calibrations were performed using both torque data from the joint motors, and torques estimated from a gravitational model. A simple model of kinetic friction was also studied and a calibration algorithm was developed to estimate the friction parameters.The calibrated elastic models were able to reduce positional and orientational error by an order of magnitude. The result could, however, not conclusively show whether the inclusion of the coupling term yielded a better model. Furthermore, the friction model suffered from identifiability issues, and was likely too simple to be useful for reducing measured torque uncertainty, though some insights on the distribution of friction on the gearbox were obtained.