The objective of this Master's thesis was to improve end-effector control for an industrial robotic manipulator operating with flexible tools. A REIS RV15 manipulator including different sensors was at disposal for experiments. Improving the control involves damping the oscillations of the flexible tool, which will increase the accuracy of the TCP position and decrease the task completion time.

A dynamic model was derived for both the manipulator and the flexible tool. A sliding mode controller structure was selected to damp the tool oscillations, and the controller applied the dynamic model in the process. System identification methods were used to adapt the model to the practical tool configuration using sensor measurement. Both offline and online methods were described. Sensor information fusion was also considered to improve model state estimates. A Kalman filter was used to combine the estimates and various sensor measurements including strain of the tool.

The control system was empirically tested to measure its performance regarding TCP accuracy and settling time of the strain response. All parts were tested separately. The relay tuning method estimated the eigenfrequency of the tool to within 8-11 % of the correct value, and the Kalman filter showed noise removal properties. The controller reduced the settling time from 50,85 s to 3,85 s in one case, but did not reach sliding mode during the test. The bandwidth of the manipulator was not adequate, and the control signal had to be limited.