Systems Theory and Automatic Control

    Information related to the paper "Implementation of Nonlinear Model Predictive Path-Following Control for an Industrial Robot" submitted to IEEE Transactions on Control Systems Technology.

General setup: a KUKA Light-weight Robot IV is controlled by an external PC via Ethernet. The constrained path-following controller runs on the PC with a sampling period of 1 ms. The PC runs Ubuntu 12.04. and the constrained optimization problem is solved using ACADO toolkit.

Constrained output path following



Drawing of the word 'Hello'. Notice how the NMPC changes the speed of the tip of the marker along the path. The actual speed depends on the geometry of the path in Cartesian space and on the constraints imposed to the applied torques and angular speeds in joint space. The robot stops its motion at the end of the path.

Go to Top

Constrained output path following with large disturbance




A second run of the 'Hello' path (same NMPC controller as in the video above). This time we manually apply large disturbances to the robot. Note that the NMPC momentarily stops the motion of the robot if the disturbances are large enough to make following the path no longer possible. As soon as the large disturbance is removed, the NMPC speeds up and continues following the path.
Compare to a trajectory tracking case with large disturbances.

Go to Top

Speed-assigned constrained path following




Drawing of a three-leaved clover path with a speed reference. Notice how the NMPC changes the speed of the tip of the marker along the path. The actual speed is not always the same as the reference speed. The former depends on the geometry of the path in Cartesian space and on the constraints imposed to the applied torques and angular speeds in joint space. The path can be repeated as many times as desired without the need to stop the motion of the robot.

Go to Top