Thursday, July 14, 2022, 11:00
Room 01-012, Georges-Köhler-Allee 102, Freiburg 79110, Germany
This thesis covers the development and analysis of a model predictive control solution for point-to-point motions of a small scale woodyard crane on a pilot plant from Psiori GmbH in Freiburg. Suitable dynamical models were derived and their parameters identified, focusing on the actuator dynamics and the pendulum movements introduced from the actuator motions. An optimal control problem formulation for the given task was derived; its parameterization led to a numerically solvable nonlinear programming problem. Numerical simulation experiments were conducted to explore the solution set of the problem and possible design choices during the implementation process of the online optimization algorithm. A numerical error analysis was done for the parameters arising from the direct multiple shooting approach which led to insights in how to choose the parameters appropriately.
The ability of the developed solution to control the real-world crane was demonstrated in experiments on a real-world setup. Open-loop control experiments were conducted to evaluate the model mismatch. The results show that the derived model approximates the plant behavior sufficiently well to compute accurate trajectories for horizon of up to two minutes. Additionally, a closed-loop control scheme was implemented into the crane artificial intelligence developed by Psiori. The experiments show that the closed-loop setup effectively compensates for unmodelled disturbances while improving the accuracy with which the goal is reached.
Also online via Zoom:
Meeting ID: 627 9173 7415