To content
Department of Mechanical Engineering
Fundamentals of model predictive control

Grundlagen der modellprädiktiven Regelung

Modern control tasks are often so complex that classical control schemes (such as PID) quickly reach their limits. In particular, the consideration of multiple interacting control inputs and outputs as well as constraints on these values (such as limited valve openings or temperature intervals to be meet) constitute major challenges. Model predictive control (MPC) provides an elegant and effective solution to such problems. In fact, MPC can handle challenging system dynamics - independent of the number of inputs, states, or outputs - by explicitly taking a model of the system into account. Moreover, process constraints are incorporated as constraints of an optimal control proplem (OCP). During runtime, this OCP is recurringly initialized with the current system state and solved on a finite prediction horizon. The first part of the resulting opimal control sequence is then applied to the system and the procedure is repeated at the next sampling instant. One obtains a (nearly) optimally operated closed-loop system.

Organizational info (Summer term 2026)

 LecturesExercises
StartApril 20th 2026April 22th 2026
TimeMondays, 14:15 to 15:45Thurdays, 14:15 to 15:45
RoomMB E23/E24MB E23/E24
Lecturers / TutorsMoritz Schulze DarupJohannes van Randenborgh, Philipp Binfet and Jannik Riemann
MoodleLink to the course
LanguageGerman

Content (according to module descprition)

Die Vorlesung bietet eine anwendungsorientierte Einführung in dieses vielseitig einsetzbare und weit verbreite Regelungsverfahren. Dabei wer­den elementare Kenntnis zur Zustandsregelung aus regelungstechnischen Grundlagenveranstaltungen (wie etwa „Regelungstechnik für MB“) aufgegriffen und schrittweise hin zur MPC ent­wickelt. Ein wich­ti­ger Zwischenschritt ist diesbezüglich die so­ge­nannte linear-quadratische Re­ge­lung (LQR), die sich als MPC ohne Be­rück­sich­ti­gung von Beschränkungen auffassen lässt. Sobald die MPC konzeptionell ver­stan­den wurde, wird die Im­ple­men­tie­rung mit­hil­fe von Matlab er­läu­tert und anhand von Beispielanwendungen aus un­ter­schied­li­chen Domänen erprobt. Da die MPC eine optimierungsbasierte Re­ge­lung darstellt, wer­den an­schlie­ßend grund­le­gen­de Ein­blicke in die (konvexe) Op­ti­mie­rung gegeben. Hier sind Kennt­nisse aus „An­ge­wand­te konvexe Op­ti­mie­rung“ hilfreich, jedoch nicht zwingend er­for­der­lich. Im letzten Drittel der Ver­an­stal­tung wer­den Varianten und Erweiterungen der MPC the­ma­ti­siert. Insbesondere wird er­läu­tert, wie sich MPC ohne die Lö­sung von Optimerungsproblemen zur Lauf­zeit re­a­li­sie­ren lässt (explizite MPC) und wie Störeinflüsse kompensiert wer­den kön­nen (robuste MPC). In der ge­sam­ten Ver­an­stal­tung liegt der Fokus auf linearen Systemdynamiken.

This course offers an application-oriented introduction to this versatile and widely used control method. It builds on basic knowledge of state-space control acquired in introductory control engineering courses (such as “Regelungstechnik für MB”) and gradually develops this knowledge toward MPC. An important intermediate step in this regard is so-called linear-quadratic control (LQR), which can be understood as MPC without consideration of constraints. Once MPC has been conceptually understood, its implementation is explained using MATLAB and tested using example applications from various domains. Since MPC is an optimization-based control method, basic insights into (convex) optimization are then provided. Knowledge of “Angewandte konvexe Optimierung” is helpful here, but not strictly required. In the final third of the session, variants and extensions of MPC will be discussed. In particular, we will explain how MPC can be implemented without solving optimization problems at runtime (explicit MPC) and how disturbances can be compensated for (robust MPC). Throughout the event, the focus is on linear system dynamics.

Literature

J. B. Rawlings, D. Q. Mayne, and M. M. Diehl. Model Predictive Control: Theory, Computation, and Design. Nob Hill Publishing, 2nd Edition, 2017.

B. Kouvaritakis and M. Cannon. Model Predictive Control: Classical, Robust and Stochastic. Springer, 2016.