Most control applications need to operate under constraints regarding system inputs, states and outputs. Model Predictive Control (MPC) is an advanced control method which allows for an easy integration of input, state and output constraints into the control algorithm. However, closed-loop system properties as stability and recursive feasibility cannot be guaranteed if the internal model deviates from the true system. Robust MPC methods address this issue by explicitly considering model uncertainty inside the control algorithm. However, control performance may degrade significantly due to an overly conservative consideration of the model uncertainty.



This thesis presents methods how to incorporate adaptive and anticipative knowledge into robust MPC algorithms. It is shown how different levels of anticipative knowledge increase control performance. The computationally efficient integration of anticipative knowledge into robust tube-based MPC algorithms enables the use in real-world applications.



Many control applications are formulated as a tracking problem whereas most robust MPC schemes only consider the regulation problem. In this thesis, robust tracking MPC algorithms are presented which incorporate changing tracking targets in an effective way. Moreover, adaptive control methods are included into a robust tube-based tracking MPC algorithm. The advantages are highlighted by a simulation example for a self-propelled work machine.



Anticipative knowledge is especially useful for throughput control of self-propelled work machines. Anticipative MPC algorithms for self-propelled work machines are presented which consider constraints on ride comfort and engine load. The effectiveness of the approach is highlighted by simulation and field test data.