Model-based control allows integral design and optimization of a process (mechanism) and its controller. In this case, a dynamics model is available from the process. Otherwise, a dynamic model should be created from analysis of the controlled process. However, models never exactly describe the actual system dynamics, which limits control performance. This can be improved by online updating of the model. Advancements computational hardware and (machine) learning techniques allows updating of ever more complex models. Having accurate models is beneficial for many control strategies and applications; Adaptive estimation of dynamic forces for feedforward compensation in motion control of precision manipulators. Adaptive estimation of the isolator dynamics in disturbance feedforward control for vibration isolation. Model adaptation in model-predictive control for metal forming. Disturbance observers for ensuring passive interaction with the environment. Repetitive disturbance compensation with model inversion-based iterative learning control. Besides improved control, the updated model information can be used for other purposes like condition monitoring, quality inspection and object identification.
