Controller Design for Complex Systems

For many years, we have been developing and applying methods for controller design and validation of control concepts. This covers aspects of sensor and actor placements as well as the design of experiments, data analysis and state estimation.

Our applied spectrum of methods encloses

  • classical PID control
  • precompensation
  • Iterative learning and adaptive controllers
  • Neural and fuzzy controllers
  • Linear optimal controls
  • robust H control
  • model predictive control
  • Distributed and networked controllers

Distributed Model Predictive Control (dMPC)

As an alternative to a central model predictive control, where a single controller is used for the complete system, hierarchical or distributed communicating MPC approaches are applied.
Within a hierarchical approach the fast control is carried out by local controllers getting their performance trajectory from slower higher level controller. Here, several control levels might exist.

In distributed MPC several local controllers solve the control problem in parallel. In case of non-coordinated controller, i.e. each controllers solves the control problem independently from the others, the performance objective is optimized with respect to the local control problem but it might be not optimal for the overall system. It is also possible that the interaction of the controller destabilizes the complete system. To avoid this, communicating controllers are proposed which exchange information like states, control inputs and thus achieve an optimal control input with respect to the complete system.