The mechatronic engineering and manufacturing industry are currently facing several control challenges that simple PID controllers and the like do not address: systems are becoming increasingly complex, must meet constraints, must consider economic goals, and handle valuable preview information effectively. Model Predictive Control (MPC) is the only advanced control approach that can meet all these challenges, thanks to its model-based and optimization-based nature. Still, the optimization-based nature of MPC currently hampers widespread adoption in industrial mechatronic systems: current MPC implementations are expensive in terms of compute and memory resources, computation time is non-deterministic, and therefore MPC algorithms cannot be certified to working with a certain sampling rate, MPC development and implementation is not easy and involves high engineering costs because the right tools are lacking.

The project "Deterministic and Inexpensive Realizations of Advanced Control" (DIRAC) aims at a breakthrough of MPC in the mechatronic / mechanical engineering / manufacturing industry by solving all obstacles through performance that revolves around the three keywords in the title:

• Deterministically: New MPC algorithms and approaches will be developed that can operate reliably at a given sample rate, as well as methods to verify their worst computation times and control performance.
• Cheap:
  • Implementations will be made that approach "full" (= online non-linear optimization with high-fidelity models) MPC and can thus run on low-cost computer hardware with a quantifiable impact on the control performance that is calculated in advance.
  • A modular MPC toolbox will be developed that will allow the development, tuning and validation of advanced control at manageable engineering costs.
• Achievements: We will demonstrate the MPC toolbox and the potential of MPC on industrially relevant demonstrators and validation cases to break the status quo in control practices, stimulate acceptance and inspire Flemish industry.

The overarching, tangible, reusable overall result of this project is a toolbox that simplifies the design of non-linear MPC controllers and brings methodological advancements in solvers, approximations and validation techniques within the reach of control practitioners.