Project number: VEGA 1/0403/15
Principal investigator: M. Kvasnica
Keywords: Optimal control, process control, stability, real-time control
This research project is devoted to design, synthesis, and implementation of optimal control systems for process control applications which require rigorous guarantees that the control system will exhibit desired safety and economical properties. The parameters of safety and economical behavior are divided into theoretical properties (closed-loop stability, recursive feasibility and satisfaction of process constraints), and practical properties (guaranteed execution of the optimization algorithm on platforms with restricted computational resources, correct behavior of the control system under quantization and under failures of the communication channels). Nowadays, these properties are verified by extensive testing, which is time consuming and expensive. Therefore the main goal of the project is to develop a unified methodology which allows to design optimal control systems in which safety properties can be imposed and verified already at the design stage.
The main aim of the project is to develop novel methods and algorithms for synthesis of verifiably safe optimal control strategies. Such a synthesis must be tailored to specific hardware platform on which the control system is to be implemented, and must take into account parameters of the input/output channels, as well as properties of communication channels. The objective is to synthesize such optimization-based control strategies which satisfy these limits from the very beginning, abolishing the need to a-posteriori extensive verification. Achieving this objective requires solving three main tasks, which represent the particular goals of this project:
- Obtaining a rigorous certificate that a given optimization-based control strategy provides closed-loop stability and recursive feasibility even in the presence of the model-plant mismatch, and when measurements/control inputs are quantized or are subject to network communication.
- Synthesis of optimal control algorithms which are guaranteed to be executed correctly even on hardware platforms with prescribed constraints on computational power and available memory.
- Implementation and verification of verifiably safe optimal control systems under hard real-time constraints.
The main vision of the project is to provide a unified control design and verification methodology which will abolish the need of costly a-posteriori testing of safety properties. This will be achieved by synthesizing rigorous certificates of closed-loop stability and recursive feasibility in the form of Lyapunov functions and invariant sets. These certificates will be constructed using parametric and mixed-integer optimization. Results of this research project will be experimentally verified on a rich scale of chemical and biotechnological processes, as well as on mobile robotic platforms.