Contracts for Control System Design

Descripción del proyecto

Desarrollar una teoría contractual para el diseño de sistemas de control modulares

Tradicionalmente, los avances en los sistemas de ingeniería se han basado en una mayor automatización. Sin embargo, las técnicas de control existentes ya no pueden adaptarse al aumento del tamaño y la complejidad de los sistemas de fabricación inteligentes, los vehículos autónomos y los sistemas energéticos, lo que compromete la seguridad y la eficiencia. El equipo del proyecto COCOS, financiado por el CEI, pretende desarrollar una teoría para el diseño de sistemas de control que sea intrínsecamente modular y permita a los componentes del sistema verificar sus requisitos de forma independiente, lo que garantiza al mismo tiempo el comportamiento a nivel de sistema. Los contratos sirven como especificaciones sobre el comportamiento de los componentes y se generarán automáticamente para optimizar el rendimiento del sistema. En el proyecto se introducirá un método escalable para el funcionamiento eficiente, seguro y fiable de sistemas de ingeniería, con aplicaciones prácticas probadas en estudios de casos de fabricación de alta tecnología y vehículos autónomos.

Objetivo

The expansion and performance improvement of our engineered systems have traditionally been achieved by increasing the level of automation, driven by advances in control system design. However, due to the increasing complexity and ever-tighter performance requirements on systems such as smart manufacturing systems, autonomous vehicles, and smart energy systems, we are reaching the limits of what can be achieved by existing control techniques. Namely, existing techniques (i) do not naturally guarantee relevant requirements such as safety, resource-efficiency, or transient performance, and (ii) do not scale with the increasing size and heterogeneity resulting from integrating more and more components in our engineering systems.

I propose a new approach for control system design based on contracts, inspired by ideas from software design in computer science. A contract is a detailed quantitative specification on the dynamic behavior of a component and/or the overall system, capable of expressing relevant requirements. Crucially, contracts enable modularity as the design of a component only needs to guarantee satisfaction of its component contract, i.e. without requiring knowledge of other components in the system. For correctly designed contracts, this then guarantees desired system-level behavior. My initial results show the feasibility of using contracts for control system design.

I will develop a contract theory for control system design that includes techniques for the automatic design of component contracts aimed at optimizing system-level performance, and control design for contract satisfaction. Consequently, this project will present (control) system engineers with a scalable approach to guarantee the efficient, safe, and reliable operation of future engineering systems. The practical relevance of the theory will be tested by case studies in high-tech manufacturing systems and autonomous vehicles, and supported by efficient computational tools.