dr. ir. Sofie Haesaert

VENI Grant for the CODEC project

Last week I was awarded with a Veni grant by the Dutch Research Council (NWO). This grant is given to highly promising young scientists. It will allow me to strengthen my research on the interface between system identification, control and formal methods.

More information will follow, but if you are already interested in this project feel free to contact me.

CODEC: Correct-by-design Estimation and Control of partially observable stochastic systems

Intelligent systems learn to function safely via sensor data

The increasing availability of computational power and sensors has led to the digitalization of inter alia cars, aeroplanes, and robots. To enable long-term autonomy, these intelligent systems need to be able to adapt to physical changes safely. This research investigates safe learning from sensor data for these systems.

With the digitalization, engineering systems are evolving into new and anticipated smart systems. Governed by digital controllers, these systems execute more and more complex navigation and regulatory tasks. The increasing complexity of these control systems induces a higher risk of design errors with heavy financial and human consequences. For these safety-critical and intelligent systems, the quest for reliable controller design, fit for the new digital age, has recently led to the development of formal verification and design algorithms. Given a specification and a model of the physical system to be controlled, these design algorithms yield a controller that is correct-by-design and thus provably satisfies the specification.

However, despite the excellent work on correct-by-design control, current methods either require full model knowledge or access to the full state of the system. In many applications, the dynamics are affected by stochastic disturbances, the system state can only be partially measured, and the physical variables and dynamic environment are only partially known. Under these challenging circumstances, the CODEC proposal “COrrect-by-Design Estimation and Control of partially observable stochastic systems” meets the need for reliable control design for real control applications with uncertain models.