Data-driven learning for dynamical systems control and design problems

Project description

In control theory, good modeling is core to the success of traditional control algorithms. In many situations, decisions must be made with incomplete model knowledge, extracting information or ‘learning’ from available data alone. These data-driven approaches have been embraced by the data science and machine learning communities with applications including search engines and computer vision.
This project applies data-driven learning including evolutionary computation algorithms to dynamical systems control and design problems, whilst requiring fundamental guarantees on the convergent behaviour of the process. A strong background and interest in linear algebra and control theory is required and exposure to machine learning and optimization is most beneficial.
There will be the opportunity for exceptional students to undertake a joint PhD on this topic with the University of Birmingham.

Project team

Leader: Chris Manzie

Staff: Airlie Chapman (to be project lead from mid-July 2017), James Bailey

Collaborators: Jon Rowe (University of Birmingham)

Other projects

Networks and data in society projects


Mechanical Engineering


Networks and data in society


autonomous systems; complex systems; control and signal processing; evolutionary algorithms; optimisation