Optimization and Learning for Control of Networks
The coming years will see a continuously growing increase of the size, scale, and complexity of computer-controlled and technological systems. Many of the corresponding challenges in the field of hybrid, nonlinear, and large-scale systems and control still have to be solved. It is of utmost importance to design controllers that meet the challenging requirements for these systems in terms of computational efficiency, adaptability, autonomy, efficiency, functionality, reliability, and safety: operation and performance over extended periods of time, even in the presence of uncertainty, noise, disturbances, and model errors must be guaranteed. Finally, the evolutions in the field of mixed-integer optimization, increasing computer power, and increasing and pervasive presence of embedded monitoring, communication, and control units, open new opportunities for analysis and control of hybrid, nonlinear, and large-scale systems.
The core research methodology of the team can be characterized as optimization-based, estimation-based, and model-based systems and control methods, complemented with computer science and operation research approaches, with the following fundamental research areas:
- discrete-event systems (in particular, max-plus systems);
- hybrid systems (with a focus on piecewise affine systems and switching max-plus systems);
- model-based predictive control and scheduling;
- multi-level, multi-agent control of large-scale systems (with a focus on hybrid systems);
- adaptive control (with hybrid switching and self-tuning mechanisms).
In addition, the team focuses on societally relevant applications stemming from the theme “smart transportation and smart infrastructures in smart cities”, which involves transportation systems and networks (road, rail, and air), infrastructure networks (water, energy, and logistics), and smart buildings.
For the coming years the team has decided to focus on the following fundamental research goals:
- extending the range (both in terms of the size and the types of systems) of current methods for hybrid, multi-level, and adaptive control,
- providing quantitative and qualitative performance guarantees for wider classes of systems,
- developing efficient adaptive and/or robust control methods for hybrid systems,
- developing efficient and scaleable coordination strategies for multi-level, multi-agent control of large-scale hybrid and nonlinear systems,
- using big data methods to efficiently and effectively deal with the increasing amount of data that can be collected in large-scale and hybrid systems.
On the applications side we see new opportunities in integrated transportation and infrastructure networks (such as multi-carrier energy systems, multi-modal transportation networks, or integrated transportation and energy networks with fuel cell cars).