MSc Thesis Proposal

Cooperating-controllers model predictive control

Mentors: Rudy Negenborn, Bart De Schutter

Keywords: cooperating controllers, model predictive control, distributed control, multi-agent systems

In a conventional control setting, there is a single controller that controls the system. This controller collects information from measurements of the system to determine which inputs to select. However, many real world systems, like for example traffic, water, and power networks are large in scale, and therefore hard to control by a single controller, since this controller would have to gather information from all sensors, and process this directly to provide inputs to all actuators. This is not only hard due to technical issues like communication delays and computational requirements, but also due to practical issues like unavailability of information from one part of the network to another and restricted control access. For this type of systems a control approach with multiple controllers has to be employed.

In such a multi-controller setting several controllers, each with only limited information gathering and processing skills and moreover limited action capabilities, control the subsystems (e.g., subnetworks) of which the overall network is composed. Since the subsystems they control are part of an overall network, inputs selected by one controller influence inputs selected by other controllers. The challenge for this type of control is therefore to make the controllers cooperate, that is, work together, such that the overall network performance is as desired. In our case, we investigate how such a scheme works when the controllers employ model predictive control (see Figure 1).

Multiple controllers
controlling a network.

Figure 1: Illustration of multi-controller model predictive control. Control agents control parts of the overall network. Each of the controllers has a model of the subnetwork it controls. Controllers communicate with neighboring controllers. Through an optimization procedure they decide which inputs to implement on their subsystems.

In this project you consider a control setting in which a power or traffic network is divided into subnetworks, each controlled by a controller using model predictive control. You look at how the network can be split up into smaller parts and how the controllers have to cooperate with one another to obtain good overall system behavior (for example by exchanging predictions about what each controller expects to do in the future).

If you are interested in selecting this project as your MSc project, please come along or send us an email for more information.

This page is maintained by Bart De Schutter.