Contributions to robust estimation and distributed control in multi-agent systems

Resumen

Networked Controls Systems (NCSs) are formed by different elements connected and coordinated through a communication network. In last years the number of publications and researches has been extended thanks to the development and reduction of prices of wireless devices and smart sensors. This thesis proposes new solutions for the estimation and control of systems on a distributed framework. The different nodes of the network are called agents, and they have to communicate their information in order to achieve an appropriate estimation. The first step to familiarize with the problem is focused on a centralized paradigm. Let's start with a traditional way to deal with uncertainties, H∞ theory. This approach, that is very extended in the literature, gives way to the set-membership approach, still under a centralized scheme. Once the set-membership methodology has been presented, it is time to focus on the distributed scheme, where the agents will exchange information in order to solve the different problems raised. Firstly, the problem of distributed estimation is addressed. Next step is to add some particularities in the communications, like saving energy consumption or delays on the network. Afterwards, the new step considers a new capability: how to deal in a distributed way with the whole set of agent does not measure all the states of the plant, a Kalman-structure and a consensus-based observer are described. The following steps deal with the joint problem of control and estimation of the systems. Based on the same idea of increasing the difficulty at each step, the first step in the joint problem is a distributed estimation with a centralized control. A star topology is considered on the communication network, the central agent applies control and the others estimate the control signal with the information that they send. After that, a complete distributed estimation and control scheme takes up the next step. More than one agent can apply control to the plant. And they estimates the control signal not only using the information that they send, but also using the distance among them and the one that applies the control signal. And finally, the last step tackles interconnected systems estimation and control. And it takes into account the negotiation problem with the exchange of information on the communication, giving a solution when there are constraint on the amount of information that can be sent over the network.