FeedNetBack - D05.04 - Design methodologies for event-based control systems

Nicolas Cardoso de Castro (1) , Federica Garin (1) , Carlos Canudas de Wit (1) , Peter Hokayem (2) , John Lygeros (2) , Carlos Vivas (3) , Pablo Millan (3) , Luis Orihuela (3) , Francisco Rubio (4) , Ubaldo Tiberi (5) , Carlo Fischione (5) , Karl Henrik Johansson (6)

1 NECS - Networked Controlled Systems
2 IFA - Institut für Automatik - Automatic Control Laboratory
3 ISA - Departamento de Ingenieria de Sistemas y Automatica [Sevilla]
4 ISA - Department of Automatic Control and Systems Engineering [Sevilla]
5 KTH - KTH Royal Institute of Technology [Stockholm]
6 Department of Electromagnetic Engineering [Stockholm]

This is a Deliverable Report for the FeedNetBack project (www.feednetback.eu). Networked Control Systems (NCS) are systems in which the sensors or/and the actuators communicate with the controller through a network. Energy saving and robustness to unreliable channels are major challenges in networked control, notably in wireless scenarios. Energy efficiency and in particular asynchronous design methodologies are studied in this deliverable. The presence of a channel between the sensors measuring the plant and the controller generating the control inputs implies that the measurements should be quantized. As a preliminary step, the problem of finding a stabilizing policy with quantized measurements and bounded control inputs is considered. It is common to assume that the different nodes of a Network Control System use a periodic synchronized clock, this simplifies the model which may take into account some transmission delays. However, this assumption is strong and energy consuming. Indeed, the periodic sampling time is often chosen to ensure given performance in the worst case scenario, wasting energy when the system is running around its working point. To relax the assumption of synchronized nodes, the rest of the deliverable introduces two asynchronous design methodologies, event-based and self-triggered methodologies. The former consists in limiting the transmissions between the nodes when a given condition holds, or, in other words, when an event occurs. Not only this approach relaxes the assumption of synchronized nodes, but it also limits the transmissions which save energy. In the following, event-based approach is applied to a feedback control case and an estimation case. However, by its nature, event-based approach forces the communicating node to watch for the occurrence of the triggering event. This is relaxed in self-triggered approach where each node decides, at the end of an action (e.g. measuring, transmitting, controlling), when the next action will take place. In between these times, the node usually goes to down mode to save energy. In the last part of this deliverable, this approach is applied to a variable sample rate control and to the case of IEEE 802.15.4 protocol.

Domaines: Automatique / Robotique

Nicolas Cardoso de Castro, Federica Garin, Carlos Canudas de Wit, Peter Hokayem, John Lygeros, et al.. FeedNetBack - D05.04 - Design methodologies for event-based control systems. [Research Report] 2011. ⟨hal-00785741⟩