Event-triggered control over sensor/actuator wireless networks (NSF 0841216)
The energy obstacle to control over sensor/actuator wireless networks.
Distributed implementations of control systems require the placement of sensors and actuators in different physical locations. The sensors, actuators, and computing units (in charge of computing the control action) need to exchange information and are typically interconnected through a communication network. It is through the communication network that the computing units obtain measurements from the sensors and feed computed control signals to the actuators.
Traditionally, these networks have been wired across the plant to be controlled. With the advent of advanced wireless technologies, wired networks are been replaced by their wireless counterparts due to easier and cheaper deployment and maintenance. Despite its clear advantages over wired networks, the wireless paradigm poses many new interesting challenges. Arguably the most notable of these challenges, is how to best manage the scarce energy reserves of each network node. Since the process of communicating information is, in general, the most expensive, the reduction of communication requirements is paramount to obtain energy efficient control over sensor/actuator networks. The less frequently the control signal needs to be updated, the less often measurements need to be transmitted though the network. Hence, energy expenditures are reduced and network lifetime maximized. Reducing the actuation necessary to stabilize the plant is also a desired feature on the real-time scheduling of controllers on embedded platforms
From time-triggered to event-triggered control.
Control signals are typically updated periodically since this results in simpler design and analysis of feedback controllers. Periodicity also simplifies the scheduling of messages required for control over wired networks. However, aperiodic implementations have been receiving increasing attention as an alternative aimed at reducing the frequency of controller updates. In this project we study one particular aperiodic controller implementation termed event-triggered control.
Event-triggered control is based on the idea that control systems do not require the same level of attention at every state of the plant. A system can be operating on a stable manifold of the state space and therefore require little or no attention, while if it operated on an unstable manifold, it might demand more control attention. Periodic implementations cannot take advantage of this fact because they disregard the information contained in the state: the control signals are updated at the same rate independently of the state. Moreover, the period is usually selected to guarantee a desired performance under worst case conditions even though these might rarely occur. Contrastingly, event-triggered implementations require the use (or update) of actuation only when certain events occur. These events are generated when a certain condition on the state of the plant is violated so as to guaranty stability and desired levels of performance.
Distributed event-triggered control.
Implementing event-triggered controllers over sensor/actuator networks leads to reduced communication. But the detection of these state dependent events poses new challenges on its own due to the distributed nature of the problem. In a sensor/actuator network implementation these events need to be designed so that they can be detected in a completely distributed fashion with a low communication burden. It is this challenge what this project focuses on. For more information on our current progress please see our publications below.
Publications.To learn more about these techniques please see the links below (in reverse chronological order)
Self-Triggered Control: trading actuation for computation
Manuel Mazo Jr., Adolfo Anta and Paulo Tabuada
Submited for publication, June 2009.
Input-to-state stability of self-triggered control systems
M. Mazo Jr. and P. Tabuada
Proceedings of the 48th IEEE Conference on Decision and Control, December 2009 (To appear).
On Self-Triggered Control for Linear Systems: Guarantees and Complexity
Manuel Mazo Jr., Adolfo Anta and Paulo Tabuada
Proceedings of the European Control Conference, August 2009.
Event-triggered and Self-triggered control over sensor/actuator networks
Manuel Mazo Jr. and Paulo Tabuada
Proceedings of the 47th IEEE Conference on Decision and Control, December 2008. PDF file
Preliminary results on state-triggered scheduling of stabilizing control tasks
P. Tabuada and X. Wang
Proceedings of the 45th IEEE Conference on Decision and Control, December 2006.