Reliable Control System Design despite Byzantine Actuators

Sensor-actuator networks are increasingly being used in distributed control of large scale systems. Often these applications are mission-critical and are required to maintain satisfactory performance in the presence of component failures. On the one hand, sensor-actuator network components are becoming inexpensive but they also tend to be unreliable, especially when deployed in harsh or unpredictable environments. The various component failures can manifest themselves in the form of arbitrary actuator behavior in which case their effect on the underlying systems can be severe. In this paper we focus our attention on applications of sensor networks in control of linear systems and show how to deal with Byzantine faults of actuators. We first describe a fault-tolerant control scheme using locally redundant actuators. We then relax the requirement on actuators to be at the same location and design a fault-tolerant scheme where the actuator redundancies are further reduced as well. We demonstrate our methodologies using a beam vibration control application as a case study.