EVENT-BASED STATE ESTIMATION VIA SLIDING MODE OBSERVER

Abstract

The central idea of this thesis is to estimate the states of a linear time-invariant (LTI) plant robustly under various resource constraints, such as communication and computational burdens. It has been recently observed that these constraints have become an integral part of stabilizing controller design in networked control systems (NCSs) or cyber-physical systems (CPSs). In all these cases, the observer plays an important role for not only in controlling the plant but also securing it against malicious attacks. The robust observers can estimate the state exactly (or with some accuracy) in the presence of uncertainties. This thesis proposes a robust sliding mode observer estimation scheme using the event-triggering technique to address the above challenges. In the first part of the thesis, a new sliding mode observer is proposed under some standard assumptions to achieve better steady-state accuracy (in its discrete realization) compared to the existing ones. The observer is designed in a decoupled manner with a linear and a discontinuous parts. The proposed observer design technique employs a discontinuous output error injection vector of same order as that of the disturbance (for rejecting disturbances). As a result of this, the numerical chattering in discrete realization is attenuated, and thus the steady-state accuracy of the estimation process is improved. The proposed observer also enables to design a robust output feedback controller via the separation principle approach. This is achieved by designing the (sliding mode) control law based on the certainty equivalence principle where the observer states are used in the state feedback law when the actual states are not measurable. In the second part of the thesis, the observer is designed by adopting an event-triggering data transmission protocol to estimate the states of an (uncertain) LTI plant in the practical sense. Two scenarios are considered for designing this observer in the event-triggering framework: the actuator side implementation and the sensor side implementation. In the former case, it is assumed that the observer is run at the actuator side, whereas in the second case, the observer is located at the sensor side. In both cases, the event-triggering mechanism is placed in the sensor end to regulate the transmission of output/estimated values. It is shown under standard observability assumption that the event-triggered observer estimates the states practically despite the presence of the disturbance input. The event-triggering implementation of the robust sliding mode observer is also extended for reconstructing the actuator faults, resulting in the malfunctioning of the system components. The observer estimates the states using the sampled output information despite the presence of fault signals. It is shown that by appropriately designing the switching gain and the triggering parameter, the states of the observer converge to a predefined bound in close vicinity of the actual system states. The fault signals can be reconstructed with arbitrary accuracy depending on the steady state error bound of the estimation error and switching component. Overall, the thesis contributes to the area of sliding mode observers. It explored the robust state estimation problem in a sampled-data setup using the event-triggering technique. These contributions not only mark as a significant development in this field but also open the door for future investigations.