Adaptive Approach To Networked Control Systems

In recent years, control over networks, also called networked control systems (NCSs), has become a very challenging and promising research field. An NCS can be defined as a feedback control system wherein the control loops are closed through a real-time network. One common problem to be addressed when considering NCSs is whether there is sufficient communication bandwidth to feed back information to the controller and then send the control commands to the actuators and the plant. Another one is that there may be delays in transferring the information and the control commands to the controller and the actuator, respectively. Also, some packets not only suffer from transmission delays but may be lost during transmission for the worse case. These constraints make the analysis of NCSs more complex than traditional control systems and may cause destabilization. Therefore, the network-free state feedback adaptive control theories are re-evaluated in this thesis to be applied to NCSs.This thesis addresses the adaptive control problem of linear time-invariant NCSs. The objective is to design state feedback adaptive stabilizing controllers for linear time-invariant NCSs tolerant to actuator failures in the presence of network-induced delay and data-packet dropout. Specifically speaking, the main task is to find upper bounds on the sampling period, the network-induced delay, and the data-packet dropout that guarantee stability of the overall system in the presence of actuator failures. The resulting upper bounds on sampling period, network-induced delay, and data-packet dropout are time varying and can be estimated online. The analysis in the thesis is in the continuous-time domain in which no discretization of the system model is needed for controller design.The adaptive stabilization problem of NCSs is firstly studied without delays or packet dropout to obtain a sufficient condition on the transmission period that guarantees the stability of the adaptive NCS. Next, the problem is reformulated in the presence of network-induced delays. Two sources of network-induced delays are considered in our thesis: the sensor-to-controller delay, and the controller-to-actuator delay. A delay-dependent stabilization approach is proposed to provide an upper bound of the delay such that the closed-loop adaptive NCS is stable for any delay less than the upper bound. As the adaptive controller stabilizer is time-varying, the two sources of delays cannot be lumped together in all situations. This makes the analysis in our thesis more difficult than that used in the literature for completely known systems.Since data-packet dropout might be potential source to instability and poor performance of NCSs, the adaptive stabilization of NCSs in the presence of data-packet dropout is developed. Again, as the adaptive controller stabilizer is time-varying, and assuming that the data packets may lost in sensor-controller network and controller-actuator network, the model of data-packet dropout proposed here is more complex than that used in the literature.The combination of delays and data-packet dropouts is an interesting research point In this thesis, a combination of delay and data-packet dropout is investigated, in which the two sources of complexity; delay modelling and data-packet dropout modelling; resulting from the time-varying property of the adaptive stabilizer are exist in this case.Finally, a state-feedback adaptive stabilizer for NCSs with unknown actuator failures that some unknown inputs are stuck at some unknown fixed values at unknown time instants is designed. The obstacles of NCSs such as; transmission delays and data-packets dropout induced by the insertion of data networks in the feedback adaptive control loops are also considered. A sufficient condition for the existence of adaptive actuator failure compensation is derived.The mathematical proofs of the proposed controllers rely heavily on Lyapunov-stability criterion. It is verified that the adaptive control parameters of the adaptive controller are bounded and the origin of the NCSs is asymptotically stable. Simulation results are given to illustrate the effectiveness of our design stabilizers.