Model-Based Control over Networks: Architecture and Performance

Control systems that transmit information over a shared digital communication network offer improved efficiency, flexibility, and reliability compared to traditional closed loop feedback systems. The use of a limited bandwidth network for control communication presents new challenges on the design and analysis of network interconnected systems due to quantization, time delays, and the absence of feedback measurements for long intervals of time.

In this dissertation we use the Model-Based Networked Control Systems (MB-NCS) configuration for control of uncertain Networked Control Systems (NCS). We consider plant-model mismatch in NCS since the absence of continuous feedback significantly increases sensitivity of the system to model uncertainties, which are commonly present in most control applications. A nominal model of the system is used in MB-NCS to generate an estimate of the plant state when no measurement information is available. The central idea in this approach is to obtain a comparable system performance with respect to the non-networked closed loop system but limiting the use of network resources. Then the overall performance of NCS has to consider the response of the control and output signals along with the excessive use of network bandwidth.

Different approaches that consider dynamical uncertain systems and allow longer sensor update intervals are discussed in this dissertation. They are primarily based on the implementation of event-based strategies that lead the sensor to send feedback measurements to the controller only when it is necessary to do so. This strategy is also used for decentralized control of networked systems and for computation of optimal sensor update instants and for design of optimal MB-NCS controllers. On-line estimation of the system parameters also permits longer update intervals by upgrading the model of the system that is used to generate the control input between sensor updates. Additionally, lifting techniques are used for stability analysis of MB-NCS. By using these methods we are able to consider model-based multi-rate periodic systems and MB-NCS that are implemented using the network in the sensor-controller and the controller-actuator channels. Finally, control architectures for reference input tracking with limited feedback information are also discussed in this dissertation.