Search

Networked control systems (NCSs) are such control systems where the control loop is closed via some form of communication networks. These control systems are widely applicable in remote and distributed control applications. The inserted network however presents great challenges to conventional control theory as far as the design and analysis of NCSs are concerned. These challenges are caused primarily by the communication constraints in NCSs, e.g., network-induced delay, data packet dropout, data packet disorder, network access constraint, etc., which significantly degrade the system performance or even destabilize the system. When applying conventional control approaches to NCSs, considerable conservativeness is inevitable due to the failure to exploit network characteristics. Therefore, the co-design approach to NCSs in which control approaches and characteristics of NCSs are both fully considered, is believed to be the best way forward for the design of NCSs. In this thesis, we investigate the packet-based transmission of the network being used in NCSs, and propose a packet-based control (PB-control) approach to NCSs. In this approach, the 'packet' structure of data transmission in NCSs which is distinct from conventional control systems, is taken advantage of where, the control signals are first 'packed' and then sent as a sequence instead of one at a time as done in conventional control systems. \Vith the efficient use of the 'packet' structure, we can then actively compensate for the communication constraints in NCSs including the network-induced delay, data packet dropout and data packet disorder simultaneously. After determining the PB-control structure, we then extend its application to several categories of problems as follows. j • The first application is to two types of special nonlinear systems described by a Hammerstein model and a Wiener model respectively. A 'two-step' approach is adopted in this situation to separate the nonlinear process from the whole system which then enables the PB-control approach to be implemented. • It is observed that the communication constraints in NeSs are stochastic in nature, and thus a stochastic analysis of the PB-control approach is presented -----'''-'--'--~-• .:.o'... '-~.::C''c:....'..:..'..;...';';;;'~~.~'----' ......;.''''- ---'- ..-..;.;.~~ / iii under the Markov jump system framework, by modeling the network-induced delay and data packet dropout as a homogeneous ergodic Markov chain. The sufficient and necessary conditions for stochastic stability and stabilization in this situation are also obtained. • Continuous-time plant and continuous network-induced delay are observed to be more difficult to handle when implementing the PB-control approach. For this challenge, a discretization technique is introduced for the continuous network-induced delay and as a result, a novel model for NCSs is derived which is different to that obtained by conventional analysis from time delay system theory. A stabilized controller is also obtained in this situation by using delay-dependent analysis. • The last application is to deal with the situation where a set of NeSs share the network and thus the network access constraint has to be considered. For this situation, a PB-control and scheduling co-design approach is proposed where, PB-control is still applied to each subsystem while scheduling algorithms are applied to schedule the network resources among the subsystems to guarantee the stability of the whole system. We also point out in the thesis that further research on the PB-control approach is still needed as far as nonlinear, continuous-time systems and stochastic analysis are concerned.