Research On The Medium Access Constraint Issue In Networked Control Systems

In the last few decades, the rapid developments of computer technology, network tech-nology and intelligent sensing technology make the control systems change from the tradi-tional point to point control architecture to the networked control counterpart. Networkedcontrol systems (NCSs) have some nice advantages such as reduced wire, low cost, easymaintenance and expanding, and high flexibility, and NCSs have found a lot of applicationsin industrial fields. Therefore NCSs are receiving increasing attention and becoming one ofthe hot research topics in the control society.However, NCSs sacrifice the real time performance of the measurement and controlsignal besides of the advantages. Traditional control theory assumes that the transmission ofmeasurement and control signals are real time and unconstrained and thus is no longer suit-able for NCSs. On the other hand, the scheduling policy of the network resource also affectsthe performance of the closed-loop NCSs. Therefore, how to design controller and schedul-ing policy to ensure the stability and enhance the control performance of the closed-loopNCSs is an important research topic in NCSs. Up to now, there have been fruitful researchresults on the study of NCSs. However, most of them are either from the controller designor from scheduling policy aspect, for example, network-induced delay, packet dropout andbandwidth allocation. Different from these problems, medium access constraint probleminvolves both controller design and scheduling policy. Therefore, medium access constraintproblem captures the essential property of NCSs and deserves investigation.With an emphasis on the medium access constraint problem, this thesis studies the fol-lowing problems with regard to the NCSs with deterministic or stochastic medium accesscontrol (MAC) protocols.1. A novel transmission policy in which multiple transmissions are allowed in onesampling period is developed. The problem of controller and scheduling policy co-design for NCSs with deterministic MAC protocols and distributed transmission delay is studiedbased on the new transmission policy. The closed-loop NCS is modeled as a switched sys-tem whose switching signal is related to the scheduling policy. Based on the the averagedwell time method, sufficient conditions are presented for the stability of the closed-loopNCS. A co-design method is presented based on the condition. Then consider the randompacket dropout caused by the transmission errors in the network and study the co-designfor NCSs with deterministic MAC protocols and random packet dropout. The closed-loopNCS is modeled as a switch system with random parameters. Stability analysis results andco-design methods are also presented.2. The problem of H∞for NCSs with deterministic MAC protocols is investigated.A new compensation scheme is developed and the attention is focus on the effects of thecompensation parameter on the H∞filtering performance. The filtering error system ismodeled as a switched system with the communication sequence being the switching sig-nal. Sufficient conditions for the filtering error system to be stable and achieve a prescribedH∞performance and an optimal H∞filter design method are presented. Then the effects ofthe random packet dropout and quantization are also considered. The filtering error systemis then modeled as an uncertain switched system with a stochastic parameter. The relationamong the communication sequence, packet dropout rate, quantization density and the H∞filtering performance are established.3. For NCSs with stochastic MAC protocols, the network resource is represented bythe medium access probabilities of the nodes. The guaranteed cost control for NCSs withstochastic MAC protocols is studied. An independent identical distributed sequence is usedto model the random communication between the actuator node and controller and theclosed-loop NCS is modeled as a stochastic system. The relation between the access prob-abilities of the nodes and quadratic performance is established. Then the H∞filteringproblem for NCSs with stochastic MAC protocols is considered. The relation between theaccess probabilities of the nodes and the H∞filtering performance is established.4. In real applications, the estimated values rather than the exact ones can be obtainedfor the medium access probabilities. The estimated error is bounded. Thus the H∞filter-ing problem for NCSs with uncertain access probabilities and quantization is considered.The filtering error system is modeled as a stochastic system with uncertainties in both themodel and probabilities. An optimal H∞filtering design method is presented. The relation between the uncertainty bound of the probabilities and the H∞filtering performance is es-tablished.5.When the feedback control is carried out by network with stochastic MAC proto-cols such as Ethernet and Internet, the NCSs are affected by random distributed delay andpacket dropout. An improved networked predictive control method is presented to achievebetter performance based on the idea of making more use of the measurements for control.The stability analysis results are given for the closed-loop NCS.The obtained results are verified by simulations of practical examples. Moreover, thenetworked predictive control method is proved to be effective by experiments on a ball andbeam system.