State Estimation And Fusion In Wireless Networked Control Systems

Compared with wired networked control systems, due to mobility of equipment, fexi-bility of network structure and lower installation costs, wireless networked control systemshas found applications in a variety of areas, and it has now been one of the most importantresearch topics in the control society.The using of wireless communication network is two-folded, as it not only strength-ens and expands the merits of NCSs, but also brings about lots of new problems andchallenging to the design and analysis of state estimation and fusion. Communicationand energy constraints, package delays and losses induced by wireless communication net-work are commonly recognized as major causes of deterioration of system performanceand system instability. Though there have been some useful results on these issues, lotsof new challenges problems remain to be unsolved. On the one hand, the combination ofmulti-sensors estimation fusion and WNCSs makes things more complicated. On the otherhand, wireless communication network has the features of scarce radio spectrum resources,time-variant channel fading and energy constraints.These are the problems to be solved in this thesis1. Regarding the problem of packet loss in WNCSs, multi-channel multi-sensor esti-mation fusion and network conditions required for estimation stability is investigated. Thispaper establishes quantitative relations among system structure parameters of each sensorand parameters characterizing packet loss of each communication channel, and estimationperformances that unsolved by previous literatures.Firstly, the packet loss process is modeled by an i.i.d. Bernoulli process. Design theestimation fusion algorithms for two scenarios: general sensor without computing powerand smart sensor equipped with computing power. The estimator of general sensor com-bines centralized structure with the idea of distributed estimation. The estimator of smartsensor adopts Covariance Intersection (CI) algorithm. The two algorithms are both basedon the fusion of local state estimates of each observable subsystem. The sufcient con-ditions for stability of expectation of estimation error covariance matrices are derived by taking into account each observable subsystem structure, and the upper bound is given.These stability conditions are described by simple strict inequalities in terms of the largestsingular values of the open loop matrix of each observable subsystem and packet successprobability of each channel. Thus, the efect of the i.i.d. packet loss and system structureon stability could be easily understood.Secondly, the packet loss process is modeled by a time-homogeneous binary Markovprocess. The estimation fusion algorithms for general sensor and smart sensor are designed。The quantitative relations between system structure parameters of each sensor and Markovtransition probability of each communication channel, and estimation performances areestablished and the upper bound of expectation of estimation error covariance matrices isgiven.2. Time-varying packet delay introduced by using a shared wireless communicationchannel of multi-sensor due to limited communication resource, estimation model, designof algorithm and analysis of performance is studied. Through analysis of the quantitativerelations between number of transmitted sensors with scarce communication resource andpacket delay, a new model is developed that can truly characterize estimation fusion prob-lem with time-varying packet delay introduced by scarce communication resource. Basedon Kalman flter and information flter, efective centralized estimation fusion algorithm isdesigned. It also gives the upper bounds of the expected estimation error covariance andestimation error covariance with one-step delay.3. Motivated by the need of saving energy when using wireless sensors, the designmethod for estimation algorithm and power control strategies that could achieve satisfac-tory performance both on energy saving and on the stability of estimation is presented.Through the analysis of the interactions between energy consumption and packet loss, es-timation algorithm based on fusion of local state estimates on the observable subsystemof each sensor with power control is given. Meanwhile, the proof of the equivalence be-tween this fusion estimate algorithm and that based on system is given. Based on givenalgorithm, the quantitative relation between energy consumption of each sensor and esti-mation performances is established. Minimal power used by each sensor that guarantee the stability of estimation is given by three power controllers.A number of illustrative examples and simulations, which are carried out by the Mat-lab and Mathematica softwares, are provided to show the efectiveness of the proposeddesign methods. Finally, the conclusion and future work are presented.