| Communication constraints on channels connecting different components of a control system are traditionally ignored. As a result of advances in communication, computation and network, many current control systems are distributed, asynchronous and networked. The traditional assumption which ignores communication constraints doesn't apply to those systems. This thesis considers the problem in general and focuses the impact of quantization on state estimation. Another problem, namely, reconstruction of a Gauss-Markov process from its quantized version would also be considered along state estimation from quantized measurement since their mathematical models are similar with each other.;Under a set of assumptions which includes ignoring time delay caused by limited communication, systems of limited communication are formulated as two classes of mathematical models. The state estimation problems for these classes of models are presented. Quantized measurement sequential Monte Carlo (QMSMC) method and Finite state approximation method are proposed as candidates for state estimator. Their properties, respective advantages and disadvantages are discussed. In particular, QMSMC's asymptotical optimality is proven under a few further assumptions regarding the system model and the quantizer.;These methods are applied to several models and their effectiveness are demonstrated through simulation-based comparison with Quantized measurement Kalman filter which is the combination of Kalman filter (or Extended Kalman filter) and treating quantization as an additive noise with certain statistical nature.;Optimization of quantization for Gauss-Markov process is also considered. Specifically, optimal quantizer in least-squares sense for second order Gauss-Markov process is numerically studied with respect to Quantized measurement Kalman filter. |
