When Decision Meets Estimation: Theory and Applications

In many practical problems, both decision and estimation are involved. This dissertationintends to study the relationship between decision and estimation in these problems, so thatmore accurate inference methods can be developed.Hybrid estimation is an important formulation that deals with state estimation and modelstructure identification simultaneously. Multiple-model (MM) methods are the most widelyusedtool for hybrid estimation. A novel approach to predict the Internet end-to-end delayusing MM methods is proposed. Based on preliminary analysis of the collected end-to-enddelay data, we propose an off-line model set design procedure using vector quantization (VQ)and short-term time series analysis so that MM methods can be applied to predict on-linemeasurement data. Experimental results show that the proposed MM predictor outperformstwo widely used adaptive filters in terms of prediction accuracy and robustness.Although hybrid estimation can identify model structure, it mainly focuses on the estimationpart. When decision and estimation are of (nearly) equal importance, a joint solutionis preferred. By noticing the resemblance, a new Bayes risk is generalized from those of decisionand estimation, respectively. Based on this generalized Bayes risk, a novel, integratedsolution to decision and estimation is introduced. Our study tries to give a more systematicview on the joint decision and estimation (JDE) problem, which we believe the work in variousfields, such as target tracking, communications, time series modeling, will benefit greatlyfrom. We apply this integrated Bayes solution to joint target tracking and classification, avery important topic in target inference, with simplified measurement models. The resultsof this new approach are compared with two conventional strategies. At last, a surveillance testbed is being built for such purposes as algorithm developmentand performance evaluation. We try to use the testbed to bridge the gap between theoryand practice. In the dissertation, an overview as well as the architecture of the testbed isgiven and one case study is presented. The testbed is capable to serve the tasks with decisionand/or estimation aspects, and is helpful for the development of the JDE algorithms.