Telecommunication networks become major parts in modern complex control systems recently. They provide many advantages over conventional point-to-point connections, such as the simplification on installation and maintenance with comparatively low cost and the nature requirement of wireless communication in remote control systems. In practice, limited resource networks are shared by multiple controllers, sensors and actuators, and they may need to serve some other information unrelated to control purpose. Consequently, the control system design in networked control systems should be revised by taking communication constraints, for example, finite precision data, time delay and noise in transmission, into account. This thesis studies the robust control and state estimation of uncertain systems, when feedback information is sent via limited capacity communication channels. It focuses on the problem of finite precision data due to the communication constraints. The proposed schemes are based on the robust set-valued state estimation and the optimal control techniques. A state estimation problem of linear uncertain system is studied first. In this problem, we propose an algorithm called coder-decoder for uncertain systems. The coder encodes the observed output into a finite-length codeword and sends it to the decoder that generates the estimated state based on the received codeword. As an illustration, we apply the results in state estimation problem to a precision missile guidance problem using sensor fusion. In this problem, the information obtained from remote sensors is transmitted through limited capacity communication networks to the guided missile. Next, we study a stabilization problem of linear uncertain systems with state feedback. In this problem, the coder-controller scheme is developed to asymptotically stabilize the uncertain systems via limited capacity communication channels. The coder encodes the full state variable into a finite-length codeword and sends it to the controller that drives the system state to the origin. To achieve the asymptotic stability, we use a dynamic quantizer so that quantization noise converges to zero. The results in both state estimation and stabilization problems can handle the problem of finite data rate communication networks in control systems.

Telecommunication networks become major parts in modern complex control systems recently. They provide many advantages over conventional point-to-point connections, such as the simplification on installation and maintenance with comparatively low cost and the nature requirement of wireless communication in remote control systems. In practice, limited resource networks are shared by multiple controllers, sensors and actuators, and they may need to serve some other information unrelated to control purpose. Consequently, the control system design in networked control systems should be revised by taking communication constraints, for example, finite precision data, time delay and noise in transmission, into account. This thesis studies the robust control and state estimation of uncertain systems, when feedback information is sent via limited capacity communication channels. It focuses on the problem of finite precision data due to the communication constraints. The proposed schemes are based on the robust set-valued state estimation and the optimal control techniques. A state estimation problem of linear uncertain system is studied first. In this problem, we propose an algorithm called coder-decoder for uncertain systems. The coder encodes the observed output into a finite-length codeword and sends it to the decoder that generates the estimated state based on the received codeword. As an illustration, we apply the results in state estimation problem to a precision missile guidance problem using sensor fusion. In this problem, the information obtained from remote sensors is transmitted through limited capacity communication networks to the guided missile. Next, we study a stabilization problem of linear uncertain systems with state feedback. In this problem, the coder-controller scheme is developed to asymptotically stabilize the uncertain systems via limited capacity communication channels. The coder encodes the full state variable into a finite-length codeword and sends it to the controller that drives the system state to the origin. To achieve the asymptotic stability, we use a dynamic quantizer so that quantization noise converges to zero. The results in both state estimation and stabilization problems can handle the problem of finite data rate communication networks in control systems.

L'orientation des telecommunications vers les hauts-debits fait de la technique de modulation OFDM l'un des centres d'intérêts privilégies de la recherche actuelle. Cette technique basée sur le principe d'orthogonalité des "filtres" réalisant la modulation ne nécessite pas d'égalisation a proprement parler, mais requiert une estimation de la réponse fréquentielle du canal pour chaque symbole transmis. Les contextes de propagation rencontres aujourd'hui en communications mobiles a hauts debits peuvent s'avérer extrêmement difficiles a estimer précisément. Nous proposons dans cet mémoire de thèse deux méthodes d'estimation de canal très sélectif en temps et en fréquence bases sur le critère du Maximum a Posteriori traitant le signal reçu par blocs. Ces algorithmes reposent sur un modèle de canal obtenu suivant la decomposition orthogonale de la matrice d'auto corrélation du canal selon le théorème de decomposition orthogonale de Karhunen¬Loève. Nous présenterons également les performances de ces nouvelles techniques comparées a celles de méthodes classiques d'estimation de canal ainsi que la robustesse de ces techniques a l'erreur d'estimation des statistiques du canal.

This thesis contains two essays that use state-of-the-art econometric methods to estimate the implicit prices of various housing and vehicle attributes in hedonic price analysis. The additive nonparametric regression proposed by Hastie and Tibshirani (1990) is applied to capture a series of nonlinearities relating prices to their attributes that cannot be captured by conventional parametric approach. Due to its additive structure, the additive nonparametric regression retains an important interpretative feature of the linear model and avoids the drawbacks of a fully nonparametric design such as slow rates of convergence and the "curse of dimensionality." The "benchmark" parametric specification for the hedonic price function is carefully chosen via the estimation of the Box and Cox (1964) and Wooldridge (1992) transformations. The additive nonparametric model provides smaller price prediction errors than the benchmark parametric specification in standard goodness of fit measures. The first study examines the effects on housing prices of the structural and environmental attributes using residential sales data from Portland, Oregon. The overall estimation results verify that most housing attributes that are generally linked to the perception of quality, such as larger total structure square footage and higher elevation, have significant positive implicit prices. Attributes that reduce house quality, such as age of house and distance to environmental amenities, discount the value of a house. Complex price effects of various housing attributes are revealed by the additive nonparametric regression. The second study uses a hedonic price approach to estimate the effects on used car prices of vehicle emission attributes, such as hydrocarbon and carbon monoxide emissions, using data from the Vehicle Inspection Program in Portland, Oregon. The estimation results show that used car value is on average higher for vehicles with lower hydrocarbon and carbon monoxide emissions, ceteris paribus. This empirical finding is consistent with recent reports from the U.S. Environmental Protection Agency, which indicate that used vehicles failing to pass required emission tests face potentially high repair costs and frequent smog-check requirements. More cylinders and larger engine size are highly valued. Higher mileage receives relatively little discount compared to age of vehicle. / Graduation date: 2000

We present an algorithm that uses multiple cues to recover shading and reflectance intrinsic images from a single image. Using both color information and a classifier trained to recognize gray-scale patterns, each image derivative is classified as being caused by shading or a change in the surface's reflectance. Generalized Belief Propagation is then used to propagate information from areas where the correct classification is clear to areas where it is ambiguous. We also show results on real images.

The thesis aims at showing some important methods, theories and applications about non-parametric density estimation via regularization in univariate setting. It gives a brief introduction to non-parametric density estimation, and discuss several well-known methods, for example, histogram and kernel methods. Regularized methods with penalization and shape constraints are the focus of the thesis. Maximum entropy density estimation is introduced and the relationship between taut string and maximum entropy density estimation is explored. Furthermore, the dual and primal theories are discussed and some theoretical proofs corresponding to quasi-concave density estimation are presented. Different the numerical methods of non-parametric density estimation with regularization are classified and compared. Finally, a real data experiment will also be discussed in the last part of the thesis. / Statistics

Urban Search and Rescue (USAR) involves having to enter and explore partially collapsed buildings in search for victims trapped by the collapse. There are many hazards in doing this, because of the possibility of additional collapses, explosions, fires, or flooding of the area being searched. The use of robots for USAR would increase the safety of the operation for the humans involved, and make the operation faster, because the robots could penetrate areas inaccessible to human beings. Teleoperated robots have been deployed in USAR situations to explore confined spaces in the collapsed buildings and send back images of the interior to rescuers. These deployments have resulted in the identification of several problems found during the operation of these robots. This thesis addresses a problem that has been encountered repeatedly in these robots: the determination of the scale of unrecognizable objects in the camera views from the robot. A procedure that would allow the extraction of size using a laser pointer mounted on the robot's camera is described, and an experimental setup and results that verify this procedure have been shown. Finally, ways to extend the procedure have been explored

Flow speed describes general traffic operation conditions on a segment of roadway. It is also used to diagnose special conditions such as congestion and incidents. Accurate speed estimation plays a critical role in traffic management or traveler information systems. Data from loop detectors have been primary sources for traffic information, and single loop are the predominant loop detector type in many places. However, single loop detectors do not produce speed output. Therefore, speed estimation using single loop outputs has been an important issue for decades. This dissertation research presents two methodologies for speed estimation using single loop outputs. Based on findings from past studies and examinations in this research, it is verified that speed estimation is a nonlinear system under various traffic conditions. Thus, a methodology of using Unscented Kalman Filter (UKF) is first proposed for such a system. The UKF is a parametric filtering technique that is suitable for nonlinear problems. Through an Unscented Transformation (UT), the UKF is able to capture the posterior mean and covariance of a Gaussian random variable accurately for a nonlinear system without linearization. This research further shows that speed estimation is a nonlinear non-Gaussian system. However, Kalman filters including the UKF are established based on the Gaussian assumption. Thus, another nonlinear filtering technique for non-Gaussian systems, the Particle Filter (PF), is introduced. By combining the strengths of both the PF and the UKF, the second speed estimation methodology - Unscented Particle Filter (UPF) is proposed for speed estimation. The use of the UPF avoids the limitations of the UKF and the PF. Detector data are collected from multiple freeway locations and the microscopic traffic simulation program CORSIM. The developed methods are applied to the collected data for speed estimation. The results show that both proposed methods have high accuracies of speed estimation. Between the UKF and the UPF, the UPF has better performance but has higher computation cost. The improvement of speed estimation will benefit real-time traffic operations by improving the performance of applications such as travel time estimation using a series of single loops in the network, incident detection, and large truck volume estimation. Therefore, the work enables traffic analysts to use single loop outputs in a more cost-effective way.

Flow speed describes general traffic operation conditions on a segment of roadway. It is also used to diagnose special conditions such as congestion and incidents. Accurate speed estimation plays a critical role in traffic management or traveler information systems. Data from loop detectors have been primary sources for traffic information, and single loop are the predominant loop detector type in many places. However, single loop detectors do not produce speed output. Therefore, speed estimation using single loop outputs has been an important issue for decades. This dissertation research presents two methodologies for speed estimation using single loop outputs. Based on findings from past studies and examinations in this research, it is verified that speed estimation is a nonlinear system under various traffic conditions. Thus, a methodology of using Unscented Kalman Filter (UKF) is first proposed for such a system. The UKF is a parametric filtering technique that is suitable for nonlinear problems. Through an Unscented Transformation (UT), the UKF is able to capture the posterior mean and covariance of a Gaussian random variable accurately for a nonlinear system without linearization. This research further shows that speed estimation is a nonlinear non-Gaussian system. However, Kalman filters including the UKF are established based on the Gaussian assumption. Thus, another nonlinear filtering technique for non-Gaussian systems, the Particle Filter (PF), is introduced. By combining the strengths of both the PF and the UKF, the second speed estimation methodology—Unscented Particle Filter (UPF) is proposed for speed estimation. The use of the UPF avoids the limitations of the UKF and the PF. Detector data are collected from multiple freeway locations and the microscopic traffic simulation program CORSIM. The developed methods are applied to the collected data for speed estimation. The results show that both proposed methods have high accuracies of speed estimation. Between the UKF and the UPF, the UPF has better performance but has higher computation cost. The improvement of speed estimation will benefit real-time traffic operations by improving the performance of applications such as travel time estimation using a series of single loops in the network, incident detection, and large truck volume estimation. Therefore, the work enables traffic analysts to use single loop outputs in a more cost-effective way.

In this thesis, we try to provide a broad econometric analysis of a class of risk measures, distortion risk measures (DRM). With carefully selected functional form, the Value-at-Risk (VaR) and Tail-VaR (TVaR) are special cases of DRMs. Besides, the DRM also admits interpretation in the sense of non-expected utility type of preferences. We first provide a unified statistical framework for the nonparametric estimators of the DRMs in a univariate case. The asymptotic properties of both the DRMs and their sensitivities with respect to the parameters representing risk aversion and/or pessimism are derived. Moreover, the relationships between the VaR and TVaR are also investigated in detail, which, we hope, can shed new lights on the way passing one risk measure to another. Then, the analysis of DRMs are extended to a multi-dimensional framework, where the DRM is computed for a portfolio consisting of many primitive assets. Analogous to the mean-variance frontier analysis, we study the efficient portfolio frontier when both objective and constraint are replaced by the DRMs. We call this the DRM-DRM framework. Under a nonparametric setting, we propose three asymptotic test statistics for evaluating the efficiency of a given portfolio. Finally, we discuss the criteria used for evaluating models used to forecast the VaRs. More precisely, we propose a criterion which takes into account the loss levels beyond the VaRs.