abstract: In today's global market, companies are facing unprecedented levels of uncertainties in supply, demand and in the economic environment. A critical issue for companies to survive increasing competition is to monitor the changing business environment and manage disturbances and changes in real time. In this dissertation, an integrated framework is proposed using simulation and online calibration methods to enable the adaptive management of large-scale complex supply chain systems. The design, implementation and verification of the integrated approach are studied in this dissertation. The research contributions are two-fold. First, this work enriches symbiotic simulation methodology by proposing a framework of simulation and advanced data fusion methods to improve simulation accuracy. Data fusion techniques optimally calibrate the simulation state/parameters by considering errors in both the simulation models and in measurements of the real-world system. Data fusion methods - Kalman Filtering, Extended Kalman Filtering, and Ensemble Kalman Filtering - are examined and discussed under varied conditions of system chaotic levels, data quality and data availability. Second, the proposed framework is developed, validated and demonstrated in `proof-of-concept' case studies on representative supply chain problems. In the case study of a simplified supply chain system, Kalman Filtering is applied to fuse simulation data and emulation data to effectively improve the accuracy of the detection of abnormalities. In the case study of the `beer game' supply chain model, the system's chaotic level is identified as a key factor to influence simulation performance and the choice of data fusion method. Ensemble Kalman Filtering is found more robust than Extended Kalman Filtering in a highly chaotic system. With appropriate tuning, the improvement of simulation accuracy is up to 80% in a chaotic system, and 60% in a stable system. In the last study, the integrated framework is applied to adaptive inventory control of a multi-echelon supply chain with non-stationary demand. It is worth pointing out that the framework proposed in this dissertation is not only useful in supply chain management, but also suitable to model other complex dynamic systems, such as healthcare delivery systems and energy consumption networks. / Dissertation/Thesis / Ph.D. Engineering 2010

abstract: Human operators have difficulty driving cranes quickly, accurately, and safely because of the slow response of heavy crane structures, non-intuitive control interfaces, and payload oscillations. Recently, a novel hand-motion crane control system has been proposed to improve performance by coupling an intuitive control interface with an element that reduces the complex oscillatory behavior of the payload. Hand-motion control allows operators to drive a crane by simply moving a hand-held radio-frequency tag through the desired path. Real-time location sensors are used to track the movements of the tag and the tag position is used in a feedback control loop to drive the crane. An input shaper is added to eliminate dangerous payload oscillations. However, tag position measurements are corrupted by noise. It is important to understand the noise properties so that appropriate filters can be designed to mitigate the effects of noise and improve tracking accuracy. This work discusses implementing filtering techniques to address the issue of noise in the operating environment. Five different filters are used on experimentally-acquired tag trajectories to reduce noise. The filtered trajectories are then used to drive crane simulations. Filter performance is evaluated with respect to the energy usage of the crane trolley, the settling time of the crane payload oscillations, and the safety corridor of the crane trajectory. The effects of filter window lengths on these parameters are also investigated. An adaptive filtering technique, namely the Kalman filter, adapts to the noise characteristics of the workspace to minimize the tag tracking error and performs better than the other filtering techniques examined. / Dissertation/Thesis / M.S. Bioengineering 2012

abstract: It is possible in a properly controlled environment, such as industrial metrology, to make significant headway into the non-industrial constraints on image-based position measurement using the techniques of image registration and achieve repeatable feature measurements on the order of 0.3% of a pixel, or about an order of magnitude improvement on conventional real-world performance. These measurements are then used as inputs for a model optimal, model agnostic, smoothing for calibration of a laser scribe and online tracking of velocimeter using video input. Using appropriate smooth interpolation to increase effective sample density can reduce uncertainty and improve estimates. Use of the proper negative offset of the template function has the result of creating a convolution with higher local curvature than either template of target function which allows improved center-finding. Using the Akaike Information Criterion with a smoothing spline function it is possible to perform a model-optimal smooth on scalar measurements without knowing the underlying model and to determine the function describing the uncertainty in that optimal smooth. An example of empiric derivation of the parameters for a rudimentary Kalman Filter from this is then provided, and tested. Using the techniques of Exploratory Data Analysis and the "Formulize" genetic algorithm tool to convert the spline models into more accessible analytic forms resulted in stable, properly generalized, KF with performance and simplicity that exceeds "textbook" implementations thereof. Validation of the measurement includes that, in analytic case, it led to arbitrary precision in measurement of feature; in reasonable test case using the methods proposed, a reasonable and consistent maximum error of around 0.3% the length of a pixel was achieved and in practice using pixels that were 700nm in size feature position was located to within ± 2 nm. Robust applicability is demonstrated by the measurement of indicator position for a King model 2-32-G-042 rotameter. / Dissertation/Thesis / Measurement Results (part 1) / Measurement Results (part 2) / General Presentation / M.S. Mechanical Engineering 2012

Esta dissertação apresenta novos resultados para a solução de problemas de implementação computacional na estimativa de sistemas singulares e sistemas Markovianos. São apresentados algoritmos alternativos para problemas de filtragem de maneira a minimizar problemas causados principalmente por erros de arredondamento e mal condicionamento de matrizes. O trabalho envolve basicamente algoritmos array e filtragem de informação para a estimativa de sistemas singulares nominais e robustos. Também é deduzido um algoritmo array para a filtragem de sistemas lineares sujeitos a saltos Markovianos. / This dissertation presents new results to solve computational implementation problems to estimate singular and Markovian systems. Alternative algorithms to handle computational filtering errors due rounding errors and ill-conditioned matrices are developed. This dissertation comprehends basically array algorithms and information filters for the estimate of nominal and robust singular systems. Also, it is developed an array algorithm for Markovian jump linear systems filtering.

Orientador: Emanuel Pimentel Barbosa / Dissertação (mestrado) - Universidade Estadual de Campinas, Instituto de Matematica, Estatistica e Computação Cientifica / Made available in DSpace on 2018-08-05T13:04:51Z (GMT). No. of bitstreams: 1 Gomes_DanielTakata_M.pdf: 3107518 bytes, checksum: c063e22a01c9ebf48a081c148b6138d8 (MD5) Previous issue date: 2005 / Resumo: É de grande interesse o estudo de previsão de séries temporais, ou seja, conseguir identificar características do processo num ponto futuro. Para isso, é necessário estimar com precisão, ou pelo menos com uma boa aproximação, o processo gerador dos dados. Nos últimos anos, modelos de redes neurais artificiais vêm desempenhando um papel crescente na abordagem e solução de problemas estatísticos importantes. A estrutura mais básica de rede neural, as redes progressivas ou feedforward, sem retroalimentação, têm se mostrado uma alternativa vantajosa, em alguns casos, em relação aos modelos lineares tradicionais. No entanto, algumas séries apresentam características que permitem introduzir algum tipo de realimentação na rede. Tal rede chamada de rede neural recorrente, ferramenta ainda pouco explorada pela comunidade estatística. São dois os principais objetivos da dissertação. Primeiro, o estudo de redes neurais recorrentes para predição de séries temporais, o que compreende sua fundamentação teórica, principais arquiteturas e algoritmos de aprendizagem e sua implementação computacional. Segundo, estudo comparativo da performance preditiva dessas redes para séries temporais, tanto no caso de séries temporais de memória curta quanto de memória longa, tomando-se como referência modelos padrão tipo ARIMA e ARFIMA. São utilizadas séries espaciais (dados de variáveis do solo) e séries temporais (séries de inflações de países desenvolvidos). Os resultados mostram que o uso das redes para predição é vantajoso em relação a modelos lineares para diversas séries. Também é desenvolvido um novo modelo (redes neurais com pesos variáveis) e seu correspondente algoritmo de estimação baseado nas características dos dados / Abstract: Forecasting of time series is a topic of great interest nowadays. To do so, the data generating process needs to be estimated with a good degree of accuracy. In the last years, artificial neural networks are becoming more important in the statistical community. The more basic structure of a neural network, the feedforward neural nets, without feedback, can be a profitable alternative, in some cases, comparing to linear traditional models. However, some time series present characteristics that allow to introduce some kind of feedback in the network. Such network is called recurrent neural network, a tool not so popular in the statistical community.Two of the main concerns of this work are the study of recurrent neural networks for prediction of time series (theoretical fundamental, main architectures and learning algorithms) and the comparative study of the predictive performance of these networks for short and long memory time series. Spatial series (solo science data) and time series (inflation data) are used. The results show that the networks have good prediction performance comparing to linear models in several series. A new model (neural networks with varying coefficients) and its estimation algorithm are proposed, based in the data characteristics / Mestrado / Mestre em Estatística

Orientadores: Luiz Koodi Hotta, Fernando Antônio da Silva Moura / Dissertação (mestrado) - Universidade Estadual de Campinas, Instituto de Matemática, Estatística e Computação Científica / Made available in DSpace on 2018-08-20T15:43:54Z (GMT). No. of bitstreams: 1 LunaHernandez_Angela_M.pdf: 2230052 bytes, checksum: 7cee5005f5f1bd9bbdd565de481db0ad (MD5) Previous issue date: 2012 / Resumo: A dissertação apresenta o processo de estimação em pesquisas repetidas sob o enfoque de séries temporais, empregando Modelos de Espaço de Estados e o filtro dos mínimos quadrados generalizados ou filtro GLS. Este filtro permite o tratamento de modelos com erros de observação autocorrelacionados de forma mais simples do que utilizando o filtro de Kalman e, além disso, possibilita a modelagem conjunta de várias subpopulações ou domínios sob restrições de benchmark obtidas a partir da mesma pesquisa. Isto não só permite manter a coerência entre as estimativas obtidas pelo método, e estimativas agregadas baseadas no planejamento da amostra, como ajuda na proteção contra possíveis erros de especificação dos modelos. Considerando o caso de amostras com rotação, também é abordado o processo de estimação da estrutura de autocorrelação dos erros amostrais empregando o método dos pseudo-erros. Via simulação, é replicado todo o procedimento de estimação, comparando resultados obtidos empregando os filtros GLS e de Kalman. Adicionalmente, é ilustrada a aplicação do filtro sob restrições de benchmark empregando a série de taxa de desemprego da Pesquisa Mensal de Emprego do IBGE, no período de março de 2002 a fevereiro de 2012 / Abstract: This work presents the estimation process in repeated surveys using State Space Models and the generalized linear squares filter, GLS filter, under the time series approach. This filter deals with autocorrelated errors in the observation equation, in a simpler way than the well-known Kalman filter. Additionally, it allows for modeling jointly several domains under benchmark constraints obtained from the same survey. The benchmarking not only achieves coherence between the model-based estimates and the corresponding design-based aggregated estimates, but also provides protection against possible model failures. For the scenario of samples with a rotation scheme, the estimation of the autocorrelation structure of observational errors, using the pseudo-errors method, is also addressed. Simulation are used to compare the GLS and Kalman filter estimators. Moreover, the application of GLS filter under benchmark restrictions is illustrated, using the unemployment rate time serie from the Brazilian monthly labor force survey, from March 2002 to February 2012 / Mestrado / Estatistica / Mestre em Estatística

Direct measurements of external forces acting on a structure are infeasible in many cases. The Augmented Kalman Filter (AKF) has several attractive features that can be utilized to solve the inverse problem of identifying applied forces, as it requires the dynamic model and the measured responses of structure at only a few locations. But, the AKF intrinsically suffers from numerical instabilities when accelerations, which are the most common response measurements in structural dynamics, are the only measured responses. Although displacement measurements can be used to overcome the instability issue, the absolute displacement measurements are challenging and expensive for full-scale dynamic structures. In this paper, a reliable model-based data fusion approach to reconstruct dynamic forces applied to structures using heterogeneous structural measurements (i.e., strains and accelerations) in combination with AKF is investigated. The way of incorporating multi-sensor measurements in the AKF is formulated. Then the formulation is implemented and validated through numerical examples considering possible uncertainties in numerical modeling and sensor measurement. A planar truss example was chosen to clearly explain the formulation, while the method and formulation are applicable to other structures as well.

Abstract The purpose of this research has been to show with applications that polyhedral scenes can be modeled in real time with a single video camera. Sometimes this can be done very efficiently without any special image processing hardware. The developed vision sensor estimates its three-dimensional position with respect to the environment and models it simultaneously. Estimates become recursively more accurate when objects are approached and observed from different viewpoints. The modeling process starts by extracting interesting tokens, like lines and corners, from the first image. Those features are then tracked in subsequent image frames. Also some previously taught patterns can be used in tracking. A few features in the same image are extracted. By this way the processing can be done at a video frame rate. New features appearing can also be added to the environment structure. Kalman filtering is used in estimation. The parameters in motion estimation are location and orientation and their first derivates. The environment is considered a rigid object in respect to the camera. The environment structure consists of 3-D coordinates of the tracked features. The initial model lacks depth information. The relational depth is obtained by utilizing facts such as closer points move faster on the image plane than more distant ones during translational motion. Additional information is needed to obtain absolute coordinates. Special attention has been paid to modeling uncertainties. Measurements with high uncertainty get less weight when updating the motion and environment model. The rigidity assumption is utilized by using shapes of a thin pencil for initial model structure uncertainties. By observing continuously motion uncertainties, the performance of the modeler can be monitored. In contrast to the usual solution, the estimations are done in separate state vectors, which allows motion and 3-D structure to be estimated asynchronously. In addition to having a more distributed solution, this technique provides an efficient failure detection mechanism. Several trackers can estimate motion simultaneously, and only those with the most confident estimates are allowed to update the common environment model. Tests showed that motion with six degrees of freedom can be estimated in an unknown environment. The 3-D structure of the environment is estimated simultaneously. The achieved accuracies were millimeters at a distance of 1-2 meters, when simple toy-scenes and more demanding industrial pallet scenes were used in tests. This is enough to manipulate objects when the modeler is used to offer visual feedback.

The work described in this thesis contributes to the development of a system to evaluate sailing performance. This work was motivated by the lack of tools available to evaluate sailing performance. The goal of the work presented is to detect and classify the turns of a sailing yacht. Data was collected using a BlackBerry PlayBook affixed to a J/24 sailing yacht. This data was manually annotated with three types of turn: tack, gybe, and mark rounding. This manually annotated data was used to train classification methods. Classification methods tested were multi-layer perceptrons (MLPs) of two sizes in various committees and nearest- neighbour search. Pre-processing algorithms tested were Kalman filtering, categorization using quantiles, and residual normalization. The best solution was found to be an averaged answer committee of small MLPs, with Kalman filtering and residual normalization performed on the input as pre-processing.

<div>Self driving cars becoming more popular nowadays, which transport with it's own intelligence and take appropriate actions at adequate time. Safety is the key factor in driving environment. A simple fail of action can cause many fatalities. Computer Vision has major part in achieving this, it help the autonomous vehicle to perceive the surroundings. Detection is a very popular technique in helping to capture the surrounding for an autonomous car. At the same time tracking also has important role in this by providing dynamic of detected objects. Autonomous cars combine a variety of sensors such as RADAR, LiDAR, sonar, GPS, odometry and inertial measurement units to perceive their surroundings. Driver-assistive technologies like Adaptive Cruise Control, Forward Collision Warning system (FCW) and Collision Mitigation by Breaking (CMbB) ensure safety while driving.</div><div>Perceiving the information from environment include setting up sensors on the car. These sensors will collect the data it sees and this will be further processed for taking actions. The sensor system can be a single sensor or multiple sensor. Different sensors have different strengths and weaknesses which makes the combination of them important for technologies like Autonomous Driving. Each sensor will have a limit of accuracy on it's readings, so multi sensor system can help to overcome this defects. This thesis is an attempt to develop a multi sensor multi object tracking method to perceive the surrounding of the ego vehicle. When the Object detection gives information about the presence of objects in a frame, Object Tracking goes beyond simple observation to more useful action of monitoring objects. The experimental results conducted on KITTI dataset indicate that our proposed state estimation system for Multi Object Tracking works well in various challenging environments.</div>