This thesis presents a framework for a mobile robot to escort a human to their destination successfully and efficiently. The proposed technique uses accurate path prediction incorporating human intention to locate the robot in front of the human while walking. Human intention is inferred by the head pose, an effective past-proven implicit indicator of intention, and fused with conventional physics-based motion prediction. The human trajectory is estimated and predicted using a particle filter because of the human's nonlinear and non-Gaussian behavior, and the robot control action is determined from the predicted human pose allowing for anticipative autonomous escorting. Experimental analysis shows that the incorporation of the proposed human intention model reduces human position prediction error by approximately 35% when turning. Furthermore, experimental validation with an omnidirectional mobile robotic platform shows escorting up to 50% more accurate compared to the conventional techniques, while achieving 97% success rate. / Master of Science / This thesis presents a method for a mobile robot to escort a human to their destination successfully and efficiently. The proposed technique uses human intention to predict the walk path allowing the robot to be in front of the human while walking. Human intention is inferred by the head direction, an effective past-proven indicator of intention, and is combined with conventional motion prediction. The robot motion is then determined from the predicted human position allowing for anticipative autonomous escorting. Experimental analysis shows that the incorporation of the proposed human intention reduces human position prediction error by approximately 35% when turning. Furthermore, experimental validation with an mobile robotic platform shows escorting up to 50% more accurate compared to the conventional techniques, while achieving 97% success rate. The unique escorting interaction method proposed has applications such as touch-less shopping cart robots, exercise companions, collaborative rescue robots, and sanitary transportation for hospitals.

Traditional methods of analyzing functional Magnetic Resonance Images use a linear combination of just a few static regressors. This work demonstrates an alternative approach using a physiologically inspired nonlinear model. By using a particle filter to optimize the model parameters, the computation time is kept below a minute per voxel without requiring a linearization of the noise in the state variables. The activation results show regions similar to those found in Statistical Parametric Mapping; however, there are some notable regions not detected by that technique. Though the parameters selected by the particle filter based approach are more than sufficient to predict the Blood-Oxygen-Level-Dependent signal response, more model constraints are needed to uniquely identify a single set of parameters. This illposed nature explains the large discrepancies found in other research that attempted to characterize the model parameters. For this reason the final distribution of parameters is more medically relevant than a single estimate. Because the output of the particle filter is a full posterior probability, the reliance on the mean to estimate parameters is unnecessary. This work presents not just a viable alternative to the traditional method of detecting activation, but an extensible technique of estimating the joint probability distribution function of the Blood-Oxygen-Level-Dependent Signal parameters. / Master of Science

The New Macroeconometrics may succinctly be described as the application of Bayesian analysis to the class of macroeconomic models called Dynamic Stochastic General Equilibrium (DSGE) models. A prominent local example from this research area is the development and estimation of the RAMSES model, the main macroeconomic model in use at Sveriges Riksbank.   Bayesian estimation of DSGE models is often computationally demanding. In this thesis fast algorithms for Bayesian inference are developed and tested in the context of the state space model framework implied by DSGE models. The algorithms discussed in the thesis deal with evaluation of the DSGE model likelihood function and sampling from the posterior distribution. Block Kalman filter algorithms are suggested for likelihood evaluation in large linearised DSGE models. Parallel particle filter algorithms are presented for likelihood evaluation in nonlinearly approximated DSGE models. Prefetching random walk Metropolis algorithms and adaptive hybrid sampling algorithms are suggested for posterior sampling. The generality of the algorithms, however, suggest that they should be of interest also outside the realm of macroeconometrics.

Terrain referenced navigation is a method of navigation that uses measurements of altitude above ground to infer the position of the vehicle, mainly aerial or underwater. This method provides an alternative to the commonly used satellite-based navigation. Satellite-based navigation methods rely on positional information being sent from an external source, which can be jammed or tampered with, a problem terrain referenced navigation does not have. Both satellite-based and terrain based navigation methods often work in conjunction with inertial navigation systems, which are accurate for short periods of time but suffer from large errors due to accumulation of errors when used for longer missions. In this thesis, several state-of-the-art methods of terrain referenced navigation are studied and evaluated, with the main focus being the different estimation methods employed. Five of the studied estimators were implemented and tested on simulated flight data from a generic aerial vehicle, resulting in improved navigation accuracy compared to using inertial navigation on its own. For the terrain referenced navigation to work well, the ground needs to be relatively unique in order to contain useful information, thus making the estimation more uncertain when flying over flat regions. To deal with this, path planning was used to alter the flight path to increase the expected information gain. Using a grid based planning algorithm together with the original route yielded a modified path with more potential information. When following this new path, the terrain referenced navigation systems are shown to estimate the position more accurately compared to the original path. The study shows that terrain referenced navigation is a viable alternative to satellite based navigation. It also indicates that modifying the path to increase the expected information gain can result in better robustness and precision. / Terrängbaserad navigering är en navigeringsmetod där mätningar av höjd över mark används för att fastställa fordonets position, huvudsakligen från luften eller under vattnet. Denna metod är ett alternativ till den allmänt använda satellitbaserade navigeringen. Satellitbaserad navigering är beroende av att positionsinformation skickas från en extern källa, som kan störas eller manipuleras, vilket är ett problem som terrängbaserad navigering inte har. Både satellitbaserade och terrängbaserade navigeringsmetoder används ofta tillsammans med tröghetsnavigeringssystem, som är noggranna under kortare tid, men som lider av stora fel på grund av ackumulering av fel när de används under längre uppdrag. I denna rapport studeras och utvärderas flera moderna metoder för terrängbaserad navigering, med huvudfokus på de olika skattningsmetoder som används. Fem av de studerade skattningsmetoderna implementerades och testades på simulerade flygdata från ett generiskt flygfarkost, vilket resulterade i förbättrad navigeringsnoggrannhet jämfört med att använda tröghetsnavigering på egen hand. För att den terrängbaserade navigeringen ska fungera bra måste marken vara relativt unik för att innehålla användbar information, vilket gör uppskattningen mer osäker när man flyger över plana områden. För att hantera detta användes banplanering för att välja en flygbana som ger maximalt informations innehåll. Genom att använda en rutnätsbaserad planeringsalgoritm tillsammans med den ursprungliga rutten erhölls en modifierad bana med mer potentiell information. Genom att följa denna nya bana uppskattas positionen bättre av de terrängbaserade navigationssystemen jämfört med den ursprungliga banan. Studien visar att terrängbaserad navigering är ett gångbart alternativ till satellitbaserad navigering. Den visar också att en ändring av banan för att öka den förväntade informationsvinsten kan leda till bättre robusthet och precision.

Detecting and tracking moving objects are important topics in computer vision research. Classical methods perform well in applications of steady cameras. However, these techniques are not suitable for the applications of moving cameras because the unconstrained nature of realistic environments and sudden camera movement makes cues to object positions rather fickle. A major difficulty is that every pixel moves and new background keeps showing up when a handheld or car-mounted camera moves. In this dissertation, a novel estimation method of camera motion parameters will be discussed first. Based on the estimated camera motion parameters, two detection algorithms are developed using Bayes' rule and belief propagation. Next, an MCMC-based feature-guided particle filtering method is presented to track detected moving objects. In addition, two detection algorithms without using camera motion parameters will be further discussed. These two approaches require no pre-defined class or model to be trained in advance. The experiment results will demonstrate robust detecting and tracking performance in object sizes and positions.

Mark-recapture methods use repeated captures of individually identifiable animals to provide estimates of properties of populations. Different models allow estimates to be obtained for population size and rates of processes governing population dynamics. State-space models consist of two linked processes evolving simultaneously over time. The state process models the evolution of the true, but unknown, states of the population. The observation process relates observations on the population to these true states. Mark-recapture models specified within a state-space framework allow population dynamics models to be embedded in inference ensuring that estimated changes in the population are consistent with assumptions regarding the biology of the modelled population. This overcomes a limitation of current mark-recapture methods. Two alternative approaches are considered. The "conditional" approach conditions on known numbers of animals possessing capture history patterns including capture in the current time period. An animal's capture history determines its state; consequently, capture parameters appear in the state process rather than the observation process. There is no observation error in the model. Uncertainty occurs only through the numbers of animals not captured in the current time period. An "unconditional" approach is considered in which the capture histories are regarded as observations. Consequently, capture histories do not influence an animal's state and capture probability parameters appear in the observation process. Capture histories are considered a random realization of the stochastic observation process. This is more consistent with traditional mark-recapture methods. Development and implementation of particle filtering techniques for fitting these models under each approach are discussed. Simulation studies show reasonable performance for the unconditional approach and highlight problems with the conditional approach. Strengths and limitations of each approach are outlined, with reference to Soay sheep data analysis, and suggestions are presented for future analyses.

To develop and maintain a modern combat aircraft it is important to have simple, yet accurate, threat models to support early stages of functional development. Therefore this thesis develops and evaluates a model of an active radio frequency (RF) seeker for a missile with data-link capability. The highly parametrized MATLAB-model consists of a pulse level radar model, a tracker using either interacting multiple models (IMM) or particle filters, and a guidance law. Monte Carlo simulations with the missile model indicate that, under the given conditions, the missile performs well (hit rate>99%) with both filter types, and the model is relatively insensitive to lost data-link transmissions. It is therefore under normal conditions not worthwhile to use the more computer intense particle filter today, however when the data-link degrades the particle filter performs considerably better than the IMM filter. Analysis also indicate that the measurements generated by the radar model are neither independent, white nor Gaussian. This contradicts the assumptions made in this, and many other radar applications. However, the performance of the model suggests that the assumptions are acceptable approximations of actual conditions, but further studies within this are recommended to verify this.

Statistical Natural Language Processing (NLP) lies at the intersection of Computational Linguistics and Machine Learning. As linguistic models incorporate more subtle nuances of language and its structure, standard inference techniques can fall behind. One such application is research on the unsupervised induction of part-of-speech tags. It has the potential to improve both our understanding of the plausibility of theories of first language acquisition, and Natural Language Processing applications such as Speech Recognition and Machine Translation. Sequential Monte Carlo (SMC) approaches, i.e. particle filters, are well suited to approximating such models. This thesis seeks to determine whether one application of SMC methods, particle Gibbs sampling, is capable of performing inference in otherwise intractable NLP applications. Specifically, this research analyses the benefits and drawbacks to relying on particle Gibbs to perform unsupervised part-of-speech induction without the flawed one-tag-per-type assumption of similar approaches. Additionally, this thesis explores the affects of type-based supervision with tag-dictionaries extracted from annotated corpora or from the wiktionary. The semi-supervised tag dictionary improves the performance of the local Gibbs PYP-HMM sampler enough to nearly match the performance of the particle Gibbs type-sampler. Finally, this thesis also extends the Pitman-Yor HMM tagger of Blunsom and Cohn (2011) to include an explicit model of the lexicon which encodes those tags from which a word-type may be generated. This has the effect of both biasing the model to produce fewer tags per type and modelling the tendency for open class words to be ambiguous between only a subset of the available tags. Furthermore, I extend the type based particle Gibbs inference algorithm to simultaneously resample the ambiguity class as well as tags for all of the tokens of a given word type. The result is a principled probabilistic model of part-of-speech induction that achieves state-of-the-art performance. Overall, the experiments and contributions of this thesis demonstrate the applicability of the particle Gibbs sampler and particle methods in general to otherwise intractable problems in NLP.

Particle filtering has been a very popular method to solve nonlinear/non-Gaussian state estimation problems for more than twenty years. Particle filters (PFs) have found lots of applications in areas that include nonlinear filtering of noisy signals and data, especially in target tracking. However, implementation of high dimensional PFs in real-time for large-scale problems is a very challenging computational task. Parallel & distributed (P&D) computing is a promising way to deal with the computational challenges of PF methods. The main goal of this dissertation is to develop, implement and evaluate computationally efficient PF algorithms for target tracking, and thereby bring them closer to practical applications. To reach this goal, a number of parallel PF algorithms is designed and implemented using different parallel hardware architectures such as Computer Cluster, Graphics Processing Unit (GPU), and Field-Programmable Gate Array (FPGA). Proposed is an improved PF implementation for computer cluster - the Particle Transfer Algorithm (PTA), which takes advantage of the cluster architecture and outperforms significantly existing algorithms. Also, a novel GPU PF algorithm implementation is designed which is highly efficient for GPU architectures. The proposed algorithm implementations on different parallel computing environments are applied and tested for target tracking problems, such as space object tracking, ground multitarget tracking using image sensor, UAV-multisensor tracking. Comprehensive performance evaluation and comparison of the algorithms for both tracking and computational capabilities is performed. It is demonstrated by the obtained simulation results that the proposed implementations help greatly overcome the computational issues of particle filtering for realistic practical problems.

Les informations contenues dans les cartes routières numériques revêtent une importance grandissante dans le domaine des véhicules intelligents. La prise en compte d’environnements de plus en plus complexes a augmenté le niveau de précision exigé des informations cartographiques. Les cartes routières numériques, considérées ici comme des bases de données géographiques, contiennent des informations contextuelles sur le réseau routier, facilitant la compréhension correcte de l’environnement. En les combinant avec les données provenant des capteurs embarqués, une représentation plus fine de l’environnement peut être obtenue, améliorant grandement la compréhension de contexte du véhicule et la prise de décision. La performance des différents capteurs peut varier grandement en fonction du lieu considéré, ceci étant principalement dû à des facteurs environnementaux. Au contraire, une carte peut fournir ses informations de manière fiable, sans être affectée par ces éléments extérieurs, mais pour cela, elle doit reposer sur un autre élément essentiel : une source de localisation. Le secteur automobile utilise les systèmes de localisation globale par satellite (GNSS) à des fins de localisation absolue, mais cette solution n’est pas parfaite, étant soumise à différentes sources d’erreur. Ces erreurs sont elles aussi dépendantes de l’environnent d’évolution du véhicule (par exemple, des multi-trajets causés par des bâtiments). Nous sommes donc en présence de deux systèmes centraux, dont les performances sont d´dépendantes du lieu considéré. Cette étude se focalise sur leur dénominateur commun : la carte routière numérique, et son utilisation en tant qu’outil d’évaluation de leur performance. L’idée développée durant cette thèse est d’utiliser la carte en tant que canevas d’apprentissage, pour stocker des informations géoréférencées sur la performance des diésèrent capteurs équipant le véhicule, au cours de trajets répétitifs. Pour cela, une localisation robuste, relative à la carte, est nécessaire au travers d’une méthode de map-matching. La problématique principale réside dans la différence de précision entre la carte et le positionnement GNSS, créant des situations ambigües. Durant cette thèse, un algorithme de map-matching a été conçu pour gérer ces ambigüités en fournissant des hypothèses multiples lorsque nécessaire. L’objectif est d’assurer l’intégrité de l’algorithme en retournant un ensemble d’hypothèses contenant l’hypothèse correcte avec une grande probabilité. Cet algorithme utilise les capteurs proprioceptifs dans une approche de navigation à l’estime aidée d’informations cartographiques. Une procédure d’évaluation de cohérence, utilisant le GNSS comme information redondante de positionnement est ensuite appliquée, visant à isoler une hypothèse cohérente unique qui pourra ainsi être utilisée avec confiance dans le processus d’écriture dans la carte. L’utilisation de la carte numérique en écriture/lecture a été évaluée et la procédure complète d’écriture a été testée sur des données réelles, enregistrées par des véhicules expérimentaux sur route ouverte. / The information stored in digital road maps has become very important for intelligent vehicles. As intelligent vehicles address more complex environments, the accuracy requirements for this information have increased. Regarded as a geographic database, digital road maps contain contextual information about the road network, crucial for a good understanding of the environment. When combined with data acquired from on-board sensors, a better representation of the environment can be made, improving the vehicle’s situation understanding. Sensors performance can vary drastically depending on the location of the vehicle, mainly due to environmental factors. Comparatively, a map can provide prior information more reliably but to do so, it depends on another essential component: a localization system. Global Navigation Satellite Systems (GNSS) are commonly used in automotive to provide an absolute positioning of the vehicle, but its accuracy is not perfect: GNSS are prone to errors, also depending greatly on the environment (e.g., multipaths). Perception and localization systems are two important components of an intelligent vehicle whose performances vary in function of the vehicle location. This research focuses on their common denominator, the digital road map, and its use as a tool to assess their performance. The idea developed during this thesis is to use the map as a learning canvas, to store georeferenced information about the performance of the sensors during repetitive travels. This requires a robust localization with respect to the map to be available, through a process of map-matching. The main problematic is the discrepancy between the accuracy of the map and of the GNSS, creating ambiguous situations. This thesis develops a map-matching algorithm designed to cope with these ambiguities by providing multiple hypotheses when necessary. The objective is to ensure the integrity of the result by returning a hypothesis set containing the correct matching with high probability. The method relies on proprioceptive sensors via a dead-reckoning approach aided by the map. A coherence checking procedure using GNSS redundant information is then applied to isolate a single map-matching result that can be used to write learning data with confidence in the map. The possibility to handle the digital map in read/write operation has been assessed and the whole writing procedure has been tested on data recorded by test vehicles on open roads.