Search

The Bayesian model selection and parameter estimation framework developed by Khalil et al. (J Sound Vib 332(15):3670–3691, 2013, Bayesian inference for complex and large-scale engineering systems. Ph.D. thesis, Carleton University, 2013) and Sandhu et al. (J Comput Methods Appl Mech Eng 282:161–183, 2014, Bayesian model selection and parameter estimation for a nonlinear fluid–structure interaction problem. Master’s thesis, Carleton University, 2012, 54th AIAA/ASME/ASCE/AHS/ASC structures, structural dynamics, and materials conference, 2013) is extended to handle strongly nonlinear systems. The evidence required to estimate the posterior probability of each proposed model is computed using the Chib–Jeliazkov method. The posterior parameter samples generated using Metropolis–Hastings (M–H) Markov Chain Monte Carlo (MCMC) simulations are required by this method. The M–H MCMC-based parameter estimation procedure is complemented by an efficient particle filter and an ensemble Kalman filter for the strongly non-Gaussian state estimation problem. A strongly nonlinear system having multiple fixed (equilibrium) points is analyzed to demonstrate the efficacy of the algorithm.

The process personnel always seek the opportunity to improve the processes. One of the essential steps for the process improvement is to recognize the starting time or the change point of a process disturbance quickly. Different from the traditional normally distributed assumption for a process, this study considers a process that follows a gamma distribution. The proposed approach combines the commonly used control chart with the Bayesian estimation technique to estimate the change point. Two Bayes estimators corresponding to an informative and a noninformative prior along with maximum likelihood estimation (MLE) are considered. Their efficiency is compared through a series of simulations. The results show that the Bayes estimator with the informative prior is more accurate and more precise when the means of the process before and after the change point time are not too closed. Additionally, the efficiency of the Bayes estimator with the informative prior increases as either the sample size incre...

Bayesian experimental design is a fast growing area of research with many real-world applications. As computational power has increased over the years, so has the development of simulation-based design methods, which involve a number of algorithms, such as Markov chain Monte Carlo, sequential Monte Carlo and approximate Bayes methods, facilitating more complex design problems to be solved. The Bayesian framework provides a unified approach for incorporating prior information and/or uncertainties regarding the statistical model with a utility function which describes the experimental aims. In this paper, we provide a general overview on the concepts involved in Bayesian experimental design, and focus on describing some of the more commonly used Bayesian utility functions and methods for their estimation, as well as a number of algorithms that are used to search over the design space to find the Bayesian optimal design. We also discuss other computational strategies for further research in Bayesian optimal design.

We present a new approach to detection and tracking of moving objects with a 2D laser scanner for autonomous driving applications. Objects are modelled with a set of rigidly attached sample points along their boundaries whose positions are initialized with and updated by raw laser measurements, thus allowing a non-parametric representation that is capable of representing objects independent of their classes and shapes. Detection and tracking of such object models are handled in a theoretically principled manner as a Bayes filter where the motion states and shape information of all objects are represented as a part of a joint state which includes in addition the pose of the sensor and geometry of the static part of the world. We derive the prediction and observation models for the evolution of the joint state, and describe how the knowledge of the static local background helps in identifying dynamic objects from static ones in a principled and straightforward way. Dealing with raw laser points poses a significant challenge to data association. We propose a hierarchical approach, and present a new variant of the well-known Joint Compatibility Branch and Bound algorithm to respect and take advantage of the constraints of the problem introduced through correlations between observations. Finally, we calibrate the system systematically on real world data containing 7,500 labelled object examples and validate on 6,000 test cases. We demonstrate its performance over an existing industry standard targeted at the same problem domain as well as a classical approach to model-free object tracking.

Purpose: The purpose of this study is to apply the constrained Bayesian estimation methodology for product quality control process and prove the effectiveness of the product management by comparing with the well-known Bayes estimator through data performance result.Methods: The Bayes and constrained Bayes estimators were produced based on the theoretical background and for confirming the effectiveness of suggested application, the deviation index was defined and calculated for the comparison.Results: The statistical analysis result shows that applying the suggested estimation methodology, that is, constrained Bayes estimator improves the effectiveness of the index with regard to reduce the error by matching the first two empirical moments.Conclusion: Considering the advanced Bayesian approaches such as constrained Bayes estimation for the product quality control process, the newly defined deviation index reduces the error for estimating the param-eter histogram which is reflected both location and deviation parameters and furthermore various Bayesian perspective approaches seems to be meaningful for managing the product quality control process.Key Words: Bayesian Estimation, Constrained Bayes Estimator, Quality Control Process

The note considers the identification of multi-input multi-output errors-in-variables Hammerstein systems, in which both the input and output can be observed but with additive noises being ARMA processes with unknown coefficients. With the help of stochastic approximation combined with the deconvolution kernel function, the recursive algorithms are proposed for estimating coefficients of the linear subsystem and for the values of the nonlinear function. Under some reasonable conditions, all the estimates are proved to converge to the true values with probability one. These results include identification of the errors-in-variables linear systems as a special case. A simulation example is given justifying the theoretical analysis.

Recursive Bayesian estimation using sequential Monte Carlos methods is a powerful numerical technique to understand latent dynamics of nonlinear non-Gaussian dynamical systems. It enables us to reason under uncertainty and addresses shortcomings underlying deterministic systems and control theories which do not provide sufficient means of performing analysis and design. In addition, parametric techniques such as the Kalman filter and its extensions, though they are computationally efficient, do not reliably compute states and cannot be used to learn stochastic problems. We review recursive Bayesian estimation using sequential Monte Carlo methods highlighting open problems. Primary of these is the weight degeneracy and sample impoverishment problem. We proceed to detail synergistic computational intelligence sequential Monte Carlo methods which address this. We find that imbuing sequential Monte Carlos with computational intelligence has many advantages when applied to many application and problem domains.

Localization and tracking 1 are key components to enable personalized service in smart homes such as lifestyle support, energy management, and safety. To locate a person in a smart home, various sensors are installed. In this work, sensor data from only non Wifi-based devices and sensors such as PIR, pressure mat etc. are considered. The Particle filter has been adopted by many works because it has more advantages, e.g. accuracy, robustness, and efficiency, over the other techniques e.g. Kalman filter. Thus, we aim to analyze a set of motion models including random walk, random waypoint and Gaussian Markov model to find how applicable they are for an indoor setting and how they affect the accuracy of particle filter for indoor location estimation. We also analyze a set of sensor models. The result shows that the chosen motion models are applicable for an indoor environment if their parameters are adjusted properly. Random based models are more appropriate in this setting. However, information, such as behavior, activity, and location etc., should be considered to improve motion models. The accuracy of the particle filter is influenced by both motion and sensor models and their parameters such as walking speeds and degree of randomness, etc. Thus, the inaccurate prediction is produced if an improper model and its parameters are used and it is very difficult to correct. In addition, improper sensor models can cause each particle to be weighted incorrectly. Therefore, the estimation is not accurate.

Tracking-by-detection is a widely used practice in recent tracking systems. These usually rely on independent single frame detections that are handled as observations in a recursive estimation framework. If these observations are imprecise the generated trajectory is prone to be updated towards a wrong position. In contrary to existing methods our novel approach uses a Dynamic Bayes Network in which the state vector of a recursive Bayes filter, as well as the location of the tracked object in the image are modelled as unknowns. These unknowns are estimated in a probabilistic framework taking into account a dynamic model, and a state-of-the-art pedestrian detector and classifier. The classifier is based on the Random Forest-algorithm and is capable of being trained incrementally so that new training samples can be incorporated at runtime. This allows the classifier to adapt to the changing appearance of a target and to unlearn outdated features. The approach is evaluated on a publicly available benchmark. The results confirm that our approach is well suited for tracking pedestrians over long distances while at the same time achieving comparatively good geometric accuracy.

This chapter presents a noise adaptive cubature information filter based on variational Bayesian approximation (VB-ACIF). This filter can jointly estimate the dynamic state and time varying measurement noise for Gaussian nonlinear state space models. In the framework of recursive Bayesian estimation, the noise adaptive information filter propagating the information matrix and information state are derived. The integration of recursive Bayesian estimation is approximated by cubature integration rule. The inverse of measurement noise matrix, which is called measurement information matrix, is modeled as a Wishart distribution. So the joint distribution of posterior state and measurement noise can be approximated by the product of independent Gaussian and Wishart. As the parameters are coupled, the updated equation can be solved by fixed point iteration. The corresponding square root version of VB-ACIF is also derived to improve numerical precision and stability. Simulations are used to verify the performance of the proposed algorithms. Results demonstrate the improved performance of the proposed algorithms over conventional cubature information filter and square root version. When both state function and measurement function are nonlinear, the VB-ASCIF outperforms the VB-ACIF.

This chapter introduces an appearance-based approach for topological mapping and localization named FEATMap (Feature-based Mapping). FEATMap relies on a loop closure detection scheme which makes use of local invariant features to describe images. These features are indexed using a set of randomized kd-trees, which permit seeking for matchings between the current and previous images to detect loop closures in a straightforward way. A discrete Bayes filter is added to the solution to obtain loop candidates while ensuring the temporal coherency between consecutive predictions. Finally, FEATMap comprises a method for refining the resulting maps as they are obtained, removing spurious nodes in accordance to the visual information that they contain.

Global Navigation Satellite Systems (GNSS) is the cornerstone for positioning and timing applications across land, sea and air. GNSS carrier phase observations can be used to determine the orientation of a vehicle when this is equipped with a setup of multiple GNSS antennas. Attitude Determination seeks to find the rotation between two coordinate frames. In this contribution, we propose a formulation for the joint estimation of the positioning and attitude using quaternion rotations. The estimation is expressed in the recursive Bayesian estimation framework as an Extended Kalman Filter, whose equations and Jacobians matrices are derived. We evaluate the Performance of the proposed methodology with actual data collected from a dynamic scenario under the influence of severe multipath effects.

This paper presents power system frequency estimation by using an Improved Recursive Newton Type (IRNTA) algorithm. The proposed approach uses Jacobian and covariance matrices for updating the unknown parameters. The recursive form of unknown parameters and covariance matrix are incorporated in the algorithm to have faster convergence. The performance of the proposed algorithm is studied through simulations and experiments for several critical cases that often arise in a power system. Efficacy of the proposed algorithm is also compared with other signal processing techniques such as Recursive Least Square (RLS) and Kalman Filter (KF). Studies made on industrial data also support for the superiority of the proposed algorithm.

This paper investigates the problem of state estimation for discrete-time stochastic linear systems, where additional knowledge on the unknown inputs is available at an aggregate level and the knowledge on the missing measurements can be described by a known stochastic distribution. Firstly, the available knowledge on the unknown inputs and the state equation is used to form the prior distribution of the state vector at each time step. Secondly, to obtain an analytically tractable likelihood function, the effect of missing measurements is broken down into a systematic part and a random part, and the latter is modeled as part of the observation noise. Then, a recursive filter is obtained based on Bayesian inference. Finally, a numerical example is provided to evaluate the performance of the proposed methods.

This paper reviews a recursive Bayesian methodology for optimal data cleaning and filtering of economic time series data with the aim of using the Kalman filter to estimate the parameters of a specified state space model which describes an economic phenomena under study. The Kalman filter, being a recursive algorithm, is ideal for usage on time-dependent data. As an example, the yearly measurements of eight key economic time series data of the Nigerian economy is used to demonstrate that the integrated random walk model is suitable for modeling time series with no clear trend or seasonal variation. We find that the Kalman filter is both predictive and adaptive, as it looks forward with an estimate of the variance and mean of the time series one step into the future and it does not require stationarity of the time series data considered.

Nonlinear Bayesian filtering has been utilized in numerous fields and applications. One of the most popular class of Bayesian algorithms is Particle Filters. Their main benefit is the ability to estimate complex posterior density of the state space in nonlinear models. This paper presents the application of particle filtering to the problem of parameter estimation and calibration of a nonlinear power system model. The parameters of interest for this estimation problem are those of a turbine governor model. The results are compared to the performance of a heuristic method. Estimation results have been validated against real-world measurement data collected from staged tests at a Greek power plant. QC 20160408

Outdoor, GNSS enables continuous positioning 24 hours a day, if pseudoranges of at least four satellites are available. For the positioning in indoor environments, such a ubiquitous navigation solution does not exist. Nowadays, typical indoor solutions either depend on special infrastructure or do not fulfill the requirements of an accurate and reliable positioning. Consequently, reliable indoor positioning is based on sensor integration combining as much information as possible. Within this paper, a filter approach for pedestrian positioning is presented, which integrates fingerprinting-based WLAN-positions, motion parameters gained from a pedestrian dead reckoning (PDR) algorithm, barometer-based height information as well as the current activity determined within a pedestrian activity classification (PAC) approach. Geometric constraints, which are derived from a map of the indoor environment, further support the filter in its positioning task.

This thesis introduces a multi-Bayesian framework for detection and classification of features in environments abundant with error-inducing noise. The approach takes advantage of Bayesian correction and classification in three distinct stages. The corrective scheme described here extracts useful but highly stochastic features from a data source, whether vision-based or otherwise, to aid in higher-level classification. Unlike many conventional methods, these features’ uncertainties are characterized so that test data can be correctively cast into the feature space with probability distribution functions that can be integrated over class decision boundaries created by a quadratic Bayesian classifier. The proposed approach is specifically formulated for road crack detection and characterization, which is one of the potential applications. For test images assessed with this technique, ground truth was estimated accurately and consistently with effective Bayesian correction, showing a 33% improvement in recall rate over standard classification. Application to road cracks demonstrated successful detection and classification in a practical domain. The proposed approach is extremely effective in characterizing highly probabilistic features in noisy environments when several correlated observations are available either from multiple sensors or from data sequentially obtained by a single sensor. Master of Science Humans have an outstanding ability to understand things about the world around them. We learn from our youngest years how to make sense of things and perceive our environment even when it is not easy. To do this, we inherently think in terms of probabilities, updating our belief as we gain new information. The methods introduced here allow an autonomous system to think similarly, by applying a fairly common probabilistic technique to the task of perception and classification. In particular, road cracks are observed and classified using these methods, in order to develop an autonomous road condition monitoring system. The results of this research are promising; cracks are identified and correctly categorized with 92% accuracy, and the additional “intelligence” of the system leads to a 33% improvement in road crack assessment. These methods could be applied in a variety of contexts as the leading edge of robotics research seeks to develop more robust and human-like ways of perceiving the world.

The Battery Management System (BMS) is respon­sible for accurate monitoring of the status of the battery (State-of-Charge (SOC)) for maintaining optimal battery performance in Battery Electric Vehicles (BEVs). Ambient temperature is a significant factor that influences the accuracy of SOC estimation, hence electrochemical combined model dependents of tempera­ture was utilized for simulating the dynamic behavior of battery in BMS. Unknown parameters of the battery model are iden­tified using the least square algorithm for Dynamic Stress Test (DST), validation of estimation is conducted for Federal Urban Driving Schedule (FUDS) and concluded that the error between predicated terminal voltage form model and voltage from DST profile was less than 0.08V for defined conditions. Then, for SOC estimation, recursive Bayesian estimation method based Extended Kalman Filtering (EKF), and Sigma-Point Kalman Filtering (SPKF) approaches were adopted. To quantify the performance of the estimators, Root Mean Square Error (RMSE) and execution time at different temperature were evaluated. The evaluation results indicate that maximum error in case of EKF is 2.43% whereas for SPKF is 1.2% and maximum execution time taken by EKF is 3.57 sec whereas for SPKF is 4.53 sec. The results reported that SPKF provides accurate and robust SOC estimation in compared EKF and could be efficiently applied in BMS for BEVS.

The STARTS (Stable Trait, AutoRegressive Trait, and State) model decomposes individual differences in psychological measurement across time into 3 sources of variation: a time-invariant stable component, a time-varying autoregressive component, and an occasion-specific state component. Previous simulation research and applications of the STARTS model have shown that serious estimation problems such as nonconvergence or inadmissible estimates (e.g., negative variances) frequently occur for STARTS model parameters. This article introduces a general approach to estimating the parameters of the STARTS model by employing Bayesian methods that use Markov Chain Monte Carlo (MCMC) techniques. With the specification of appropriate prior distributions, the Bayesian approach offers the advantage that the model estimates will be within the admissible range, and it should be possible to avoid estimation problems. Furthermore, we show how Bayesian methods can be used to stabilize STARTS model estimates by specifying weakly informative prior distributions for the model parameters. In a simulation study, the statistical properties (bias, root mean square error, coverage rate) of the parameter estimates obtained from the Bayesian approach are compared with those of the maximum-likelihood approach. A data example is presented to illustrate how the Bayesian approach can be used to estimate the STARTS model. Finally, further extensions of the STARTS model are discussed, and suggestions for applied research are made. (PsycINFO Database Record

Within the context of road estimation, the present paper addresses the problem of the fusion of several sources with different reliabilities. Thereby, reliability represents a higher-level uncertainty. This problem arises in automated driving and ADAS due to changing environmental conditions, e.g., road type or visibility of lane markings. Thus, we present an online sensor reliability assessment and reliability-aware fusion to cope with this challenge. First, we apply a boosting algorithm to select the highly discriminant features among the extracted information. Using them we apply different classifiers to learn the reliabilities, such as Bayesian Network and Random Forest classifiers. To stabilize the estimated reliabilities over time, we deploy approaches such as Dempster-Shafer evidence theory and Influence Diagram combined with a Bayes Filter. Using a big collection of real data recordings, the experimental results support our proposed approach.

This paper addresses the issue of the optimal protection strategy with minimum protection cost against false data injection attacks, considering multiple power grid topologies. The model of robust optimal protection strategy, that can make power grids immune from false data injection attacks even if the grid topology changes, is first proposed. And then combining graph theory and recursive Bayesian estimation, we solve the model from the standpoint of pattern recognition. Finally, by testing on the IEEE 14-bus system, the correctness and effectiveness of our proposed robust optimal protection strategy model and its solving algorithm are demonstrated.

This paper presents a solution to the problem of tracking multiple targets from multistatic Doppler measurements. To the best of our knowledge our solution is the first principled multitarget tracker for multistatic Doppler radar. The tracker is based on a closed form solution to the labeled multitarget Bayes filter, which can be implemented with linear complexity in the number of measurements. Numerical examples are presented to illustrate the effectiveness of the solution on simulated data.

This paper studies the state estimation problem for a class of discrete-time nonlinear complex networks. The purpose is to design a recursive state estimator by using the variance-constrained approach such that the variance of the estimation error is not more than the prescribed upper bound. By adopting the structure of the extended Kalman filter (EKF), the gain matrix is determined by minimizing the trace of the prescribed upper bound matrix. It is shown that the estimator can be developed by solving two Riccati-like difference equations. A numerical example is provided to illustrate the effectiveness of the proposed estimator.

International audience; – A general approach is proposed to determine the occupancy in a room using sensor data and knowledge coming respectively from observation and questioning are determined. Means to estimate occupancy include motion detections, power consumptions and acoustic pressure rewarded by a microphone. The proposed approach is inspired from machine learning. It starts by determining the most useful measurements in calculating information gains. Then, a non supervised estimation algorithm is proposed: it relies on Hidden Markov Model and Bayesian Network algorithms to model a human behaviour with probabilistic cause-effect relations and states based on knowledge and questioning. Bayesian Network(BN) and Hidden Markov Model (HMM) based approaches have been applied to an office setting, with an average estimation error of 0.1-0.08 and an accuracy of 89%-91%. This approach avoids the usage of a camera to determine the actual occupancy required for supervised learning.

© 2017 IEEE. This paper presents a generalized multi-stage Bayesian approach for an unmanned aerial vehicle to estimate the location of a mobile target. The major hardware components of the proposed approach are a camera with a fisheye lens and another camera with a normal lens and a pan/tilt unit. With wide angle of view (AOV), the fisheye lens camera first detects the bearing of the target, and the PT camera next captures the target in its AOV. The recursive Bayesian estimation steadily locates the target in a globally defined space. The paper also proposes a multi-stage detection method for the fisheye lens camera. The level of confidence is defined in association with the probability of detection (POD) for each detection technique, and the fisheye lens enables continuous detection by gradually increasing the POD. The observation likelihood is finally derived from the POD in a generalized manner. The proposed approach was applied to the detection of a mobile target by a multi-rotor helicopter, and results have demonstrated the effectiveness of both the proposed multi-stage Bayesian approach and multi-stage fisheye lens detection method.

Robots are expected to operate autonomously in dynamic environments. Understanding the underlying dynamic characteristics of objects is a key enabler for achieving this goal. In this paper, we propose a method for pointwise semantic classification of 3D LiDAR data into three classes: non-movable, movable and dynamic. We concentrate on understanding these specific semantics because they characterize important information required for an autonomous system. Non-movable points in the scene belong to unchanging segments of the environment, whereas the remaining classes corresponds to the changing parts of the scene. The difference between the movable and dynamic class is their motion state. The dynamic points can be perceived as moving, whereas movable objects can move, but are perceived as static. To learn the distinction between movable and non-movable points in the environment, we introduce an approach based on deep neural network and for detecting the dynamic points, we estimate pointwise motion. We propose a Bayes filter framework for combining the learned semantic cues with the motion cues to infer the required semantic classification. In extensive experiments, we compare our approach with other methods on a standard benchmark dataset and report competitive results in comparison to the existing state-of-the-art. Furthermore, we show an improvement in the classification of points by combining the semantic cues retrieved from the neural network with the motion cues., 8 pages to be published in IROS 2017

The web is a source of a large amount of arguments and their acceptability statuses (e.g., votes for and against the arguments). However, relations existing between the fore-mentioned arguments are typically not available. This study investigates the utilisation of acceptability semantics to statistically estimate an attack relation between arguments wherein the acceptability statuses of arguments are provided. A Bayesian network model of argument-based reasoning is defined in which Dung's theory of abstract argumentation gives the substance of Bayesian inference. The model correctness is demonstrated by analysing properties of estimated attack relations and illustrating its applicability to online forums.

By introducing suitable loss random variables of detection, we obtain optimal tests in terms of the stopping time or alarm time for Bayesian change-point detection not only for a general prior distribution of change-points but also for observations being a Markov process. Moreover, the optimal (minimal) average detection delay is proved to be equal to $1$ for any (possibly large) average run length to false alarm if the number of possible change-points is finite.

In this paper, we consider the nonlinear filtering by using information geometric approach. Under the principle of Bayesian, the filtering problem has been converted to Bayesian estimation. Based on the estimation conditional on the measurement, the posterior probability density functions (PDFs) have constructed a statistical manifold. With the information geometric approach, the nonlinear characteristic has been convert to the Riemannian metric tensor, and the Bayesian estimation has been obtained by the natural gradient descent technique. Further, the information geometric nonlinear filtering has been induced. For the linear and Gaussian case, the metric tensor is constant, and the proposed method will be equivalent to the traditional Kalman filter. While the nonlinear case, the metric tensor is variable in which need be computed at each point on the statistical manifold, and the proposed method will be convergence along the optimal direction. The numerical experiment will show the better performance of our proposed method.

Brain-machine interfaces (BMIs) are a rapidly progressing technology with the potential to restore function to victims of severe paralysis via neural control of robotic systems. Great strides have been made in directly mapping a user’s cortical activity to control of the individual degrees of freedom of robotic end-effectors. While BMIs have yet to achieve the level of reliability desired for widespread clinical use, environmental sensors (e.g., RGB-D cameras for object detection) and prior knowledge of common movement trajectories hold great potential for improving system performance. Here, we present a novel sensor fusion paradigm for BMIs that capitalizes on information able to be extracted from the environment to greatly improve the performance of control. This was accomplished by using dynamic movement primitives to model the 3-D endpoint trajectories of manipulating various objects. We then used a switching unscented Kalman filter to continuously arbitrate between the 3-D endpoint kinematics predicted by the dynamic movement primitives and control derived from neural signals. We experimentally validated our system by decoding 3-D endpoint trajectories executed by a nonhuman primate manipulating four different objects at various locations. Performance using our system showed a dramatic improvement over using neural signals alone, with median distance between actual and decoded trajectories decreasing from 31.1 to 9.9 cm, and mean correlation increasing from 0.80 to 0.98. Our results indicate that our sensor fusion framework can dramatically increase the fidelity of neural prosthetic trajectory decoding.

With the ubiquity of information distributed in networks, performing recursive Bayesian estimation using distributed calculations is becoming more and more important. There are a wide variety of algorithms catering to different applications and requiring different degrees of knowledge about the other nodes involved. One recently developed algorithm is the distributed Kalman filter (DKF), which assumes that all knowledge about the measurements, except the measurements themselves, are known to all nodes. If this condition is met, the DKF allows deriving the optimal estimate if all information is combined in one node at an arbitrary time step. In this paper, we present an information form of the distributed Kalman filter (IDKF) that allows the use of explicit system inputs at the individual nodes while still yielding the same results as a centralized Kalman filter.

With the rise of end-to-end learning through deep learning, person detectors and re-identification (ReID) models have recently become very strong. Multi-camera multi-target (MCMT) tracking has not fully gone through this transformation yet. We intend to take another step in this direction by presenting a theoretically principled way of integrating ReID with tracking formulated as an optimal Bayes filter. This conveniently side-steps the need for data-association and opens up a direct path from full images to the core of the tracker. While the results are still sub-par, we believe that this new, tight integration opens many interesting research opportunities and leads the way towards full end-to-end tracking from raw pixels., First two authors have equal contribution. This is initial work into a new direction, not a benchmark-beating method. v2 only adds acknowledgements and fixes a typo in e-mail

International audience; Wind turbine noise is often annoying for humans living in close proximity to a wind farm. Reliably estimating the intensity of wind turbine noise is a necessary step towards quantifying and reducing annoyance, but it is challenging because of the overlap with background noise sources. Current approaches involve measurements with on/off turbine cycles and acoustic simulations, which are expensive and unreliable. This raises the problem of separating the noise of wind turbines from that of background noise sources and coping with the uncertainties associated with the source separation output. In this paper we propose to assist a black-box source separation system with a model of wind turbine noise emission and propagation in a recursive Bayesian estimation framework. We validate our approach on real data with simulated uncertainties using different nonlinear Kalman filters.

In this paper we present a Bayesian framework for parameter estimation and model selection for nonlinear aeroelastic oscillatory systems using their measured response. The framework is motivated by the fact that (1) the oscillations involve the complex nonlinear fluid–structure interaction whose numerical modeling poses both conceptual and computational challenges, (2) several plausible models fit the experimental data with reasonable accuracy, and (3) the model calibration involves estimating the joint probability density function (pdf) of high-dimensional (between 10 and 22) parameter vectors including the variance of model prediction error (modeling error). A four-stage process is followed. First, a candidate model set consisting of parameterized stochastic models is proposed based on the prior understanding of the physics. Second, the joint posterior pdf of the model parameters of each model is estimated using Bayesian inference which involves Markov Chain Monte Carlo (MCMC) sampling complemented by a nonlinear filter for state estimation. Owing to the dense observational data, the state estimation is performed by the extended Kalman filter (EKF). The unnormalized, high-dimensional, joint posterior pdf is sampled through an MCMC method, whereby a parallel version of single-block, adaptive proposal-based Metropolis–Hastings (MH) algorithm is employed to tune the proposal pdf. Third, using Bayes’ theorem each model is weighted according to the evidence provided by the data in its favor. For single-block MCMC sampling, an estimate of the evidence is readily available through the Chib–Jeliazkov method using the posterior samples generated after the burn-in. Fourth, the evidence is used to compute model plausibilities of each model, which are later incorporated in the Bayesian model averaging to make robust predictions. The resulting prediction incorporates uncertainties in the measurements, model parameters, model structure and model selection. A correction term handling the use of improper priors is shown to mimic the effect that diffuse (but proper) priors have on Bayesian model selection. The methodology is applied to wind-tunnel data of a nonlinear aeroelastic oscillator consisting of a rigid wing performing self-sustained small amplitude ( 6°) limit cycle oscillation (LCO) in pitch involving complex fluid–structure interactions.

Bayesian learning provides a firm theoretical basis of the design and exploitation of algorithms in data-streams processing (preprocessing, change detection, hypothesis testing, clustering, etc.). Primarily, it relies on a recursive parameter estimation of a firmly bounded complexity. As a rule, it has to approximate the exact posterior probability density (pd), which comprises unreduced information about the estimated parameter. In the recursive treatment of the data stream, the latest approximate pd is usually updated using the treated parametric model and the newest data and then approximated. The fact that approximation errors may accumulate over time course is mostly neglected in the estimator design and, at most, checked ex post. The paper inspects the estimator design with respect to the error accumulation and concludes that a sort of forgetting (pd flattening) is an indispensable part of a reliable approximate recursive estimation. The conclusion results from a Bayesian problem formulation complemented by the minimum Kullback-Leibler divergence principle. Claims of the paper are supported by a straightforward analysis, by elaboration of the proposed estimator to widely applicable parametric models and illustrated numerically.

This paper considers estimators for distributed systems and packet delivery prediction based on recursive Bayesian estimation.

This paper proposes a Bayesian method to address the farfield narrowband direction-of-arrival (DOA) estimation problem with time-varying arrays, whose elements relatively move in an arbitrary but known way. The measurements associated with different array geometries are formulated with distinct and spatially overcomplete observation systems, and a joint Bayesian model is established to combine those measurements and yield unified DOA estimates. The joint reconstruction process of the multiple measurements falls into the multitask learning category; thus, the proposed method is named DOA estimation via multitask learning (DEML). Theoretical results focusing on the uniqueness of the solution and the global convergence of the Bayesian learning process are also given, which indicate the maximal separable signal number and the global convergence of the proposed method in the considered array processing scenarios. Numerical examples are also provided to demonstrate the DOA estimation performance of the proposed method and support the theoretical results.

Multipath and shadow fading are the primary cause for positioning errors in a Received Signal Strength Indicator (RSSI) based localization scheme. While fading, in general, is detrimental to localization accuracy, cross-correlation and divergence properties of shadow fading residuals may be utilized to improve localization and tracking accuracy of mobile IEEE 802.15.4 transmitters. Therefore, this paper begins by presenting a stochastic filter that models the fast changing multipath fading as a mean reverting Ornstein-Uhlenbeck (OU) process followed by a Generalized Auto Regressive Conditional Heteroskedasticity (GARCH) filtering to isolate the slow changing shadow fading residuals from measured RSSI values. Subsequently, a novel wireless transmitter localization scheme that combines the measured cross-correlation in shadow fading residuals between adjacent receivers using a Student-t Copula likelihood function is proposed. However, the long convergence time for this highly non-convex copula function might render our method unsuitable for tracking applications. Therefore, we present a faster tracking method where the velocity and heading of a mobile transmitter are estimated from α-Divergence between shadow fading signals and an onboard gyroscope respectively. To bind the localization error in this tracking method, the transmitter location estimates are smoothed by a Bayesian particle filter. The performance of our proposed localization and tracking method is validated over simulations and hardware experiments.

This technical note concerns the control of partially observable Markov decision processes characterized by a prior distribution over the underlying hidden Markov model parameters. In such instances, the control problem is commonly simplified by first choosing a point estimate from the model prior, and then selecting the control policy that is optimal with respect to the point estimate. Our contribution is to demonstrate, through a tractable yet nontrivial example, that even the best control policies constructed in this manner can significantly underperform the Bayes optimal policy. While this is an operative assumption in the Bayes-adaptive Markov decision process literature, to our knowledge no such illustrative example has been formally proposed.

A low signal to noise ratio (SNR), single source/receiver, broadband, frequency-coherent matched-field inversion procedure recently has been proposed. It exploits coherently repeated transmissions to improve estimation of the geoacoustic parameters. The long observation time improves the SNR and creates a synthetic aperture due to relative source-receiver motion. To model constant velocity source/receiver horizontal motion, waveguide Doppler theory for normal modes is necessary. However, the inversion performance degrades when source/receiver acceleration exists. Furthermore processing a train of pulses all at once does not take advantage of the natural incremental acquisition of data along with the ability to assess the temporal evolution of parameter uncertainty. Here a recursive Bayesian estimation approach is developed that coherently processes the data pulse by pulse and incrementally updates estimates of parameter uncertainty. It also approximates source/receiver acceleration by assuming piecewise constant but linearly changing source/receiver velocities. When the source/receiver acceleration exists, it is shown that modeling acceleration can reduce further the parameter estimation biases and uncertainties. The method is demonstrated in simulation and in the analysis of low SNR, 100-900 Hz linear frequency modulated (LFM) pulses from the Shallow Water 2006 experiment.

Many tracking systems rely on independent single frame detections that are handled as observations in a recursive estimation framework. If these observations are imprecise the generated trajectory is prone to be updated towards a wrong position. In contrary to existing methods our novel approach suggests a Dynamic Bayes Network in which the state vector of a recursive Bayes filter, as well as the location of the tracked object in the image are modelled as unknowns. These unknowns are estimated in a probabilistic framework taking into account a dynamic model, prior scene information, and a state-of-the-art pedestrian detector and classifier. The classifier is based on the Random Forests-algorithm and is capable of being trained incrementally so that new training samples can be incorporated at runtime. This allows the classifier to adapt to the changing appearance of a target and to unlearn outdated features. The approach is evaluated on a publicly available dataset captured in a challenging outdoor scenario. Using the adaptive classifier, our system is able to keep track of pedestrians over long distances while at the same time supporting the localisation of the people. The results show that the derived trajectories achieve a geometric accuracy superior to the one achieved by modelling the image positions as observations.

This paper provided the E-Bayes estimation and hierarchical Bayes estimation of Rayleigh distribution parameter under q-symmetric entropy loss function, when the prior distribution is Beta distribution, and the relation between E-Bayes estimation and hierarchical Bayes estimation was discussed. Numerical simulation shows that the results of these two types of estimation are close, and E-Bayes estimation is more convenient in terms of calculation complexity.

Numerical methods for recursive Bayesian estimation are widespread in position tracking of robotics. However, challenges arise to indoor network positioning due to the limited processing power and inaccurate ranging measurements of low-end network nodes. For efficient and robust indoor position tracking, we incorporate a recursive bound to a grid-based filter namely RBGF, which approximates the posterior of the target's position by a grid of weighted cells over a bounded state-space. The state-space (the set in which the state samples can take) is recursively confined based on both the previous estimation and current measurements, therefore, the grid cells converge to the true state and the effect of non-line-of-sight (NLOS) measurements is bounded. Experimental results by an indoor sensor test-bed demonstrate RBGF achieves the average and the worst-case of positioning errors about 1 meter and 3 meters, respectively on condition that the average ranging error is about 3 meters.