Search

Highlights • Proposing new algorithm that combines KLD sampling with Gmapping proposal distribution. • Reducing the number of required particles with an accurate proposal distribution. • Simulation evaluation to show performance improvement over KLD sampling. • Experimental evaluation to show greater accuracy with a smaller number of particles. Abstract The paper proposes an algorithm for mobile robot navigation that integrates the Gmapping proposal distribution with the Kullback–Leibler divergence for adapting the number of particles. This results in a very effective particle filter with adaptive sample size. The algorithm has been evaluated in both simulation and experimental studies, using the standard KLD—sampling MCL as a benchmark. Simulation results show that the proposed algorithm achieves higher localization accuracy with a smaller number of particles compared to the benchmark algorithm. In a more realistic scenario using experimental data and simulated robot odometry with drift, the proposed algorithm again has greater accuracy using a lower number of particles. [ABSTRACT FROM AUTHOR]

The paper is devoted to Bayesian target motion analysis (TMA) for the case when the variance of additive white zero-mean Gaussian measurement noise is unknown. Two Rao–Blackwellised particle filters for TMA are developed, which jointly estimate the target state and the measurement variance. The error performance of the two particle filters is compared against the theoretical Cramer–Rao lower bound. The bound suggests the error in target state estimation is not affected by the ignorance of the measurement noise variance. Both developed TMA algorithms reach this theoretical bound, however, one is significantly faster. [ABSTRACT FROM AUTHOR]

The problem is to estimate the parameters of a source continuously releasing hazardous material into the atmosphere. The concentration measurements are collected at a number of known locations by a moving binary sensor, characterised by an unknown threshold . The paper formulated a solution in the Bayesian framework, using a dispersion model of Poisson distributed particle encounters in a turbulent flow and assuming the environmental parameters (wind velocity, diffusivity, particle lifetime) are known. The method is implemented using an importance sampling technique and successfully validated with three experimental datasets under different wind conditions. In this context, the estimates of the source release rate are not of practical use, being scaled with an unknown constant related to the binary threshold. [ABSTRACT FROM AUTHOR]

Localisation of a source of a toxic release of biochemical aerosols in the atmosphere is a problem of great importance for public safety. Two main practical difficulties are encountered in this problem: the lack of knowledge of the likelihood function of measurements collected by biochemical sensors, and the plethora of candidate dispersion models, developed under various assumptions (e.g. meteorological conditions, terrain). Aiming to overcome these two difficulties, the paper proposes a likelihood-free approximate Bayesian computation method, which simultaneously uses a set of candidate dispersion models, to localise the source. This estimation framework is implemented via the Monte Carlo method and tested using two experimental datasets. [ABSTRACT FROM AUTHOR]

The paper presents a framework for autonomous search for a diffusive emitting source of a tracer (e.g., aerosol, gas) in an environment with an unknown map of randomly placed and shaped obstacles. The measurements of the tracer concentration are sporadic, noisy and without directional information. The search domain is discretised and modelled by a finite two-dimensional lattice. The links in the lattice represent the traversable paths for emitted particles and for the searcher. A missing link in the lattice indicates a blocked path due to an obstacle. The searcher must simultaneously estimate the source parameters, the map of the search domain and its own location within the map. The solution is formulated in the sequential Bayesian framework and implemented as a Rao-Blackwellised particle filter with entropy-reduction motion control. The numerical results demonstrate the concept and its performance. [ABSTRACT FROM AUTHOR]

The problem considered in this paper is detection and estimation of multiple radiation sources using a time series of radiation counts from a collection of sensors. A Bayesian framework is adopted. Source detection is approached as a model selection problem in which competing models are compared using partial Bayes factors. Given the number of sources, the posterior mean is the minimum mean square error estimator of the source parameters. Exact calculation of the partial Bayes factors and the posterior mean is not possible due to the presence of intractable integrals. Importance sampling using progressive correction is proposed as a computationally efficient method for approximating these integrals. Previously proposed algorithms have been restricted to one or two sources. A simulation analysis shows that the proposed methods can detect and accurately estimate the parameters of four sources with reasonable computational expense. [ABSTRACT FROM AUTHOR]

A theoretical framework is presented for target classification based on the belief theory on the continuous space. The proposed approach is applicable when class-conditioned densities of feature/attribute measurements are known only partially, as subjective models of a potential "betting" behaviour. Prior class probabilities may also be unknown. Numerical examples are provided to illustrate how the proposed approach is more cautious in decision making and produces very different results from those obtained using the Bayesian classifier. [ABSTRACT FROM AUTHOR]

In our recent paper [1] we demonstrated that the standard Bayes classifier, when applied to a problem characterised by imprecise likelihood functions, produces results which are inconsistent with our intuition. As a more appropriate alternative to the standard Bayes classifier, we proposed in [1] a classification method based on the transferrable belief model (TBM). Mahler [2, ch. 4–8] recently proposed a novel approach to Bayesian estimation, fusion, and classification, applicable to situations where the information (priors, measurements, likelihoods) is imprecise and vague in addition to being random. The purpose of this letter is to demonstrate that Mahler's approach can produce identical results to those obtained using the TBM classifier. [ABSTRACT FROM PUBLISHER]

Source term estimation for atmospheric dispersion deals with estimation of the emission strength and location of an emitting source using all available information, including site description, meteorological data, concentration observations and prior information. In this paper, Bayesian methods for source term estimation are evaluated using Prairie Grass field observations. The methods include those that require the specification of the likelihood function and those which are likelihood free, also known as approximate Bayesian computation (ABC) methods. The performances of five different likelihood functions in the former and six different distance measures in the latter case are compared for each component of the source parameter vector based on Nemenyi test over all the 68 data sets available in the Prairie Grass field experiment. Several likelihood functions and distance measures are introduced to source term estimation for the first time. Also, ABC method is improved in many aspects. Results show that discrepancy measures which refer to likelihood functions and distance measures collectively have significant influence on source estimation. There is no single winning algorithm, but these methods can be used collectively to provide more robust estimates. [ABSTRACT FROM AUTHOR]