Optimal filtering problems are ubiquitous in signal processing and related fields. Except for a restricted class of models, the optimal filter does not admit a closed-form expression. Particle filtering methods are a set of flexible and powerful sequential Monte Carlo methods designed to solve the optimal filtering problem numerically. The posterior distribution of the state is approximated by a large set of Dirac-delta masses (samples/particles) that evolve randomly in time according to the dynamics of the model and the observations. The particles are interacting; thus, classical limit theorems relying on statistically independent samples do not apply. In this paper, our aim is to present a survey of recent convergence results on this class of methods to make them accessible to practitioners.

“particle filters, ” refers to a general class of iterative algorithms that performs Monte Carlo approximations of a given sequence of distributions of interest (πt). We establish in this paper a central limit theorem for the Monte Carlo estimates produced by these computational methods. This result holds under minimal assumptions on the distributions πt, and applies in a general framework which encompasses most of the sequential Monte Carlo methods that have been considered in the literature, including the resample-move algorithm of Gilks and Berzuini [J. R. Stat. Soc. Ser. B Stat. Methodol. 63 (2001) 127–146] and the residual resampling scheme. The corresponding asymptotic variances provide a convenient measurement of the precision of a given particle filter. We study, in particular, in some typical examples of Bayesian applications, whether and at which rate these asymptotic variances diverge in time, in order to assess the long term reliability of the considered algorithm. 1. Introduction. Sequential Monte Carlo methods form an emerging

powerful tool to perform computations in general state space models. We discuss and compare the accept–reject version with the more common sampling importance resampling version of the algorithm. In particular, we show how auxiliary variable methods and stratification can be used in the accept–reject version, and we compare different resampling techniques. In a second part, we show laws of large numbers and a central limit theorem for these Monte Carlo filters by simple induction arguments that need only weak conditions. We also show that, under stronger conditions, the required sample size is independent of the length of the observed series. 1. State space and hidden Markov models. A general state space or hidden Markov model consists of an unobserved state sequence (Xt) and an observation sequence (Yt) with the following properties: State evolution: X0,X1,X2,... is a Markov chain with X0 ∼ a0(x)dµ(x) and Xt|Xt−1 = xt−1 ∼ at(xt−1,x)dµ(x). Generation of observations: Conditionally on (Xt), the Yt’s are independent and Yt depends on Xt only with Yt|Xt = xt ∼ bt(xt,y)dν(y). These models occur in a variety of applications. Linear state space models are equivalent to ARMA models (see, e.g., [16]) and have become popular Received January 2003; revised August 2004. AMS 2000 subject classifications. Primary 62M09; secondary 60G35, 60J22, 65C05. Key words and phrases. State space models, hidden Markov models, filtering and smoothing, particle filters, auxiliary variables, sampling importance resampling, central limit theorem. This is an electronic reprint of the original article published by the

We propose a new approach to study the stability of the optimal filter w.r.t. its initial condition, by introducing a "robust" filter, which is exponentially stable and which approximates the optimal filter uniformly in time. The "robust" filter is obtained here by truncation of the likelihood function, and the robustification result is proved under the assumption that the Markov transition kernel satisfies a pseudo-mixing condition (weaker than the usual mixing condition), and that the observations are "sufficiently good". This robustification approach allows us to prove also the uniform convergence of several particle approximations to the optimal filter, in some cases of nonergodic signals.

Abstract. Stability problem of the Wonham filter with respect to initial conditions is addressed. The case of ergodic signals is revisited in view of a gap in the classic work of H. Kunita (1971). We give new bounds for the exponential stability rates, which do not depend on the observations. In the non-ergodic case, the stability is implied by identifiability conditions, formulated explicitly in terms of the transition intensities matrix and the observation structure. Key words. Nonlinear filtering, stability, Wonham filter

The aim is to model "activity" performed by a group of moving and interacting objects (which can be people or cars or different rigid components of the human body) and use the models for abnormal activity detection. Previous approaches to modeling group activity include co-occurrence statistics (individual and joint histograms) and Dynamic Bayesian Networks, neither of which is applicable when the number of interacting objects is large. We treat the objects as point objects (referred to as "landmarks") and propose to model their changing configuration as a moving and deforming "shape" (using Kendall's shape theory for discrete landmarks). A continuous state Hidden Markov Model (HMM) is defined for landmark shape dynamics in an activity. The configuration of landmarks at a given time forms the observation vector and the corresponding shape and the scaled Euclidean motion parameters form the hidden state vector. An abnormal activity is then defined as a change in the shape activity model, which could be slow or drastic and whose parameters are unknown. Results are shown on a real abnormal activity detection problem involving multiple moving objects.

Target tracking is usually performed using data from sensors such as radar, whilst the target identification task usually relies on information from sensors such as IFF, ESM or imagery. The differing nature of the data from these sensors has generally led to these two vital tasks being performed separately. However, it is clear that an experienced operator can observe behaviour characteristics of targets and, in combination with knowledge and expectations of target type and likely activity, can more knowledgeably identify the target and robustly predict its track than any automatic process yet defined. Most trackers are designed to follow targets within a wide envelope of trajectories and are not designed to derive behaviour characteristics or include them as part of their output. Thus, there is potential scope for both applying target type knowledge to improve the reliability of the tracking process, and to derive behavioural characteristics which may enhance knowledge about target identity and/or activity. In this paper we introduce a Bayesian framework for joint tracking and identification and give a robust and computationally efficient particle filter based algorithm for numerical implementation of the resulting recursions. Simulation results illustrating algorithm performance are presented.

The nonlinear filtering equation is said to be stable if it “forgets” the initial condition. It is known that the filter might be unstable even if the signal is ergodic Markov chain. In general, the filtering stability requires stronger signal ergodicity provided by, so called, mixing condition. The latter is formulated in terms of the transition probability density of the signal. The most restrictive requirement of the mixing condition is a uniform positiveness of this density. We show that this requirement might be weakened regardless of an observation process structure.

Abstract: The basic nonlinear ltering problem for dynamical systems is considered. Approximating the optimal lter estimate by particle lter methods has become perhaps the most common and useful method in recent years. Many variants of particle lters have been suggested, and there is an extensive literature on the theoretical aspects of the quality of the approximation. Still, a clear cut result that the approximate solution, for unbounded functions, converges to the true optimal estimate as the number of particles tends to in nity seems to be lacking. It is the purpose of this contribution to give such a basic convergence result.

We study the change detection problem in partially observed nonlinear dynamic systems. We assume that the change parameters are unknown and the change could be gradual (slow) or sudden (drastic). For most nonlinear systems, no finite dimensional filters exist and approximation filtering methods like the Particle Filter are used. Even when change parameters are unknown, drastic changes can be detected easily using the increase in tracking (output) error or the negative log of observation likelihood (OL). But slow changes usually get missed. We propose in this paper, a statistic for slow change detection which turns out to be the same as the Kerridge Inaccuracy between the posterior state distribution and the normal system prior. We show asymptotic convergence (under certain assumptions) of the bounding, modeling and particle filtering errors in its approximation using a particle filter optimal for the normal system. We also demonstrate using the bounds on the errors that our statistic works in situations where observation likelihood (OL) fails and vice versa.