Forgetting in Marginalized Particle Filtering and its Relation to Forward Smoothing, Report no. LiTH-ISY-R-3009
نویسنده
چکیده
The problem of degeneracy in marginalized particle filtering is addressed. In particular, we note that the degeneracy is caused by loss of entropy of the posterior distribution and design maximum entropy estimates to prevent this. The main technique used in this report is known as forgetting. It is shown that it can be used to suppress the problem with degeneracy, however, it is not a proper cure for the problem of stationary parameters. The problem of marginal-marginalized particle filter for sufficient statistics is also studied. The resulting algorithm is found to have remarkable similarities with the algorithm known as forward smoothing.
منابع مشابه
Rao-Blackwellised particle smoothers for mixed linear/nonlinear state-space models, Report no. LiTH-ISY-R-3018
We consider the smoothing problem for a class of mixed linear/nonlinear state-space models. This type of models contain a certain tractable substructure. When addressing the ltering problem using sequential Monte Carlo methods, it is well known that this structure can be exploited in a Rao-Blackwellised particle lter. However, to what extent the same property can be used when dealing with the s...
متن کاملComplexity Analysis of the Marginalized Particle Filter, Report no. LiTH-ISY-R-2680
In this paper the computational complexity of the marginalized particle filter is analyzed and a general method to perform this analysis is given. The key is the introduction of the equivalent flop measure. In an extensive Monte Carlo simulation different computational aspects are studied and compared with the derived theoretical results.
متن کاملThe Marginalized Auxiliary Particle Filter, Report no. LiTH-ISY-R-2934
In this paper we are concerned with nonlinear systems subject to a conditionally linear, Gaussian sub-structure. This structure is often exploited in high-dimensional state estimation problems using the marginalized (aka Rao-Blackwellized) particle lter. The main contribution in the present work is to show how an e cient lter can be derived by exploiting this structure within the auxiliary part...
متن کاملEstimation of General Nonlinear State-Space Systems, Report no. LiTH-ISY-R-3001
This paper presents a novel approach to the estimation of a general class of dynamic nonlinear system models. The main contribution is the use of a tool from mathematical statistics, known as Fishers' identity, to establish how so-called particle smoothing methods may be employed to compute gradients of maximum-likelihood and associated prediction error cost criteria.
متن کاملInference in Mixed Linear/Nonlinear State-Space Models using Sequential Monte Carlo, Report no. LiTH-ISY-R-2946
In this work we apply sequential Monte Carlo methods, i.e., particle lters and smoothers, to estimate the state in a certain class of mixed linear/nonlinear state-space models. Such a model has an inherent conditionally linear Gaussian substructure. By utilizing this structure we are able to address even high-dimensional nonlinear systems using Monte Carlo methods, as long as only a few of the ...
متن کامل