Use this URL to cite or link to this record in EThOS:
Title: Sequential Gaussian mixture techniques for target tracking applications
Author: Robbiati, Stefano Andrea
ISNI:       0000 0004 2743 3316
Awarding Body: Imperial College London
Current Institution: Imperial College London
Date of Award: 2006
Availability of Full Text:
Access from EThOS:
Access from Institution:
Target tracking is a well known state estimation problem which has a simple solution in the form of the Kalman filter under Gaussian and linear model assumptions. The challenge arises in problems where these assumptions are no longer valid. To tackle these cases, a variety of suboptimal moment-matching techniques, such as the extended Kalman filter, have been developed over the years. More recently, randomised sequential Monte Carlo techniques, generally referred to as particle filters, have shown how large numbers of sample simulations can be used to produce very accurate estimates (at the cost of significant computational requirements). This thesis focuses on the use of Gaussian mixture techniques for the solution of target tracking problems in which model nonlinearity arises in consequence of switching model parameters. Problems of this nature, encountered for example in manoeuvring target tracking or target classification, involve model structures well suited to Gaussian mixture approximations. These structures have been earlier exploited in deterministic algorithms such as pseudo-Bayesian and interacting multiple model techniques. Rao-Blackwellised particle filters, in which random selection is carried out only with respect to the system modes, also employ Gaussian mixture approximations. In this thesis, efficient mixture filters with features associated with both particle fil-ters and deterministic algorithms will be discussed. These techniques provide a more accurate representation of the posterior distribution of the target, by permitting the complexity of the conditional distribution to expand and contract. This is carried out by means of efficient mixture reduction techniques, a discussion of which will tie in with clustering methods, normally used for data association. Furthermore, it will be shown how these algorithms can be directly extended to take into account of extra non-linearities, such as multiple and/or nonlinear measurements, while still retaining their simplicity and their computational efficiency.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available