Thesis - Leigh Moody.pdf - Bad Request - Cranfield University

Chapter 1 / Introduction
_ _
radar target state observer in §4. Bar-Shalom categorises tracking filters
considered as:
• Static, non-switching models
• Optimally Multiple-Model (MM) estimators
• Sub-optimal MM estimators with Markov switching
Fixed gain filters tend to be inflexible and ignore cross coupling effects
resulting in poor performance against complex targets. Since processor
loading is less of an issue that it once was, these are rarely in modern state
observers having been replaced by time-varying stochastic alternatives. Of
these the sub-optimal Interacting-Multiple-Model (IMM) was chosen for its
flexibility, stability through re-initialisation, and its potential for expansion
into the variable-structured algorithm suggested by Li [L.5] .
The core of the IMM comprises filters with different dynamic models driven
by the same radar measurements. Although the IMM can accommodate
many types of filter the reasons for using the conventional Extended
Kalman Filter (EKF) are discussed. Whilst other options were considered,
EKF iterative and adaptive forms, Second Order Filters (SOF), many were
discounted due to computational burden, indirect access to covariance data,
inability to process range and range-rate data, etc. The suitability of the
EKF and its iterated form is discussed in the context of operating in
Cartesian or polar state-space, measurement conversion, and initialisation
using α−β−γ filters and least squares algorithms. §4 concludes with a
description of the tracking simulator, its controls, and interaction with the
target and sensor simulators.
§5 extends the state observation started in §4 with a review of centralised
and decentralised observer architectures covering the benefits of track and
measurement data fusion. The hybrid architecture chosen for this
application combines the up-linked IMM target track with the radar’s
missile measurements, and the missile gyroscope, accelerometer and seeker
measurements. The track is fused into the missile observer target state using
information filtering techniques favoured by Lobbia [L.8] and Durrant-
White [J.2] .
The measurements are serial processed in a conventional EKF whose
process model is purposely simple so as to promote high frequency
propagation. State cross-coupling is introduced using directed process noise
and pseudo-measurement updates. Constraining the process model through
the measurement update provides a flexible observer since executive control
can be exercised over the type and frequency of the pseudo-measurements
used. Pseudo-measurements are derived from aerodynamic constraints,
circular target models, PN goal orientation, and quaternion normalisation.
Protecting the observers from spurious measurements and failing sensors is
crucial. Commonly employed integrity algorithms are discussed noting the
work of Daum and Fitzgerald [D.3] , Kerr [K.6] amongst others in this field. §5
1-9