IDENTIFICATION OF UNDER WATER VEHICLE HYDRODYNAMIC EFFICIENTS USING FREE DECAY TESTS

[attachment=18367]

INTRODUCTION
In recent years there has been an ever increasing number
of applications for unmanned underwater vehicles
(UUV) in various tasks, for instance in surveying and
exploration, or in missions such as the positioning
of underwater laboratories (Aguiar 1997). Improvements
in the evaluation of the hydrodynamic models
of UUV’s result in more effective control system
implementation, leading to increased capability and
performance in underwater operations. The efcient
identication of hydrodynamic coefcients is a task
which is difcult and oftentimes expensive to carry
out, with many examples of how to measure or estimate
them. For example, the use of towing tanks in
a marine laboratory (Aage 1994) is well-established,
as is the hydrodynamic modelling of underwater vehicles
in computational uid dynamics programs such as
WAMIT.



2. FREE DECAY TEST
The MIMO free decay test is performed according to
ow chart in Figure 1.
2.1 State Measurements
For the experiments envisaged in this paper, an underwater
camera system is to be used. This system
works by identifying pre-determined markings on the
UUV, and measures at 20 Hz. Using cameras offers
measurements of only the generalised position, , and
therefore suitable methods must be applied in order to
generate the estimates of the states  and _ .



3. SYSTEM IDENTIFICATION
The task of system identication is essentially matching
a model of some form to experimental data, in
order that the model explains, in some fashion, the
experimental data.



5. CONCLUSIONS
The new methods are presented, with simulated experimental
results for the longitudinal model, demonstrating
the combination of simulation with signal processing
and system identication. The implications of the
paper must be veried through actual experimentation
to gain a clearer picture of the applicability and usefulness
of the methods presented. The fact that the evaluation
process converges to the correct parameters with
full state knowledge implies identiability, and so the
problem is primarily one of achieving effective state
measurement. The penalty of using signal processing
method is that, although cheap, errors arise in
state measurement, leading to the innacuracies noted
in the previous section.