Another modeling technique considers recursive identification methods. This
considers calculating a model simultaneous to obtaining data. However, this is not
a requirement for this specific project but can be useful in different applications.
Certain applications include having an up to date model in order to consider these
parameters when making decisions about what the system is to do next. This is
typically referred to as an adaptive modeling technique because the input and output
signals are calculated in order to be used as they become available.
An example of a recursive model which can be used for system identification in
Matlab is the RARMAX model. This uses a recursive technique of an ARMAX model
which considers the noise in its calculations. However, this technique only provides
models for single-input, single-output systems. Similarly, another technique is the
RARX model which estimates parameters recursively of a single-output system.
Therefore, for this project an initial linear approximation was done using an ARX
technique. The initial step was to design an experiment which consisted of specified
maneuvers such as doublets to the morphing and rudder servos. These were done in
order to consider the roll and yaw rate responses of the system.
The data is then collected and processed before considering it for modeling. The
data processing included using an algorithm which plotted, filtered and removed the
bias in the data. The filtering was done using a low pass Butterworth filter on all the
parameters and the bias was removed from the parameters by subtracting the mean.
This processed data is then used in the ARX modeling approximation. The roll
rate and yaw rate responses are then compared to the simulation responses. This is
done for both the morphing and rudder servos. The orders and delays are selected for
the parameter estimation.