28.02.2014 Views

The Development of Neural Network Based System Identification ...

The Development of Neural Network Based System Identification ...

The Development of Neural Network Based System Identification ...

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

206 CHAPTER 7 FLIGHT CONTROL SYSTEM DESIGN: RESULTS AND DISCUSSION<br />

7.4 SUMMARY<br />

<strong>The</strong> flight test results for the unmanned helicopter system controlled by multiple NNAPC<br />

controllers are presented in this chapter. <strong>The</strong> automatic hovering flight controllers are<br />

designed according to the TITO architecture which result in three NNAPC controllers<br />

i.e. the coupled roll-pitch controller, the yaw rate controller and the altitude speed<br />

controller. Several flight tests have been conducted in the early parts <strong>of</strong> controller<br />

development to identified the control tuning parameters such as r w , N p , K Ψ and K w<br />

that would produced satisfactory control performance. <strong>The</strong> HMLP network presented<br />

in the Subsection 4.2.2 is used in the NNAPC controller prediction due the network<br />

capability to model the flight dynamics with fewer weight connections compared with<br />

the standard MLP network. <strong>The</strong> control tuning results show that tuning parameters<br />

such as r w = 1.5, N p = 10, K Ψ = 1.0 and K w = 0.75 produce the best flight control<br />

response compared with other tuning setting. Using the tuning parameters identified<br />

from the control tuning tests, a full 4 DOF NNAPC controller is successfully realised<br />

in the hovering flight condition with good compensation performance. In the hovering<br />

flight test, the unmanned helicopter is commanded to track a step response in altitude<br />

for a duration <strong>of</strong> 19.8 s while stabilising the helicopter in roll, pitch and yaw channels.<br />

After the successful implementation <strong>of</strong> the 4 DOF flight controllers, the NNAPC<br />

controllers are further tested under the physical parameter variations that involve<br />

changes in the collective pitch curve setting, throttle curve setting and the test rig’s<br />

counterbalance weight. <strong>The</strong> changes in the counterbalance weight can be equated to the<br />

changes in the unmanned helicopter payload during flight. Similar to the approach taken<br />

by Samal [2009], the changes in both collective and throttle curves <strong>of</strong> the helicopter<br />

blades are used as the test reference. In real flight condition, the total lift and drag<br />

forces acting on the helicopter vary during different flight and environment conditions.<br />

This can be treated similarly in the indoor test by setting changes in the collective pitch<br />

and throttle setting during flight. <strong>The</strong> performance <strong>of</strong> the NNAPC-Offline controller<br />

is then compared with the NNAPC-Online controller under parameter variation tests.<br />

<strong>The</strong> <strong>of</strong>f-line NN model and the initial model for recursive HMLP are obtained using the

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!