NONLINEAR CONTROLLER COMPARISON ON A BENCHMARK ...

NONLINEAR CONTROLLER COMPARISON ON A BENCHMARK ... NONLINEAR CONTROLLER COMPARISON ON A BENCHMARK ...

24.08.2013 Views

Response in cm 4 3 2 1 0 −1 −2 −3 Linearized Optimal Open Loop Response −4 0 1 2 3 4 5 Time 6 7 8 9 10 Figure 4.3: Linearized Optimal vs. Open Loop Response solutions to plants that are di cult to regulate using the standard linear approaches. Figure 4.3 shows the LQR design's performance plotted against the open loop re- sponse. This linearized H 2 approach gave almost the best performance in simulations. This is reasonable since the initial disturbance was small, and the system stayed within a region where linearization yields a good estimate of the true system dynamics. It should be said, though, that this control relied on relatively high gains (Kc = [;658 69:8 26 3:7]), and it was later discovered that such gains resulted in actuator dysfunction and instability on the hardware system. This really highlights the goal of nonlinear control design research: to nd control laws that will e ectively regulate systems when they operate in the presence of real-world nonlinearities or when they operate well beyond their linear region. 32

Response in cm 4 3 2 1 0 −1 −2 −3 Linearized Robust Open Loop Response −4 0 1 2 3 4 5 Time 6 7 8 9 10 4.2.2 Linearized H1 Figure 4.4: Linear H1 vs. Open Loop Response The linearized H1 control also gave good performance, though it could not match the performance of the linearized H 2 controller. In fact, this linearized H1 control could match neither the H 2 SGA nor the Backstepping control's performance. Figure 4.5 shows that the price of a more robustness is less performance. The gains produced by the linearized H1 design (Kc = [;62 16:7 1:5:46]) were muchlower than those of the LQR design. These gains were, in fact, more robust at least in the sense that they produced a control that succeeded in stabilizing the actual exible beam system unlike the gains produced by the linearized H 2 approach. 4.2.3 Passivity Based Control As seen in in Figure 4.6, the passivity based control exhibited some strange behavior: after suddenly attenuating the vibration at about t = 6 the oscillations then increase a little before dying out. Though this control is still a dramatic improvement on the open loop response, both of the linearized controllers provided better damping 33

Response in cm<br />

4<br />

3<br />

2<br />

1<br />

0<br />

−1<br />

−2<br />

−3<br />

Linearized Robust<br />

Open Loop Response<br />

−4<br />

0 1 2 3 4 5<br />

Time<br />

6 7 8 9 10<br />

4.2.2 Linearized H1<br />

Figure 4.4: Linear H1 vs. Open Loop Response<br />

The linearized H1 control also gave good performance, though it could not<br />

match the performance of the linearized H 2 controller. In fact, this linearized H1<br />

control could match neither the H 2 SGA nor the Backstepping control's performance.<br />

Figure 4.5 shows that the price of a more robustness is less performance.<br />

The gains produced by the linearized H1 design (Kc = [;62 16:7 1:5:46])<br />

were muchlower than those of the LQR design. These gains were, in fact, more robust<br />

at least in the sense that they produced a control that succeeded in stabilizing the<br />

actual exible beam system unlike the gains produced by the linearized H 2 approach.<br />

4.2.3 Passivity Based Control<br />

As seen in in Figure 4.6, the passivity based control exhibited some strange<br />

behavior: after suddenly attenuating the vibration at about t = 6 the oscillations then<br />

increase a little before dying out. Though this control is still a dramatic improvement<br />

on the open loop response, both of the linearized controllers provided better damping<br />

33

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

Saved successfully!

Ooh no, something went wrong!