24.08.2013 Views

NONLINEAR CONTROLLER COMPARISON ON A BENCHMARK ...

NONLINEAR CONTROLLER COMPARISON ON A BENCHMARK ...

NONLINEAR CONTROLLER COMPARISON ON A BENCHMARK ...

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

limitations, k 1 and k 2 are tuning parameters to properly shape the closed loop en-<br />

ergy function, and q 3 is the actuated plant variable, . The controller dynamics are<br />

computed from the interconnection constraint andthe controller's EL equation:<br />

u = ;k 2 tanh(qc + b ) ; k 1 tanh( )<br />

qc _ = ;ak2 tanh(qc + b ):<br />

(3.17)<br />

Thus the control takes the measurable output, , as its only input, and the dynamics<br />

in the control produce a pseudo-derivative of to generate the necessary damping.<br />

The mathematical model for the exible beam system was slightly di erent from the<br />

nonlinear benchmark problem due to the extra term that resulted from the fact that<br />

we used voltage as an input instead of torque. This made the implementation of the<br />

saturation constraint a little tricky. In fact, the particular software/hardware con g-<br />

uration of the exible beam system required the removal of the saturation functions<br />

in the controller dynamics. (The actual control signal was up = ;k 2(qc + b ) ; k 1 .)<br />

To implement the control was simple: nd appropriate values for a b k 1 and k 2.<br />

The constants a = 250, b = 2:7, k 1 = :29 and k 2 = 2:5 were chosen experimentally<br />

to give the best possible dynamic performance. This control was relatively easy to<br />

implement, and its great advantage is that it only uses measurable outputs of the<br />

system. This strength can also be seen as a weakness, though the controller does<br />

not utilize all of the available information { only one of the two truly measurable<br />

states is used to generate the control signal. From an informational stand-point, the<br />

ideal controller should use all of the available plant output signals. However, the<br />

passivity-based control can clearly do more than a feedback system consisting of a<br />

simple gain matrix { PBC systems implement a dynamic system in the feedback loop<br />

that, in our case, provides information about _ as well as .<br />

3.4 Backstepping<br />

Another recent approach to the problem of robustly stabilizing non-linear sys-<br />

tems is that of integrator backstepping. This control strategy is an attempt to con-<br />

struct for a given non-linear system a control that will render the closed-loop system's<br />

21

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

Saved successfully!

Ooh no, something went wrong!