31-03-2010, 07:53 PM
Abstract
Methods to synthesize controllers for nonlinear systems are developed by exploiting the fact that under mild differentiability conditions, systems of the form:
x = f (x) + G(x)u
can be represented in quasilinear form, viz:
x = A(x)x + B(x)u
Two classes of control methods are investigated:
¢ zero-look-ahead control, where the control input depends only on the current val¬ues of A(x), B{x). For this case the control input is computed by continuously solving a matrix Ricatti equation as the system progresses along a trajectory.
¢ controllers with look-ahead, where the control input depends on the future be¬havior of A(x), B(x). These controllers use the similarity between quasilinear systems, and linear time varying systems to find approximate solutions to op¬timal control type problems.
The methods that are developed are not guaranteed, to be globally stable. However in simulation studies they were found to be useful alternatives for synthesizing control laws for a general class of nonlinear systems.
Presented By:
Josef Adriaan Coetsee
read full report
http://dspace.mit.edu/bitstream/handle/1...sequence=1