Hi could some expert recommend a Liapunov function that might help control this plant represented by a nonlinear differential equation: y(t) = x(t) * ( 1 - F(s) * (|x(t)|^2) ) here F(s) = K/(1+sb), represents a 1st order low pass filter (convolution with Kexp(-bt)) with time constant 1/b, and K is a "small" positive number, so y is nearly x but not exactly; both b and K are unknown although I have a good idea of their value within a factor of 2. ( By F(s) * (|x|^2) is meant the square of x(t), x(t)^2, is low pass filtered by F(s).) I need a control law x = C[u, y],for positive u>0, such that the tracking error |y-u| is small. Thanks
nonlinear headache
Started by ●September 20, 2003
Reply by ●September 25, 20032003-09-25
robert egri wrote:> Hi > > could some expert recommend a Liapunov function that might help > control this plant represented by a nonlinear differential equation: > > y(t) = x(t) * ( 1 - F(s) * (|x(t)|^2) ) > > here F(s) = K/(1+sb), represents a 1st order low pass filter > (convolution with Kexp(-bt)) with time constant 1/b, and K is a > "small" positive number, so y is nearly x but not exactly; both b and > K are unknown although I have a good idea of their value within a > factor of 2. > > ( By F(s) * (|x|^2) is meant the square of x(t), x(t)^2, is low pass > filtered by F(s).) > > I need a control law x = C[u, y],for positive u>0, such that the > tracking error |y-u| is small. > > ThanksYou are mixing Laplace notation with time notation. Tom
Reply by ●October 16, 20032003-10-16
rge11x@netscape.net (robert egri) wrote:>Hi > >could some expert recommend a Liapunov function that might help >control this plant represented by a nonlinear differential equation: > >y(t) = x(t) * ( 1 - F(s) * (|x(t)|^2) ) > >here F(s) = K/(1+sb), represents a 1st order low pass filter >(convolution with Kexp(-bt)) with time constant 1/b, and K is a >"small" positive number, so y is nearly x but not exactly; both b and >K are unknown although I have a good idea of their value within a >factor of 2. > >( By F(s) * (|x|^2) is meant the square of x(t), x(t)^2, is low pass >filtered by F(s).) > >I need a control law x = C[u, y],for positive u>0, such that the >tracking error |y-u| is small. > >ThanksI think what you said was dx/dt = x - F x^3 Is this what you mean? If so, you get dz/dt = (-z + F )/2 with z = 1/x^2 This equation should be much easier to work with.