You are on page 1of 2

AA599: Geometric Methods for Nonlinear Control Systems Lecture Notes Lie Brackets

Kristi A. Morgansen Department of Aeronautics and Astronautics March 30, 2006

Lie brackets and switching controls


f x x

For a system of the form x = f (x) (where x = any point a is

f f (x)), x

the Taylor expansion about

1 x(t) = x(a) + x(a)(x a) + x (x a)2 + h.o.t. 2! f t2 = x(0) + f (0)t + f (0) + h.o.t. x x=0 2 where we have chosen a = 0 for simplicity. Now consider switching controls for the system x = f1 (x)u1 + f2 (x)u2 of the form u1 u1 u1 u1 = 1, u2 = 0, 0 t < = 0, u2 = 1, t < 2 = 1, u2 = 0, 2 t < 3 = 0, u2 = 1, 3 t < 4 .

Using the Taylor expansion, the system response at time t = is x( ) = x0 + f1 (x0 ) +


21

f1 f1 (x0 ) + h.o.t 2 x

where we will be assuming that 1 and neglect all terms of order more than two in . We will also suppress the evaluation notation on derivatives for brevity. At time t = 2 , we use the end point of x( ) as the initial condition for another evolution over time to get x(2 ) = x( ) + f2 (x( )) +
21

f2 f2 (x( )) + h.o.t 2 x

= x0 + f1 (x0 ) +

f1 f1 (x0 ) 2 x 1 f1 f1 (x0 ) + f2 x0 + f1 (x0 ) + 2 2 x 1 f1 1 f2 f2 x0 + f1 (x0 ) + 2 f1 (x0 ) + h.o.t. +2 2 x 2 x

21

Now note that we can use the Taylor expansion again to rewrite the term f2 : f2 x0 + f1 (x0 ) +
21

f1 f2 f1 (x0 ) = f2 (x0 ) + 2 x x
3

f1 (x0 ) +

21

f1 f1 (x0 ) 2 x

which gives (neglecting terms of order x(2 ) = x(0) + f1 (x0 ) +


21

and higher)
2 f2

f1 f1 (x0 ) + f2 (x0 ) + 2 x
(f ) x

f1 (x0 ) +

21

f2 f2 (x0 ). 2 x

Now reverse the rst control (note that x(3 ) = x(2 ) f1 (x(2 )) + = x(0) + f1 (x0 ) + f1 (x0 ) + f1 x

= f ), and repeat the above process to get x

= x(0) + f2 (x0 ) + After the nal step, we have

f1 f1 (x(2 )) + h.o.t 2 x f2 1 f2 2 1 f1 f1 (x0 ) + f2 (x0 ) + 2 f1 (x0 ) + 2 f2 (x0 ) 2 x x 2 x 1 f1 (f1 (x0 ) + f2 (x0 )) + 2 f1 (x0 ) 2 x 1 f2 f1 2 f2 f1 (x0 ) + 2 f2 (x0 ) 2 f2 (x0 ). x 2 x x

21

x(4 ) = x(3 ) f2 (x(3 )) +

f2 f2 (x(3 )) + h.o.t 2 x 1 f2 f2 f1 (x0 ) + 2 f2 (x0 ) = x(0) + f2 (x0 ) + 2 x 2 x 1 f2 f2 f2 (x0 ) + 2 f2 (x0 ) f2 (x0 ) + x 2 x f2 f1 f1 (x0 ) f2 (x0 ) = x0 + 2 x x = x0 + 2 [f1 , f2 ] (x0 ).

21

2 f1

f2 (x0 )

Thus one can consider the Lie bracket of two vector elds to be the result of switching at high frequency between the two.

You might also like