You are on page 1of 6

term -2c°.

to the nonlinear danping

14.3 Backstepping
We start with the special case of integrator backstepping. Consider the Svst

=
f7) +9(n)
(14.49
(14.50)
where In.T E RT+ is the state and u E R is the control input. The
f:D R" and g: D R" are smooth in a domain functions
0 and f(0) 0. We want to DcR" that contains
=
design a state feedback control law to stabilize the
origin (7 0, f 0). We assume that both and are
= =

be viewed as a cascade f g known. This system can


connection of two components, as shown in
the first
component is (14.49), with as input, and the Figure 14.153
mtegrator (14.50). Suppose the component second component s te
state
feedback control law = o(7), with (14.49) that can be stabilized by a smootu
Ù(0)= 0; is, the origin or
is - fn)+9(n)d(n)
asymptotically
Lyapunov functionstable. Suppose further that we
V(n) that we know a
a (smooth, positive definite
(smooth,
satisfies the P0
av
inequality
We require n)+9(n)oln)) s -W (n), V7ED
(14.51)

sinoothness of all functions


particular problem, we
only need existencefor convenier become clear,
however, that
of
derivatives up
up to a certain order
a
ce
CHAPTER 14. NONLINEAR DESIGN
TOOLS

ti
W(n) is posi nositive definite. By adding and subtracting g(7)6(7) on the right
e
where. n
o f(14.49),
of
we obtain the equivalent
representation
side
h a n d

fn) +9(7)¢(n)+9(7)E- dlr)]

is shown
Figure 14.15(b). The change of variables
which

z=6-d(7)
system
in the
results
=f7)+9(7)b(n)] +9(7)z
u-o
14.15(c). Going from Figure 14.15(b) Figure 14.15(c)
to
in Figure
hich isshown the integrator. Since f, g, and o
viewed as "backstepping" -¢n) through
ean be the expression
are known, the derivative o can be computed by using

)+9tm)k)
connection
reduces the system to the cascade
Taking v u - o
i n) +9(7)d(n)] +9(7):

the first component


we started from, except that now feature will be
which is similar to the system is zero. This
an asymptotically
stable origin when the input
has stabilize the overall system.
Using
to
exploited in the design of v
V.7,)=V(n) + }:2
we obtain
as a Lyapunov function candidate,
aV
Ve=f)
av +9(n)4(n)+9) +
On
aV
-Wn) +97)z +zv

Choosing aV kz, k> 0


V - A 9 ( n ) -

yields Vs-Wn) -kz2


14.3. BACKSTEPPING
591

9n)

(a) fn)

T g(n)

(b) -oln)
)+90))-

g)

-
c)
f)+9()o(O

Figure 14.15: (a) The block diagram of


(c) "backstepping -o(n) system (14.49)-(14.50): (b) introducing
through the integrato.
CHAPTER 14.
NONLINEAR DESIGN TOOLS
that the
whichshows tthat
the
hat
origin (7 =0, z =

0) is
hih sde that
de that the origin (n 0, = asy1nptotically stable. Sinee o(0)-0.
=
0) is
Co obtain the state feedback asymptotically stable. Substituting
e c o n c l

,
and d o,
o, we obtain
we
control law
4,
or

u
df)+g()¬) h7)-k{¢-d(7)] (14.52)
assumptions hold globally and V (7) is radially unbounded,
Fall the a s s u m p t i o n :
we can conclude
that
the origin is globally asymptotically stabke. We summarize our conclusions in
the next lemma.

Temma 14.2 Consider the system (14.49)-(14.50). Let d(n) be a stahilizing state
fedback control lawfor (14.49) with o(0) = 0, and V (n) be a Lyaprunov function that
satisfes (14.51) with some positive definite function W(7). Then, the state feedback
control law (14.52) stabzlizes the origin of (14.49)-(14.50), rith V(7) + [E-oln)"/2
asa Lyapunov function. Moreover, if all the assumptions hold globally and V (n) s
radially unbounded, the origin will be globally asynptotically stable.
Example 14.8 Consider the system

|1 -+

and = T2. We start with


the
the form (14.49)-(14.50) with 7 *1
=

which takes
scalar system

control z2 o(z1)
and proceed to design the feedback
with z2 viewed as the input
to stabilize the origin z1 0. With =

t2=plz1) = -si - z

we cancel the nonlinear term zf to obtain

and V(z,) = ri/2 satisfies


Vz ER
V---afs--i.
stable. To backstep,
exponentially
is globally
Hen
ene, the origin of |
- -

-z1 =

we use the change of variables

2Td(z1) = a2 t71tri
Exampkl3.19.)
nonlinear damping. (Sew
8 because it provides
do not cancel-#}
we
14.3. BACKSTEPPING
593

into the form


the system
to transform
-t1-aj +2
u+(1+ 21)(-z1
-

rj + 2)
g

Taking
we obtain
function,
as a composite Lyapunov

V=n(-z1-ri+z2)+ z2|u +(1+271)(-t1 +za)


-

= - i -i+zalF1 +(1+271)(-a1 - z+22) +u


Taking
u=-1 (1+271)(-t1 - ri+2) -

z2

yields
stable.
Hence, the origin is globally asymptotically A
Application of integrator backstepping in the preceding example is straightíor-
ward due to the simplicity of scalar designs. For higher order systens, we may retain
this simpicity via recursive application of integrator backstepping, as illustrated by
the next example.

Example 14.9 The third-order system

i-ri+T2
C3
3
is composed of the second-order system of the previous example with an additional
integrator at the
input side. We proceed to apply integrator backstepping as n
previous example. After one step of tne
system backstepping, we know that the second-order

-+
=

with zs as input, can be globally stabilized


by the control

and
73--(1+221)(#}-z+a) -

(z2 +1 +)o(1,72)
V(1,a)= }+zl#a+ +
CHAPTER 14. NONLINEAR
DESIGN TOOLS
correspon nonding Lyapunov function. To
backstep,
& the
we
apply the change of
va
ables

Z3 T3O(T1, t2)
o obtain

i=-zit2
i2 O(T1,T2) +z3
z3 u - i +za) (o+ zs)
Ve=V+/2 as a composite Lyapunov function, we obtain
Using

Ve -+z)+ 2 +0)
+3 u-(-7+z) - +
- - (2 +a1 +"
Ov +)+u
+ ( - ++7)-
Taking OV
( - +zo) +zs+o)- z3
yields
Ve--zi-(7a +z1+r -
A
asymptotically stable.
Hence, the origin is globally
the general system
from (14.49)(14.50) to
more
Let us move now

= f)+9n)E
(14.53)
fal7.8) + 9a(7,£)u
(14.54)
E =

the input
0 the domain of interest,
smooth. If ga(n, 5) over
where fa and ga are
transformation 1 (14.55)
Sa7,6)
if a stabilizing state feedback
ua. Therefore,
the conditions of
=

(14.54) to the integrator


will reduce
function exist such that
V(n)
and a Lyapunov and (14.55) yield
rO! law o(n) satisfied for (14.53), then the lemma
Lenma 14.2 are

=de(7.¬) (14.56)
+o(n)1-ot)--dn))- J(n.¬0
18n)
f(n)
9.(76)

You might also like