Control Lability
Control Lability
i1
A
i
(T t)
i
i!
_
_
B = 0
implies that v
/
B = 0, v
/
A
i
B = 0 pour tout i 1 or v
/
C = 0 with
v ,= 0, thus :
rank C < n.
Conversely, if rank C < n, v ,= 0 such that
v
/
C = 0.
Thus, according to what precedes, v
/
A
i
B = 0 for i = 0, . . . , n 1.
By Cayley-Hamilton Theorem, A
n+i
, i 0, is a linear combination
of the A
j
s, j = 1, . . . , n 1.
Thus v
/
A
i
B = 0, i 0 or v
/
G = 0, which proves the Theorem.
Examples
x
1
= x
2
, x
2
= u
is controllable : B =
_
0
1
_
, A =
_
0 1
0 0
_
,
C =
_
0 1
1 0
_
and rank C = 2.
x
1
= u , x
2
= u
isnt controllable : B =
_
1
1
_
, A =
_
0 0
0 0
_
,
C =
_
1 0
1 0
_
and rank C = 1.
Note : x
1
x
2
: 0, or x
1
x
2
= Cste : relation between states
independent of u.
1.2. Controllability Canonical Form
D enition Two systems
x = Ax + Bu , z = Fz + Gv
are said equivalent by change of coordinates and feedback (we note
(A, B) (F, G)) iff there exist invertible matrices M and L and a
matrix K such that
_
x = Ax + Bu
z = Mx, v = Kx + Lu
= z = Fz + Gv
and conversely.
M : change of coordinates, invertible of size n n, and K and L :
feedback gains, with L invertible of size mm and K of size mn.
Changes of coordinates preserve the state dimension and (non dege-
nerate) feedbacks preserve the input dimension.
The relation is an equivalence relation :
reexive : (A, B) (A, B) (M = I
n
, K = 0 and L = I
m
)
symmetric : x = M
1
z and u = L
1
KM
1
z + L
1
v
transitive : if (A, B) (F, G) and (F, G) (H, J), z = Mx, v =
Kx + Lu imply z = Fz + Gv and s = Tz, w = Nz + Pv imply
s = Hs + Jw. Thus s = TMx, w = (NM + PK)x + PLu.
Note that
F = M(A BL
1
K)M
1
, G = MBL
1
.
The Single Input Case
x = Ax + bu
with (A, b) controllable :
rank C = rank
_
b Ab . . . A
n1
b
_
= n.
One can construct the matrices M, K and L that transform the system
in its canonical form,
z = Fz + gv
F =
_
_
_
_
_
_
0 1 0 . . . 0
0 0 1 0
.
.
.
.
.
.
.
.
.
0 0 0 1
0 0 0 . . . 0
_
_
_
_
_
_
, g =
_
_
_
_
_
_
0
0
.
.
.
0
1
_
_
_
_
_
_
or
z
1
= z
2
, . . . , z
n1
= z
n
, z
n
= v.
The Multi Input Case (m > 1) :
x = Ax + Bu , B =
_
b
1
. . . b
n
_
.
The controllability matrix
C =
_
b
1
Ab
1
. . . A
n1
b
1
. . . b
m
Ab
m
. . . A
n1
b
m
_
has rank n and one can construct a sequence of integers n
1
, . . . , n
m
called controllability indices such that
n
1
+ . . . + n
m
= n
with
C =
_
b
1
Ab
1
. . . A
n
1
1
b
1
. . . b
m
Ab
m
. . . A
n
m
1
b
m
_
of size n n, invertible.
The controllability indices n
1
, . . . , n
m
exist for all controllable linear
systems, are dened up to permutation and are invariant by change
of coordinates and feedback.
Theorem (Brunovsky) : Every linear system with n states and m
inputs is equivalent by change of coordinates and feedback to the
canonical form
F = diagF
1
, . . . , F
m
, G = diagg
1
, . . . , g
m
i=0
a
i
_
t
T
_
i
.
with a
0
, . . . , a
2n+1
computed from the successive derivatives of y
ref
at times 0 and T :
y
(k)
ref
(t) =
1
T
k
2n+1
i=k
i(i 1) (i k + 1)a
i
_
t
T
_
ik
At t = 0 :
y
ref
(0) = a
0
y
(k)
ref
(0) =
k!
T
k
a
k
, k = 1, . . . , n 1,
v
ref
(0) =
n!
T
n
a
n
and at t = T :
y
ref
(T) =
2n+1
i=0
a
i
,
y
(k)
ref
(T) =
1
T
k
2n+1
i=k
i!
(i k)!
a
i
, k = 1, . . . , n 1,
v
ref
(T) =
1
T
n
2n+1
i=n
i!
(i n)!
a
i
Thus
a
0
= y
ref
(0) , a
k
=
T
k
k!
y
(k)
ref
(0) , k = 1, . . . , n 1, a
n
=
T
n
n!
v
ref
(0).
_
_
_
_
_
_
_
1 1 . . . 1
n + 1 n + 2 2n + 1
(n + 1)n (n + 2)(n + 1) (2n + 1)2n
.
.
.
.
.
.
(n + 1)!
(n+2)!
2
. . .
(2n+1)!
(n+1)!
_
_
_
_
_
_
_
_
_
a
n+1
.
.
.
a
2n+1
_
_
=
_
_
_
_
_
_
_
y
ref
(T)
n
i=0
a
i
.
.
.
T
k
y
(k)
ref
(T)
n
i=k
i!
(ik)!
a
i
.
.
.
T
n
v
ref
(T) n!a
n
_
_
_
_
_
_
_
1.4. Trajectory Tracking, Pole Placement
Assume that the state x is measured at every time.
We want to follow the reference trajectory y
ref
, the system being
perturbed by non modelled disturbances. Note e = y y
ref
the de-
viation between the measured trajectory and its reference. We have
e
(n)
= v v
ref
.
Note v v
ref
=
n1
i=0
K
i
e
(i)
, the gains K
i
being arbitrary. Thus
e
(n)
=
n1
i=0
K
i
e
(i)
or
_
_
e
.
.
.
e
(n)
_
_
=
_
_
_
_
_
_
0 1 0 . . . 0
0 0 1 0
.
.
.
.
.
.
.
.
.
0 0 0 1
K
0
K
1
K
2
. . . K
n1
_
_
_
_
_
_
_
_
_
_
e
e
.
.
.
e
(n1)
_
_
_
_
.
the gains K
i
are the coefcients of the characteristic polynomial of
the closed-loop matrix A + BK.
Theorem
If the system x = Ax+Bu is controllable, the eigenvalues of A+BK
may be placed arbitrarily in the complex plane by a suitable feedback
u = Kx.
Corollary
A controllable linear system is stabilizable and, by state feedback, all
its characteristic exponents can be arbitrarily chosen.
2. First Order Controllability of Nonlinear
Systems
Consider the nonlinear system
x = f(x, u)
with x X, n-dimensional manifold, and u R
m
.
Its tangent linear system around the equilibrium point ( x, u) is given
by
= A + Bv
with A =
f
x
( x, u), B =
f
u
( x, u).
Denition We say that a nonlinear system is rst order control-
lable around an equilibrium point ( x, u) if its tangent linear system at
( x, u) is controllable, i.e. iff rank C = n, with C =
_
B[AB[ [A
n1
B
_
.
Denition We say that a nonlinear system is locally controllable
around an equilibrium point ( x, u) if :
for all > 0, there exists > 0 such that for every pair of points
(x
0
, x
1
) R
n
R
n
satisfying |x
0
x| < and |x
1
x| < , there
exists a piecewise continuous control u on [0, ] such that | u(t)| <
t [0, ] and X
(x
0
, u) = x
1
, where X
(x
0
, u) is the integral curve
at time , generated from x
0
at time 0 with the control u.
Theorem If a nonlinear system is rst-order controllable at the equi-
librium point ( x, u), it is locally controllable at ( x, u).
Remark
The scalar system : x = u
3
is locally controllable but not rst-order
controllable.
To join x
0
and x
1
in duration T = , x
0
and x
1
arbitrarily chosen
close to 0, one can use the motion planning approach :
x(t) = x
0
+ (x
1
x
0
)
_
t
_
2
_
3 2
t
_
.
Thus u(t) =
_
x(t)
_1
3
=
_
6
_
x
1
x
0
_
_
t
_ _
1
t
_
_1
3
.
One easily checks that if [x
0
[ < and [x
1
[ < with <
4
3
then
[u(t)[ < , which proves the local controllability. On the contrary, at
the equilibrium point (0, 0), the tangent linear system is x = 0, and is
indeed not rst-order controllable.
3. Local Controllability and Lie Brackets
For simplicitys sake, the system is assumed afne in the control, i.e.
x = f
0
(x) +
m
i=1
u
i
f
i
(x)
with f
0
(0) = 0, (( x, u) = (0, 0) is an equilibrium point).
From the vector elds f
0
, . . . , f
m
, we construct the sequence of dis-
tributions :
D
0
= spanf
1
, . . . , f
m
, D
i+1
= [f
0
, D
i
] + D
i
, i 1
where D
i
is the involutive closure of the distribution D
i
.
Proposition 3.1 The sequence of distributions D
i
is non decreasing,
i.e. D
i
D
i+1
for all i, and there exists an integer k
and an invo-
lutive distribution D
such that D
k
= D
k
+r
= D
for all r 0.
Moreover, D
(ii) [f
0
, D
] D
.
Proof D
i
[f
0
, D
i
] + D
i
= D
i+1
D
i+1
.
Thus, there exists a largest D
, with D
= D
. But :
rank D
i+1
rank D
i
+ 1 for small i = 0, 1, . . ..
thus, there exists k
n such that D
k
= D
k
+r
= D
.
Moreover, D
0
D
: (i),
and D
= [f
0
, D
] + D
implies [f
0
, D
] D
: (ii).
Proposition 3.2 Let Dbe involutive with constant rank equal to k in
an open U, satisfying (ii). There exists a diffeomorphism such that,
if we note :
_
i
=
i
(x), i = 1, . . . , k
j
=
k+j
(x), j = 1, . . . , n k
we have :
f
0
(, ) =
k
i=1
i
(, )
i
+
nk
i=1
k+i
()
i
where the
i
s are C
functions.
Proof We have
f
0
(, ) =
k
i=1
i
(, )
i
+
nk
i=1
k+i
(, )
i
.
But, by Frobenius Theorem, D = span
1
, . . . ,
k
.
By (ii), we have [
f
0
,
i
] D for all i = 1, . . . , k. But :
[
f
0
,
i
] =
k
j=1
_
j
[
j
,
i
]
j
j
_
+
nk
j=1
_
k+j
[
j
,
i
]
k+j
j
_
=
k
j=1
nk
j=1
k+j
j
D
Thus
k+j
i
= 0 for all i, j, or
k+j
depends only of , which proves
the Proposition.
Theorem Let D
(x) = n , x U
where U is a neighborhood of the origin.
Proof By contradiction. Assume that D
(x) = k < n , x U
Using (i), the image by of f
i
, i = 1, . . . , m, is
f
i
=
k
j=1
i,j
j
.
Therefore,
_
_
f
0
+
m
i=1
u
i
f
i
_
_
= (
f
0
) +
m
i=1
u
i
(
f
i
)
=
k
j=1
_
_
j
(, ) +
m
i=1
u
i
i,j
(, )
_
_
j
+
nk
j=1
k+j
()
j
.
In other words, in these coordinates, the system reads
_
j
=
j
(, ) +
m
i=1
u
i
i,j
(, ), j = 1, . . . , k
j
=
k+j
(), j = 1, . . . , n k
and the part is not controllable, which achieves the proof.
Denote by Lief
0
, . . . , f
m
the Lie algebra generated by the linear
combinations of f
0
, . . . , f
m
and all their Lie brackets.
Lief
0
, . . . , f
m
(x) is the vector space generated by the vectors of
Lief
0
, . . . , f
m
at the point x.
By construction,
D
Lief
0
, . . . , f
m
= Lief
1
, . . . , f
m
. Thus, if f
1
, . . . , f
m
are analytic, local controllability is equivalent to
rank D
= n
in some open set.
Example
x
1
= x
2
2
, x
2
= u
Equilibrium point (x
1
, 0), x
1
arbitrary.
D
0
= spanf
1
= span
x
2
and D
0
= D
0
;
D
1
= spanf
1
, [f
0
, f
1
] = span
x
2
, x
2
x
1
, has rank 2 and is invo-
lutive if x
2
,= 0.
But [f
1
, [f
0
, f
1
]] = 2
x
1
D
1
(x
1
, 0), thus
D
= D
1
(x
1
, 0) = span
x
2
,
x
1
.
However, if x
1
> 0, one cannot reach points such that x
1
< 0. Thus
the rank condition is necessary but not sufcient for local control-
lability !
Remark For general systems :
x = f(x, u)
one can use the previous formalism by setting
u
i
= v
i
, i = 1, . . . , m
since
x = f(x, u)
u
1
= v
1
.
.
.
u
m
= v
m
has the required afne form.
Remark The rank condition for linear systems :
x = Ax +
m
i=1
u
i
b
i
Set
f
0
(x) = Ax =
n
i=1
_
n
j=1
A
i,j
x
j
_
x
i
and
f
i
(x) = b
i
=
n
j=1
b
i,j
x
j
, i = 1, . . . , m.
D
0
= spanb
1
, . . . , b
m
. D
0
= D
0
since [b
i
, b
j
] = 0 for all i, j.
[Ax, b
i
] =
n
j=1
b
i,j
x
j
_
_
n
k=1
n
l=1
A
k,l
x
l
x
k
_
_
=
n
j,k=1
b
i,j
A
k,j
x
k
j=1
(Ab
i
)
j
x
j
= Ab
i
.
Thus
D
1
= [Ax, D
0
] + D
0
= spanb
1
, . . . , b
m
, Ab
1
, . . . , Ab
m
.
and D
1
involutive (made of constant vector elds).
We have
D
k
= spanb
1
, . . . , b
m
, Ab
1
, . . . , Ab
m
, . . . , A
k
b
1
, . . . , A
k
b
m
for all k 1.
There exists k
= D
and rank D
i=1
k
i
j=1
a
i,j
d
j
x
i
dt
j
=
n
i=1
k
i
j=1
a
i,j
x
(j)
i
with k
1
, . . . , k
n
arbitrary integers. Mis a nitely generated K-module.
Let K be the eld of meromorphic functions of t, K = K[
d
dt
] and let
x
1
, . . . , x
n
be a basis of R
n
. Let Mbe made of all vectors of the
form
n
i=1
k
i
j=1
a
i,j
(t)x
(j)
i
with k
1
, . . . , k
n
arbitrary integers and the a
i,j
s meromorphic func-
tions. Then Mis a nitely generated K-module.
Let M be a K-module. An element m M, m ,= 0, is said to be
torsion if there exists K, ,= 0, such that m = 0.
We denote by T the set of all torsion elements of M.
It is obviously a submodule of M, called the torsion submodule.
We say that a K-module F is free if and only if for every m F,
m ,= 0, m = 0, with K, implies = 0.
We have
Proposition 1 Let K = K[
d
dt
] where K is a eld.
(i) A nitely generated module M can be uniquely decomposed into
a torsion sub-module T and a free sub-module F : M = T F.
(ii) It is free if and only if T = 0.
4.2. Linear Systems
Let K = K[
d
dt
], where K is a eld, and M a nitely generated K-
module.
Let A() be a (n m) n polynomial matrix of =
d
dt
with coef-
cients in K. We consider the linear system in M
A()x = 0.
The quotient of Mby the lines of this system is again a nitely gene-
rated K-module.
Therefore, the data of a linear system is equivalent to the one of a
nitely generated K-module.
Denition. (Fliess, 1990) Given a nitely generated K-module M,
we say that the associated linear system is controllable if and only if
Mis free.
Example
Consider the n-dimensional time-varying linear system
x = F(t)x + G(t)u
with u R
m
, F and G meromorphic w.r.t. t in a given open interval
of R, and rank G(t) = m for every t in this interval.
It reads (
d
dt
I F(t))x = G(t)u.
Let C be a (nm)n matrix such that C(t)B(t) = 0 with rank C(t) =
n m for every t.
We have
A(
d
dt
)x
def
= C(t)(
d
dt
I F(t))x = 0
thus A(
d
dt
) = C(t)(
d
dt
IF(t)) is a (nm)n matrix with coefcients
in K = K[
d
dt
], with K the eld of meromorphic functions of t.
The linear system is equivalent to the set of linear equations
A(
d
dt
)x = 0.
We denote by Mthe nitely generated K-module generated by a ba-
sis of R
n
and M
0
its quotient submodule generated by the lines of
A(
d
dt
)x = 0.
Assume that the pair (F, G) is not controllable. By Kalmans decom-
position, there exists a k-dimensional subspace, described by the set
of n k independent equations
K(t)x = 0
such that the n k-dimensional quotient subsystem is controllable.
Clearly, K K and the set x[Kx = 0 is the torsion submodule of
M
0
which is thus not free.
Consequently, M
0
is free if the pair (F, G) is controllable in the usual
sense.
The converse follows the same lines.
Remarks.
This denition is independent of the system variables description.
For 1st order controllable nonlinear systems, the tangent linear ap-
proximation at any point constitutes a linear time-varying system
whose module is free.