Beruflich Dokumente
Kultur Dokumente
Mrida, Yucatn,
CONTENT
INTRODUCTION
REFERENCES
MODEL DESCRIPTION
CONCLUSIONS
INTRODUCTION
-Servomechanism
- Open loop vs Closed loop
- Identification
- PBC
REFERENCES
Y. Zhou, A. Han, S. Yan, X. Chen, "A fast method for online closed loop
system identification", The International Journal of Advanced Manufacturing
Technology, vol. 31, no. 1-2, pp. 78-84, 2006.
MODEL DESCRIPTION
ALGORITHM
Jq(t ) fq (t ) ku (t ) v1 (t ) (t ) v1 (t )
where
J Inertia
f Viscous friction coefficient
k Amplifier gain
u Input voltage
v1 Perturbation signal
Torque input
q(t ) aq (t ) bu (t ) v(t )
where : a f , b k , v v
1
(1)
(2)
u (t ) k p e(t ) k d q
(t )
where :
e(t ) qd (t ) q (t )
(3)
qd reference
The real servomechanism in closed loop with u (t ) leads to the closed loop
dynamics:
: q(t ) (a bk
1
) q (t ) bk p e(t )
(4)
where :
(a bk d )
Using Routh- Hurwitz criterion, the control law u (t ) stabilizes
q
e a q e (t ) bue (t )
(5)
ue (t ) k p ee (t ) kd q e (t )
where :
ee (t ) qd (t ) qe (t )
(6)
The estimated model of 1 in closed loop with ue (t ) leads to the closed loop
dynamics:
:q
(7)
e (a bk d )q e (t ) bk p ee (t )
is stable
Error dynamics
Let output error between the real servomechanism and its model be given as:
(t ) q qe
(8)
Evaluating its second time derivate and employing the real servomechanism
and the model estimated can be established the error dynamics as follows:
~
(t ) c bk p (t ) T (t )
where :
.
~ parameter error
regressor vector
q e
a a
~
(t )
, (t )
b b
ue
(9)
Stability analisys
x ( , )T
be positive definite if
bk p
(10)
then the time derivate of H1along the trajectories of the dynamics error:
yields
H 1 ( , ) ~ T ( ) c / 2( ) 2
(11)
therefore, defines an output strictly passive map (OSP)
(~ T ) ( )
(12)
Its intuitive to consider the parameter error dynamics as the following gradient
like parameter updating law:
~
3 : ( )
where : T M 2 X 2
(13)
(t ) c (t ) bk p (t ) ~ T (t )
~
( )
(14)
(t ), (t ),~(t )
Exponential convergence
(t ) c (t ) bk p (t ) ~ T (t ) then the system (A, B, C) is
Considerating
~ ( )
~ 0
4 : y T (t )~(t )
2
~ ( )
5 : y T (t )~(t )
2
Theorem: Let the state space description of the error dynamics and assume that the
following inequalities hold:
4bk p c
1 2bk p
min bk p , ,
,
2c c
4bk p c 2
Let w ( x, ~ ) be an equilibrium point of the system. Then w is an exponentially
stable equilibrium point of the system.
: q(t ) (a bk )q (t ) bk e(t )
2 : qe (a bkd )q e (t ) bk p ee (t )
~ ( )
:
3
d
~T
(t ) c (t ) bk p (t ) (t )
~
( )
CLOSED-LOOP IDENTIFICATION:
PERTURBED CASE
~
T
Theorem: If the equilibrium w 0 of the error dynamics (t ) c bk p (t ) (t )
is exponentially stable, then:
L stable
(15)
(bk p c 1) (bk p c 1) 2 4 2
so that:
min (M 2 )
(b k p c 1) - (bk p c - 1) 2 4 2
2
max (M 2 )
(b k p c 1) (bk p c - 1) 2 4 2
2
CONCLUSION
Its important to pay attention not only to the values of the gain matrix
k dgains and
. and ,but also to the value of the controllers
kp
.
Future work can be to design more complicated controllers and ensures the
overall exponential stability.
THANKS !!!