Generalized Digital Redesign Method for Linear
Feedback System Based on N – Delay Control
Objective:
The Objective of this project to design digital state
feedback gain matrix, Kd and feed forward gain matrix, Ed
using Digital Redesign approach based on N-Delay control.
What is Digital Redesign?
Digital Redesign is a controller design Technique. The
approach converts a satisfactorily designed continuous -time
controller into an equivalent digital controller retaining
completely the properties (performance) of the analog
system.
Features of Proposed Methods
•The N-delay control is employed, and the lth plant input
is changed Nl times during one sampling period.
•The states of the redesigned sampled – data system
completely match those of the original continuous-time
closed-loop system at every sampling period.
•The proposed redesign method can be applicable for a
static state-feedback or a dynamic controller.
I. PROBLEM OF DIGITAL REDESIGN BY COMPLETE STATE MATCHING
Consider the linear continuous time system
o
X (t) = A C X(t) + BC u(t) (1)
The continuous - time state - feedback control law be
u(t) = f C x(t)
(2)
The continuous - time closed – loop system becomes
o
X (t) = (A C + BC f C )x(t) (3)
Sampled – data system with the sampling period T is
(A C + bC f C )(T)
x((i+1)T) = e x(iT) (4)
o
y (t)
u(t) x(t ) = A C x(t) + b C u(t)
y(t) = C C x(t)
+ x(t)
fC
Fig. 1. Continuous – time state – feedback control system.
u[i] y(t)
u(t) o
Z. O. Hold x(t ) = A C x(t) + b C u(t)
y(t) = C C x(t)
+
T x(t)
x[i]
f (T)
Sampler
Fig. 2. Discrete-time state-feedback control system
The sampled - data system for the sampling period T becomes
_ _ _ _
X c ((i+1)T) = A X c (iT) + B r(iT) (17)
where
_ _
_
_ _
A A T
B1
A = e AC T = _
11 12 _ _ _
B = e AC
B c d = _
_
21
A A 22
0
B2
Definition 1: Generalized controllability indexes of (Acp, Bcp). If (Acp, Bcp)
is a controllable pair, np linearly independent vectors can be
selected from
b c1 ,..., b cmp , A cp bc1 , ...,A cp bcmp ,...,A cp np-1bcmp
where
Bcp = [bc1 ,..., bcmp ] be a set of these np vectors, l
l = number k/A k-1cp bcl
The discrete-time plant using N-Delay input control is given by
mp Nl
X dp [i+1] = A dp X dp [i] + B lj u l j [i ]
l 1 j1 (18)
where
(1-μl(j-1)T) A
A cp T
A dp = e Blj = Cp
e Bcld
(1-μlj)T
Controller Plant
r uck
yck + ucp
Gck + o
X ck = A ck X ck + Bck u ck
o
X cp = A cp X cp + Bcp u cp ycp
Gcp
y ck = Cck X ck + Dck u ck y cp = C cp X cp
+
Xcp
Fcp
Fck
Fig. 3. Continuous – time control system
Controller Plant
r ulj[i]
X dk [i+1] = L1X dp +L 2 X dk +L3 r o ydp(t)
HOLD X dp = A cp X dp +ycpBcp u lj
u lj = FljX dp +G ljX dk +H ljr
y dp = Ccp X dp
+
Xdp(t)p
Xdp[i]
Fig. 4. Discrete – time control system
The dynamic of the discrete - time controller be
X dk [i+1] = L1X dp [i] +L 2 X dk [i] +L 3 r[i] (19)
The discrete - time control law be
u lj [i] = Flj X dp [i] +G lj X dk [i] +H lj r[i] (20)
The discrete - time closed - loop augmented system is represented by
_ _ _ _
X d [i+1] = A d X d [i] + Bd r[i] (21)
where
_
A +BF BG
_ BH
A d = dp Bd =
L1 L 2 L
3
B = [b11 ,...,b1 N1 ,b 21 ,...,b mp N mp ]
F = [f11T ,...,f1T N1 ,f 21T ,...,f mp T N mp ]T
G = [g11T ,...,g1T N1 ,g 21T ,...,g mp T N mp ]T
H = [h11T ,...,h1T N1 ,h 21T ,...,h mp T N mp ]T
N = N1 +N 2 + ........+N mp n p
Comparing (17) and (21) the states of the digitally controlled system
completely match the states of continuous-time system
_ _
_
A11 A 12
= A dp +BF BG B1 = BH
L1 L 2 _ L (b)
_ _ (a) 3
A 21 A 22 B2
The necessary and sufficient condition for solution of the linear matrix
equation is
rank A = rank [A,b].
The necessary and sufficient condition for the existence of F, G, H in (a)
and (b) are given by
_
rank B = rank[B, A11 - A dp ]
_ _
= rank[B, A12 ] = rank [B, B1 ]
l Is the generalized controllability index
Theorem1: Let (Acp, Bcp)be a controllable pair .If the input multiplicities
satisfy Nl 1
The matrix B has full row rank,
rank B = np
The existence of F,G,H in (a) and (b), the redesigned parameters are
given by
_
-
F = B ( A11 - A dp )
_
-
G = B A12
_
-
H = B B1
_
L1 A12
_
L 2 A 22
_
L3 B2
ucp
Plant
ycp
o
X cp = A cp X cp + Bcp u cp
ycp = Ccp X cp
+
o
X cp = A cp X cp (t) + Bcp u cp (t) +
K cp (y cp - Ccp X cp )
Fcp
Observer
X cp
Fig. 5. Continuous time observer
The sampled data system for the sampling period T is represented by
ecp ((i+1)T) = e
(A cp -K cp Ccp )T
ecp (iT) (24)
The multirate output sampling control is employed, in which the qth plant
output is detected Mq times during one sampling period
The selections of output multiplices Mq are made by the following condition.
M q q (25)
where ( 1,... Pp ) are the set of generelized observability indexes of (Acp, Ccp)
Definition 2: Generalized observability indexes of (Acp, Ccp) are defined as follows
If (Acp, Ccp) is an observable pair, np linearly independent vectors can be
selected from
Cc1 ,..., Ccpp , Cc1A cp , ...,Ccpp A cp ,...,Ccpp A np-1cp
where
be a set of these np vectors, q
Ccp = [Cc1T ,..., Ccpp T ]
q = number K/Ccq A cp k-1
The discrete - time plant using N - delay input control and M - delay output
control is given by
mp Nl
X dp [i+1] = A d P X dP [i] + B l=1 j=1
lj u ij [i] (26)
mp Nl
(27)
yqk [i] = Cqk X dp [i] + D
l=1 j=1
qklj u ij [i]
where
Cqk = Ccq e AcpvqkT
lj<v : C (vqk - l(j-1)T e Acp B dτ
cq (v - lj)T
qk
qk
cl
(vqk - l(j-1)T
l(j-1)<v :C e Bcl dτ
cq 0
Acp
Dqklj = qk lj
v l(j-1): 0,
qk
0 vq1 <v q2 <....<vq M q <1
Using the intersampling outputs yqk, the discrete-time observer be
mp Nl m p Nl
X dp [i+1] = A d P X dP [i] + Blju ij[i]K qk y qk [i]-(C qk X dp [i] D qklju lj[i]
l=1 j=1 l 1 j 1 (28)
The estimation error of the discrete-time state edp is represented by
edp [i+1] = X dP [i+1] - X dp [i+1] = (A dp -KC)e dp [i] (29)