National Instruments 370757C-01 用户手册

The transfer matrix G can be viewed as a model of the underlying system 
dynamics with v and u as generalized forces that produce effects in the 
performance signals z and measured signals y. 

The weight W

 is used to model the exogenous input v by v = W

w.

Similarly, the critical performance variables in the vector z are weighted to 
form the normal critical variables e = W

z.

In general, the input weight W

 can be viewed as a dynamic model of the 

exogenous inputs and the output weight W

 as the inverse of the desired 

performance. As an illustration, consider the plant configuration in 
Figure 4-3. 

Figure 4-3.  Typical Plant Configuration

The exogenous input vectors d and n represent disturbances and sensor 
noise, respectively. These are generated by passing normalized 
unpredictable signals, 

ω

 and 

ω

, through stable transfer matrices, 

and W

, respectively. The critical performance variables are some 

regulated variables y

, as well as the actuator commands u. These are 

weighed by the transfer matrices W

 and W

act 

to form the normalized error 

variables e

 and e

. The sensed variables y

 are contaminated by 

additive noise n to form the measured signal y. The transfer matrix G

 

represents the underlying system dynamics. Observe that the transfer 
matrix G, as defined in [BBK88], consists of G

 with some special 

output/input connections among the variables n and u as depicted in 
Figure 4-3. This is in the form of the familiar LQG setup, except that 

w

d

n

y

reg

y

sens

u

e

u

y

P

W

dist

w

dist

W

in

W

reg

W

out

w

e

W

noise

w

noise

e

reg

e

act

W

act

G

dyn

G