National Instruments 370757C-01 用户手册

The four transfer matrices are labeled e/d, e/n, u/d, and u/n in the final plot. 
The plots in the top row, consisting of e/d and e/n, show the regulation or 
tracking achieved by the controller. If both these quantities are small, then 
the disturbance d and the sensor noise n will not make the error signal e 
large. 

The bottom row of plots, consisting of u/d and u/n, show the actuator effort 
used by the controller. If these are both small, then the actuator effort u, 
which results from the disturbance d and the sensor noise n will be small.

A classic trade-off in controller design boils down to a choice between 
making the top row of a 

perfplot( )

 small (good regulation/tracking) 

and making the bottom row small (low actuator effort). For example, by 
varying the design parameter 

ρ in the 

lqgltr( )

 regulator design process, 

the magnitude of the top two transfer matrices can be traded off against the 
magnitude of the bottom two. Increasing 

ρ makes the top two magnitudes 

smaller but makes the bottom two larger. 

The columns of a 

perfplot( )

 have a dual interpretation. The plots in the 

left column, e/d and u/d, show how sensitive the system is to the process 
noise or disturbance d. The plots in the right column, e/n and 

u

/n, show how 

sensitive the system is to the sensor noise n. Again, there is a trade-off 
between making the magnitudes of the transfer matrices on the left small 
(good disturbance rejection) and making the magnitudes of the transfer 
matrices on the right small (low sensitivity to sensor noise). In the 

lqgltr( )

estimator design, the parameter 

ρ controls the relative 

magnitude of the left and right plots. Increasing 

ρ makes the left two 

magnitudes smaller but makes the right two larger. Refer to Example 3-3.

Consider the simple closed-loop system shown in Figure 3-4. 

Figure 3-4.  Closed-Loop System

disturbance

e

1

s

+

+

+

–

K

n

noise