N. P. THOMPSON, B. WIDROW AND C. SCHADE 
1213 
3/2 5/71 
ELAPSED TIME = 
6 2 
MINUTES 
SICK DOG 
AUTOMATIC CONTROL 
PHASE JE 
3/2 5/71 SICK DOG 
ELAPSED TIME = 115 MINUTES 
AUTOMATIC CONTROL 
PHASE HI 
200- 
1 r 
5 6 7 8 
TIME (minutes) 
Figure 12. — An example of phase III control. 
symetrical responses. That is, the dynamics of 
adding the drug are quite similar to those seen 
while removing the drug. The present approach 
would lend itself well to the control of hyperten- 
sive crises except that the ganglionic blockers 
etc. appear to cause quite asymetrical re- 
sponses. Thus, an asymetrical model is required. 
Asymetrical modeling requires only that we use 
two models, one for each direction. We are also 
investigating multiparameter control as re- 
quired for shock control. In addition to the clin- 
3/2 5/7! SICK DOG 
ELAPSED TIME =9 8 MINUTES 
AUTOMATIC CONTROL 
PHASE m 
2oa 
TIME (minutes) 
200 
1 r 
5 6 
TIME (minutes) 
Figure 13. — An example of phase III control. 
Figure 14. — An example of phase III control. 
ical entities noted above we are also examining 
the control of cardiac arrhythmias. 
Another aspect of the whole question of con- 
trol that we are studying is what we call con- 
currence control. To make such systems more 
acceptable to physicians, we are studying the 
feasibility of allowing them to concur with all 
decisions made by the controller. Thus the con- 
troller will have to get the physician's permis- 
sion to make any change in drug rate beyond 
preset bounds. Obviously no such system will 
ever be optimum, but because of its adaptive 
qualities, it is hoped such systems will function 
adequately despite the delays caused by the phy- 
sician. 
SUMMARY 
We have been studying a relatively simple 
therapeutic control problem, the support of hy- 
potension in dogs with a vasopressor. Such a 
procedure is attractive as a starting point be- 
cause one can easily measure and effect blood 
pressure. Also, such a control problem has the 
characteristics common to most therapeutic 
control situations i.e., a nonexplicit time vari- 
able delay and response. 
ACKNOWLEDGMENTS 
This paper is the result of the Ph.D. work of 
Dr. C. M. Schade done under Dr. Widrow with 
the assistance of Dr. Thompson. 
