00100	
00200						Feb 8,1972
00300	
00400	Dr. Douglas O'Handley
00500	Jet Propulsion Laboratory
00600	4800 Oak Grove Drive
00700	Pasadena, Calif,91103
00800	
00900	
01000	
01100	
01200	
01300	Dear Doug:
01400	
01500	
01600	
01700	A proposal is enclosed, which is a  bit  informal,  but  conveys  the
01800	information that you may want to give feedback on.
01900	
02000	We  are,  independent of any contract, glad to accomodate any of your
02100	staff for extended periods in  the  lab,  and  communicate  and  help
02200	document  software.  We would also entertain technicians or engineers
02300	for replication of the arm, and guide the orders  for  machining  and
02400	parts,  aid  with  electronics, and guide assembly.  It is understood
02500	that our  part  would  be  informal  and  close  advice.  A  6  month
02600	construction  time  seems  conservative and firm. If the decision has
02700	been made, we might immediately proceed to help to start  fabrication
02800	and parts orders.
02900	
03000	
03100	
03200	
03300	
03400						Sincerely,
03500	
03600	
03700	
03800						Thomas O. Binford
03900	
04000	
04100	
     

00100	We propose a collaboration for several programs in the development of
00200	computer controlled manipulators and in visual depth perception.    A
00300	program,  with  Prof  John  McCarthy  as  principal  investigator, is
00400	outlined.
00500	
00600	In  the  first  stage   of   development   of   computer   controlled
00700	manipulators,  we  have  achieved  fast  and accurate hardware at low
00800	cost, and the computer control software to plan and  execute  smooth,
00900	fast  motions.     The software has been systematised in an extensive
01000	user package.    This first stage  might  be  called  the  trajectory
01100	control  stage.    Trajectory planning distinguishes this system from
01200	teleoperators and industrial manipulators for which a human  operator
01300	plans  motions.   There is a second stage of arm function which could
01400	be called the interactive stage, incorporating sensing, and  reaction
01500	to  sensed  external  conditions.  We have used some elementary touch
01600	and joint force sensing  and  interaction,  and  are  working  toward
01700	improved  sensors,  together  with  visual  feedback and software for
01800	avoidance, contingency planning, and  assembly  operations.   We  are
01900	constructing a hand capable of complex operations.
02000	
02100	Our  research  in  vision has made progress in several forms of depth
02200	perception, and in mathematical  representation  and  description  of
02300	shape.   We have had operating depth-ranging hardware for a year, and
02400	can make crude stick figure descriptions of some  toys,  including  a
02500	doll  and  a  horse.    Effort  is directed toward action in the real
02600	world; we have done primitive world  modelling  based  on  the  depth
02700	impression for a moving observer (motion parallax), and have inferred
02800	depth from texture gradients.   These two  provide  human  perception
02900	with  information  about  depth  beyond  the  useful  range of stereo
03000	vision.
03100	
03200	
03300	We propose a program of sensor design and implementation, development
03400	of software using touch for adaptively picking up irregular  objects,
03500	implementation of improvements to the control electronics to optimize
03600	the sharing between computer and external control, and adaptation  of
03700	the  servo software to a PDP-11. The adaptation will require purchase
03800	of a small configuration PDP-11 for our laboratory and  approximately
03900	two months by an experienced programmer from the JPL staff.
04000	
04100	We also propose to extend work in depth perception by motion parallax
04200	and stereo.     In particular, we plan to  implement  a  proposal  by
04300	Binford on efficient stereo correlation.
04400	
04500	The  total cost of the projected program, in hardware and staff would
04600	be $75,000.