00100	Abstract
00200	I	The Need for Advanced Automation
00300	II	Research Plan
00400		A.	The Pilot Assembly Station
00500			1.	Tasks
00600			2.	Hardware
00700		B.	Vision Development
00800		C.	Manipulator Development
00900		D.	Systems Support and Related Issues
01000	III	Dissemination of Results and Relation with Industry
01100	IV	Staff
01200	V	Computing Facilities
01300	VI	Budget
01400	VII	References
01500	VIII	Relevant Laboratory Output
01600		Films,THeses,Reports
     

00100			THE NEED FOR ADVANCED AUTOMATION
00200	
00300	
00400	
00500	The benefits available  from  automation  in  manufacturing  are  now
00600	widely  appreciated.   The  current  interest  in advanced (computer-
00700	oriented) automation stems  from  two  sources.   The  first  is  the
00800	decreasing   cost   of   computation   combined   with   an  improved
00900	understanding of how to  apply  computation  to  various  aspects  of
01000	manufacturing.  The second cause is the realization that many current
01100	manufacturing operations, primarily  assembly  and  inspection,  will
01200	require quite complex automation techniques.  The same considerations
01300	apply to the automation of  tasks  such  as  laboratory  and  medical
01400	procedures,   but  the  emphasis  has  been  (as  ours  will  be)  on
01500	manufacturing.  kThere has been a great deal written  lately  on  the
01600	potential  benefits of advanced automation and we will not attempt to
01700	restate  the  points  made   so   cogently   elsewhere.    A   recent
01800	ARPA-sponsored  study  []  in  which  we  participated  suggests  the
01900	following major benefits from advanced automation:
02000		reduced prices
02100		increased product quality and uniformity
02200		customization of products
02300		increased job satisfaction. We would like  to  add  just  one
02400	comment.   As we have seen with other technological developments such
02500	as numerical-controlled tools, there will be things made  which  were
02600	previously  impractical to make. There is widespread agreement in the
02700	technical , industrial and government communities  on  the  potential
02800	benefits of advanced automation.  There is much less agreement on the
02900	level of sophistication needed in automation hardware  and  software.
03000	We   believe  that  the  most  practical  solution  for  intermediate
03100	quantities will be to develop general purpose assembly and inspection
03200	devices controlled by sophisticated programs in digital computers.
03300	
03400	We  assume  that present manipulators have a small but stable market,
03500	and that most will have minicomputers  soon  (Sundstrand  does  now).
03600	Will  the  incremental  cost  of  touch,  force and visual sensors be
03700	offset by the opening of new  capabilities,  and  will  the  cost  be
03800	offset  by  reduced  cost  of  system design for machines which allow
03900	loose alignment tolerances?  We estimate hardware costs in a separate
04000	paragraph,  and estimate that by 1977, the incremental cost of visual
04100	sensors will be roughly $1100, with about $300 additional for a laser
04200	ranging  scheme  as  we  use  now.        With  the additional memory
04300	requirements, and the multiple of two in  marketing,  visual  sensors
04400	will add $4k to $6k to the selling price, or about 15%-20% of current
04500	manipulator prices.     Presumably labor costs will inflate also,  so
04600	that  cost  comparisons  will  not  change appreciably over that time
04700	scale unless significant  new  mechanical  designs  are  made.      A
04800	conclusion is that sensor-equipped robots will be available at slight
04900	additional cost.
05000	
05100	We do not expect mechanical production costs to drop dramatically  as
05200	electronics  costs  fall  with integration.   The most promising cost
05300	reduction development would be high-production standard modules  such
05400	as  robots.     As an example, consider the Stanford arm; produced in
05500	unit  quantity,  parts,  manufacturing   and   assembly   costs   are
05600	approximately  $10k.  The arm is much simpler than an automobile; its
05700	potential  selling  price  in  large   quantity   is   $3k,   without
05800	minicomputer   control.      Potentially,  a  robot  with  two  arms,
05900	minicomputer control, and touch sensors is possible for under $10k in
06000	quantity.   This  is  the range of price of present transfer machines
06100	which have limited  market  because  they  have  limited  motion,  no
06200	control,  and are thus locked out of the economy of scale.  Versatile
06300	assembly devices have an advantage in utilization in medium scale and
06400	job  shop  production;  their  utilization  efficiency  can be higher
06500	because versatility allows a variety of operations  to  be  scheduled
06600	simultaneously.     The  availability  of  computers  for  production
06700	scheduling will allow the flexibility to be used to advantage.   With
06800	simple  devices, transfer and assembly operations must be broken down
06900	into small operations  with  comparable  cost  to  a  more  dexterous
07000	system.  Robots can cut down on real estate, since several operations
07100	can be done at one station; belts and other transfer devices  require
07200	more  space than desirable.  A robot which can pick parts from a cart
07300	would save considerable space.  Robots would also  have  longer  life
07400	over  which  to amortize costs, since they could operate with greatly
07500	changed production operations,  and  evolve  with  new  hardware  and
07600	software  modules.    They  would have higher reliability, since they
07700	would have monitoring, contingency action, and error recovery.   They
07800	would be self-aligning, self-calibrating, and self-testing. There has
07900	been  difficulty  in  meeting  the  tight  tolerances  necessary  for
08000	operation  of  non-intelligent  devices.   It is hard to estimate the
08100	value of systems which  work  within  current  industrial  tolerance,
08200	rather  than  requiring radical improvements in engineering practice,
08300	but that continuity does seem important.
08400	
08500	Functional modularity has been emphasized in our studies;  modularity
08600	is  clearly  possible  and  there is a powerful economic incentive to
08700	assemble widely varying systems from one, two, or three arm  modules,
08800	force  and  touch  modules, vision and ranging modules, with planning
08900	and monitoring modules,  computer-aided  mechanical  design  and  job
09000	specification  for robots on a local or network time-sharing service,
09100	with software modularity. This approach to advanced automation  saves
09200	much  wasted effort. The economies of scale in production of assembly
09300	modules are matched by economies of intellectual scale:   the  design
09400	process  is  modularized  and  easily communicable; there is a common
09500	discipline which can be widely absorbed; developments of many sources
09600	can  be  integrated.     The  alternative  is special design for each
09700	operation.  We expect that at some point, design will be cheaper with
09800	standard  robot modules than with special purpose designs, just as it
09900	is now more economical to use mini-computer controllers than  special
10000	purpose  controllers.     To  some extent, the dehumanized industrial
10100	worker now fills the role of standard  module,  so  that  we  do  not
10200	expect   discontinuous  change.    The  continuity  with  decades  of
10300	experience with design for  human  versatility  in  assembly  can  be
10400	relied  upon;  robots  with  some  versatility,  sensing, contingency
10500	actions, and dexterity will provide a  transition  to  newer  schemes
10600	which  will  evolve  more slowly. Advances which benefit special case
10700	automation and traditional trends in automation for human  production
10800	will  also  aid general purpose devices.  For example, new techniques
10900	in bonding and fastening will simplify many  assembly  tasks  to  the
11000	capability   of   robots,   while   still   too  complex  for  simple
11100	servomechanisms. Another advantage of general purpose  techniques  is
11200	the  ease  with which the size of the robot can be scaled up or down.
11300	There appears to be no difficulty in making very large or very  small
11400	assembly stations which use essentially the same programs.
11500	
11600	The  development  of  general  purpose  assembly devices will benefit
11700	manufacturers and a broad spectrum of industry.  There does not  seem
11800	to  be  a  private  institution  with the competence or the desire to
11900	undertake this program.   It is an appropriate area for  governmental
12000	action to help meet a national need.
12100	
12200	But given sensor-equipped robots at little extra cost, they will have
12300	visual ability  to  sense  color,  strong  contrasts  in  brightness,
12400	recognize  known  objects  from two dimensional outlines (top view of
12500	flat object) or three dimensionally using depth, to a certain  extent
12600	they  will  be  able  to  recognize  three  dimensional objects using
12700	monocular recognition.   They will be able to visually  align  mating
12800	parts    in    assembly    operations,    increase   reliability   by
12900	self-calibration and flaw and disaster detection,  and  permit  large
13000	tolerances  in parts transfer, even permitting parts to be taken from
13100	a bin.       Several of these abilities are not yet available in  the
13200	laboratory,   but   with  known  conditions  and  objects,  and  some
13300	engineering (color coding) are achievable within that time span.
13400	
     

00100			VISUAL DEVELOPMENT FOR APPLICATIONS
00200	
00500	A major part of our effort will  be
00600	vision for assembly; an assembly task will be performed using vision,
00700	touch, and two hands.    The vision effort will  be  integrated  with
00800	other  abilities, and while no great generality will be stressed,
00850	new problems will be faced, new
00900	modules will be incorporated into the system, and a set of techniques
01000	will evolve.
01005	Visual feedback for alignment and insertion is a primary and typical
01010	function.  Visual procedures will have models for parts and approximate
01015	locations.
01200	The  internal  model  of  the  world  will  be used to predict the
01300	appearance of simple features in selected areas of the scene, and a set of  verification
01400	procedures  will  confirm or contradict the predictions.
01405	Other visual operations will be based on similar techniques.  This
01410	is particularly appropriate for the industrial environment for which
01415	parts are well known in advance, and positions are known well.  Simple
01420	procedures will be used for gross localization, since a little engineering
01425	of the problem can provide adequate contrast and color coding for a few
01430	such locating tasks.
01700	A typical  example  is  our
01800	current project, assembly of a water pump.  The major parts are to be
01900	located, the screws  are  supplied  as  is  standard,  by  a  feeder.
02000	Locating the body and cover of the pump require a gross localization,
02100	then a fine localization of screw  holes.   Touch  provides  a  final
02200	accuracy  and  independent coordination of the hand.   The placing of
02300	alignment pins within screw holes can be monitored visually; although
02400	the previous actions provide adequate accuracy.  Placing of the cover
02500	over alignment pins can be carried out without vision.  However,  our
02600	improved  force  sensing  will  be  useful  for  that  operation when
02700	completed.  The insertion of alignment pins in other screw holes, and
02800	the insertion of screws in holes will be monitored visually. Although
02900	the assembly has been performed up to tightening of  the  screws,  by
03000	the hand alone, this will provide a good test case for development of
03100	vision, which will increase reliability.
03200	
03300	To be precise in what vision software will be  implemented,  we  will
03400	greatly  expand  our efforts in visual feedback; at present these are
03500	adequate for polyhedra.    We will generalize these to  objects  with
03600	with curved edges, and extend the quality of the verification modules
03700	in  straightforward  ways  using  our  edge   detection   and   color
03800	facilities.  These prediction techniques involve 2d shape description
03900	techniques.   Template matching techniques  are  straightforward  and
04000	standard;  we  expect  to make more sophisticated description modules
04100	which have part/whole capability, of use in  planning.   Some  simple
04200	extensions  to 3d descriptions are planned.  We will extend the power
04300	of our region finding and color  region  analysis  programs.   Visual
04400	input  devices  will  be  interfaced to the minicomputer and existing
04500	modules converted.
04600		Vision research
04700	
04800	Visual  tasks  connected with assembly will initially be performed as
04900	special case problems.   As we have pointed out,  communication  with
05000	the  machine will be necessary for job specification and instruction.
05100	In parallel with progress on  vision  for  assembly,  to  make  these
05200	techniques  available,  we will pursue development of visual strategy
05300	programs. We describe separately the work on construction strategies,
05400	and limit this discussion to visual strategy.  The domain of assembly
05500	is particularly attractive as a subject for strategy programs,  since
05600	the context is much richer than in more artificial environments, such
05700	as the blocks world.  Although a much more  complex  problem  in  the
05800	shapes  and  structural relations, in some ways, it is simpler; there
05900	are  more  types  of  visual  information  to  use,  more  contextual
06000	constraints,a  greater  variety  of simple verification operations to
06100	perform.   This step is essential if intelligent assembly devices are
06200	to be conveniently programmed for tasks. Within the division of labor
06300	between large and small machines, this facility must  reside  in  the
06400	central  time-sharing  facility which has large memory, file storage,
06500	and problem-solving languages.  The system must have a model  of  the
06600	parts  for  assembly,  and a priority from higher up of the parts and
06700	relations to verify or perceive. It  will  be  necessary  to  have  a
06800	representation of shape and models for usual assembly operations such
06900	as insertion and alignment, and the ability to  make  predictions  of
07000	expected  visual scenes and choose a sequence of operations, all this
07100	to  enable  convenient  communication  with  the  user.   When  these
07200	facilities  exist, the addition of new features and operations can be
07300	incorporated smoothly, so that users can call on a growing library of
07400	planning  modules.   The  result  of  a planning effort is a cycle of
07500	machine planning and interaction which results in a plan network  for
07600	the mini computer which controls individual machines.
07700	
07800	
07900	We will continue with our work  with  description,representation  and
08000	recognition of complex objects using depth data.  We are able to make
08100	complete descriptions of objects such as rings and s-shaped snakelike
08200	objects.   The  programs  can  segment  dolls,  toy  horses,  etc  in
08300	intuitively appealing ways and describe the segments well.  The  next
08400	efforts will be to improve and generalize the segmentation process of
08500	describing primitive parts, to link the  descriptions  of  primitives
08600	into  joints,  then to form object descriptions from the descriptions
08700	of joints.    The  recognition  of  such  objects  depends  first  on
08800	internal  modelling  of an object, then on matching perceived objects
08900	with previously stored internal models.  Our representation is useful
09000	in  giving  a  symbolic  description  of  objects which is useful for
09100	visual memory. We intend also to apply large parts  of  our  programs
09200	for   description  and  recognition  of  curved  objects  using  only
09300	monocular intensity information.  This is a much harder problem,  but
09400	within  our paradigm, shares much of the same code, and relies on the
09500	same conceptual basis of local 3d models.   This  might  seem  overly
09600	sophisticated  to  be  relevant  to  industrial robotics, since early
09700	applications might use the simplest facilities.   Our  representation
09800	of  shape  provides a natural way of representing shape internally in
09900	symbolic form and in communicating  with  a  user.  There  are  other
10000	programs  at  the  laboratory in computer graphics (OCCULT) geometric
10100	modelling (GEOMED) which we plan to use as a basis for Computer-Aided
10200	Mechanical Design.
10300	
10400	
10500	
10600	We  will  continue  work on color region analysis and on a new system
10700	which combines region  analysis  with  semantics  about  objects  and
10800	relations.   These  programs  already  exist, but are under extensive
10900	development to integrate more knowledge and to fit  into  a  complete
11000	visual  system. We anticipate that they will function in more complex
11100	environments than have been attempted in the past. There is a growing
11200	effort  on outdoor scenes, in connection with navigation of vehicles.
11300	While this is not directly relevant to the industrial environment, it
11400	does  reflect the growing involvement in complex visual tasks, and is
11500	another force for development of useful visual programs.
11600	
11700