00100		A. THE PILOT ASSEMBLY STATION
00200	
00300	Half the effort of the proposed project is to extend our existing  robot
00400	system,  to  specialize on assembly tasks, and meanwhile, to transfer
00500	the system to a standard minicomputer, in order  to  demonstrate  its
00600	applicability  to  real  problems.  The  project will be considered a
00700	success if it has major impact on an industrial assembly system or
00800	if  an  enterprise  chooses  to  manufacture  and  market the general
00900	purpose station.
01000	
01100	Our current robot system has one arm of our design,  two  TV  cameras
01200	and  a  laser  ranging system.  A variety of vision modules have been
01300	standardized, and a hand language (HAL) has  been  used  for  several
01400	tasks  [Paul].   A demonstration of the robot's capabilities has been
01500	made by assembling a water pump, using vision, touch and a tool.
01600	
01700	We  propose  to  extend  that  system  by:  a)  doing  more  assembly
01800	experiments  to  guide  research  in developing the hand language and
01900	manipulation techniques. b) design and  implement  visual  operations
02000	especially for assembly, which would not be programmed in our general
02100	vision effort, again using assembly experiments to guide research. c)
02200	adding a second arm which is nearly completed, and improved touch and
02300	force sensing; extending the hand language  to  two  arm  operations.
02400	demonstrate  practical  speed  and  dexterity  with two arms and tool
02500	using. d) after some progress with b, implementing a  language  which
02600	combines  vision  with  the  hand  language.  show the commonality of
02700	operations among similar assembly tasks.
02800	
02900	Simultaneously, we would transfer our robot system to a  minicomputer
03000	which  we are acquiring.  One benefit of this transfer is for our own
03100	research.  Another is that we can show the relevance of our  research
03200	to systems which are cost effective for industrial applications.  Yet
03300	another is that we can transfer our research results more directly to
03400	industrial  observers  and  industrial  laboratories.  In many cases,
03500	this can be direct  transfer,  since  we  have  the  beginning  of  a
03600	robotics user's group sharing the same minicomputer (GM research lab,
03700	3M research lab, JPL, and Artificial Intelligence lab,  MIT).   Also,
03800	our  hand  language, arm control and arm trajectory planning software
03900	are suitable for a variety of arms, including the Unimate.
04000	
04100	The result of this effort is a small system we call a pilot  assembly
04200	station,  based  on  a minicomputer.  Two related projects complement
04300	this proposed project.  These projects at SRI  and  MIT,  Draper  Lab
04400	have industrial participation to work with commercial manipulators of
04500	a type already in  use  in  factories  (Unimate).   They  propose  to
04600	develop systems which perform simple two dimensional inspection tasks
04700	and systems for manipulation at about our current level of dexterity.
04800	With  their  industrial participation, and a commercial device, their
04900	impact may be short term.  Our project is longer term in  that  while
05000	we  have  those accomplishments in manipulation and assembly, we work
05100	with equipment which is not a commercial product, so that  there  may
05200	be  a delay in converting our techniques to a commercial manipulation
05300	system, or marketing  our  manipulator.     Finally,  we  are  aiming
05400	toward  a  higher level of evolution, something like the simple robot
05500	described in the ISI report [Anderson].
05600	
05700	In one configuration, the assembly station is assumed to be connected
05800	to  a  large time-shared computer facility. Not only does this appear
05900	to be a good design for our problem, but it is also necessary for  an
06000	integrated  programmed  automation  facility.   The  only  emphasized
06100	statement in the ARPA study report [Anderson 1972] is:  "The  maximum
06200	impact  of  computers  in  manufacturing  will  come  from  complete,
06300	real-time computer  cognizance  and  control  of  all  processes  and
06400	resources,  allowing  the  precise scheduling and allocation of these
06500	resources."
06600	
06700	Each assembly station will consist of a small digital  computer,  one
06800	or  more  manipulators  and cameras. Manipulators and cameras will be
06900	combined as modules in systems to suit a  given  task,  and  will  be
07000	digitally  controlled  by the station computer.  The station computer
07100	may be connected to a central computer or may stand alone. The latter
07200	configuration  is  less  efficient  in  the  long run, but allows for
07300	incremental automation.
07400	
07500	The program in the station computer will be flexible as to the number
07600	and  position of manipulators and cameras. It will control input from
07700	the cameras, provide for some pre-processing of the picture  and  the
07800	application  of  simple  tests  to  the visual data.   In the case of
07900	manipulators it will provide for simultaneous servo control. It  will
08000	also  process  other  inputs,  such  as  from  touch  sensors  on the
08100	manipulators, photocell interrupt devices, limit switches, etc.  Both
08200	analog  and binary outputs will be available for the control of other
08300	devices.
08400	
08500	Each station computer will have access to a task  program  specifying
08600	the  actions and tests it should perform to carry out its task.  This
08700	task program will be compiled by the  central  computer.  Compilation
08800	will  consist of the analysis of predicted scenes in order to specify
08900	simple tests that the station computer can carry out, the calculation
09000	of  collision free trajectories for the manipulators, the sequence of
09100	operations and simple contingency plans. etc.  We  are  planning  for
09200	tasks   whose  normal  execution  does  not  require  detailed  scene
09300	analysis.  If scene analysis becomes necessary (e.g. a dropped  part)
09400	we  envision the station sending an image to the central computer for
09500	processing.
09600	
09700	The central computer will perform overall control of  the  individual
09800	assembly  stations  and  the relative rates of production.  If at any
09900	station a planned test fails, for which a contingency plan  does  not
10000	exist,  or  any  unexpected  sensor signals are received, the central
10100	computer will also be interrupted and appropriate actions  taken.  If
10200	the  assembly  station  is not connected to a central computer, these
10300	operations will require human intervention.
10400	
10500	
10600	Most  of  the  hardware  required  for  the assembly station has been
10700	developed for the laboratory.   The major additional  problem  is  to
10800	refine  the  existing  designs  to  meet the demands of an industrial
10900	environment.
11000	
11100	We are now acquiring a small computer, a PDP-11, and  will  interface
11200	the existing arms to it together with one or more cameras. A software
11300	package for the small computer will then be  developed  in  order  to
11400	control  the  manipulators,  read  the  TV cameras and to perform low
11500	level visual analysis. Programs to plan tasks, calibrate  the  system
11600	and  to  perform  higher  level  visual  analysis, error recovery and
11700	strategy planning will be written in a high level  language  for  the
11800	time-sharing processor.
11900	
12000	The most difficult part of the assembly station, and the heart of the
12100	project is a demonstration of the feasibility of the station. We plan
12200	to do this by developing programs which actually assemble and inspect
12300	a variety of fairly complex manufactured items.
12400	
12500	Our first task was assembly of an  automobile  water  pump.   A  film
12600	shows  the  process in detail.  The assembly consists of four sets of
12700	operations: Visually locating the pump body to move  it  to  standard
12800	position;   inserting  guide  pins  and  the  gasket,  then  visually
12900	verifying that the gasket is in place; picking up the pump  head  (at
13000	an approximate standard position) and placing it over the guide pins;
13100	using a tool to pick up screws from a  feeder  and  insert  them,  by
13200	searching near the hole if necessary.  Tools, vision, touch and force
13300	feedback were used. The assembly was at  approximately  human  speed,
13400	but not as fast as a practiced worker.  The reliability is relatively
13500	high.  Assembly was programmed in the hand  language,  with  a  small
13600	program, in terms of macros like "insert a screw".  Visual operations
13700	were special case programs.  A major part of the development  of  the
13800	pump  assembly  was  implementing  tools  and  finding  manufacturing
13900	engineering techniques.  Speed and reliability  in  repetitive  tasks
14000	could  be  increased  by  using  errors  found in searching to insert
14100	screws to update target positions for screw insertion, etc., so  that
14200	these operations were almost always immediately successful.
14300	
14400	As  we move to more difficult tasks, the sophistication required goes
14500	up quite rapidly.  The  current  manipulation  routines  seem  to  be
14600	expandable  to cover anticipated problems, especially since we intend
14700	to employ special purpose manipulators (tools)  for  some  tasks.   A
14800	major  part  of  our  effort will be vision for assembly; an assembly
14900	task will be performed using vision,  touch,  and  two  hands.    The
15000	vision  effort  will be integrated with other abilities, and while no
15100	great generality will be stressed, new modules will  be  incorporated
15200	into  the  system,  and  a  set of techniques will evolve. These will
15300	depend heavily on techniques similar to those of our  recent  studies
15400	in  visual  feedback  and  hand/eye  coordination  [Gill  1972].  The
15500	internal model of the world will be used to predict visual appearance
15600	of  selected areas of the scene, and a set of verification procedures
15700	will confirm or contradict the predictions. Many  of  the  operations
15800	will  be  connected  with  supplying  feedback  to  manipulation, for
15900	alignment and insertion. These visual procedures will have models for
16000	the parts and approximate locations.
16100	
16200	This and other visual tasks connected with assembly will initially be
16300	performed as special case  problems.   In  addition,  to  make  these
16400	techniques   available,   we  will  pursue  development  of  strategy
16500	programs. This step is essential if intelligent assembly devices  are
16600	to  be conveniently programmed for complex tasks. Within the division
16700	of labor between large and small machines, this facility  resides  in
16800	the  central  time-sharing  facility  which  has  large  memory, file
16900	storage, and problem-solving languages. The system must have a  model
17000	of  the  parts  for  assembly,and  a  sequence  of  instructions  for
17100	assembly. It will be necessary to have a representation of shape  and
17200	models for usual assembly operations such as insertion and alignment,
17300	and the ability to solve simple problems  in  connection  with  these
17400	operations,  all  this  to  enable  convenient communication with the
17500	user. When these facilities exist, the addition of new  features  and
17600	operations  can be incorporated smoothly, so that users can call on a
17700	growing library of planning modules. The result of a planning  effort
17800	is  a  cycle  of  machine  planning and man-machine interaction which
17900	results in a plan  network  for  the  mini  computer  which  controls
18000	individual machines.
18100