COMMENT ⊗   VALID 00003 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	\\M1BDR30\M2BDR25\F1
C00012 00003	Application of AI toward productivity concerns sensor devices,
C00014 ENDMK
C⊗;
\\M1BDR30;\M2BDR25;\F1
\CAnthony Connole
\F2\CAdministrative Assistant, United Auto Workers

\JAny examination of  the "Frontiers of Computer Science"  must include
an  identification  of  its socio-economic  effects  and  planning to
insure that  its impacts  will be  favorable to  our  society.   This
should involve  a study of  its impact  on consumers aad  workers, on
levels of employment and unemployment, on changes and obsolencence of
manpower  skills, and  on  both  short  and long  term  security  for
affected work forces.  In short, we must do those things necessary to
assure that  computer  science  serves  people,  rather  than  people
becoming subservient to computer science. \.


\F1\CJ.L. Nevins
\F2\CCharles Stark Draper Laboratory, Inc.

\JThe applied research issues being explored for a new class of systems
for  performing  automatic  mechanical  assembly  described.    These
systems are  organized about  sensor arrays  that measure  the forces
present when two pieces interact during the process of assembly. 

The   research  issues  described  include  methods  for  classifying
mechanical assembly, assembler  system configurations of interest  to
assembly classes,  associated motion regimes  and control strategies,
task analysis,  sensor and  servo  integration, sensor  systems,  and
experiments under way to verify the proposed strategies. \.


\F1\CM.  Eugene  Merchant,
\F2\CDirector of Research Planning, Cincinnati Milacron Inc.

\JA  recent  international  Delphi-type   forecast  of  the  future  of
manufacturing   strongly   indicates  that   the  computer-integrated
automatic factory  will be  a reality  well  before the  end of  this
century.  However, because of the potential major economic and social
benefits which this development can bring to a country, some  nations
are pushing to accomplish this even earlier  than forcast.  Of these,
Japan has already made the most significant strides in this direction
and is now planning a  national program to have a prototype  unmanned
machine building factory in operation by  about 1980.  This wuld be a
factory  of about 200,000  to 300,000 square feet  floor space, which
instead of the normal work force of 700 to 800, would  be manned by a
force of approximately ten workers.  The development cost is expected
to be in excess  of $100 million,  with approximately one-third  that
sum devoted  to  the development  of the  necessary software  system.
Further details on  these developments will be presented at the panel
session. \.



\F1\CCharles Rosen
\F2\CAI Center, SRI

\JMaterial-handling,   inspection  and  assembly  processes  are  still
heavily   labor-intensive    in   many   manufacturing    industries.
Furthermore,  many of  these industrial  jobs are  dull, repetitious,
noisy, dangerous, or  otherwise undesirable.   These factors lead  to
high cost  and poor quality  of product, and  to low  productivity of
workers.     Computer-controlled  manipulators  coupled  with  visual
tacticle, and other sensors are now beginning to be programmed in the
laboratory  to perform  many operations  that in  theepast have  been
reserved  for humans,  because they were  too costly  or appeared too
difficult to do in any other way. 

Computer  programs are  now  being  developed, which,  together  with
increasingly   inexpensive  digital   hardware,  will   soon  provide
cost-effective production tools which  will enhance both the  quality
of products and jobs. \.
∂14-NOV-74  1629		network site RAND
 TO:             Lou Paul
 
 FROM:           Bob Anderson
 
 SUBJECT:        NCC Panel on Automation:
                 Title and Abstract for my Contribution
 
 
 
              PRIORITIES IN THE DEVELOPMENT OF
            INTEGRATED FACTORY AUTOMATION SYSTEMS
                    Robert H. Anderson
 
 
 Probably the most important contribution of the computer to
 increasing the productivity of discrete product manufacturing
 will be in aiding the management and control of the complex
 production process.  Automated workstations will have little
 effect unless parts and tooling are almost always available
 exactly when needed;  this implies greater control of production
 than is currently exercised (or even possible, using current
 techniques).  On the other hand, a greater degree of automation
 in workstations can provide timely, reliable data on which better
 management and control systems can be based.  Automated workstations
 can also be controlled directly by computer-based supervisory
 systems, thereby contributing to system responsiveness to
 management control.  Neither workstation automation nor computer-
 based management and control systems should be developed in
 isolation; they are highly interdependent.
 
 The best application area for demonstrating initial successes in
 computer-based manufacturing automation appears to be in the
 production of electronic subassemblies, such as avionics subsystems,
 minicomputer CPUs, and electronic consumer products.  Some reasons
 for this assessment are:  flexibility is needed due to the
 constantly changing technology; automation devices are not directly
 competing with human manipulative skills in the micro-miniaturized
 electronics realm; there is a need for complex electronic testing
 as an integral part of the manufacturing process, and both micro-
 assembly and testing can be integrated within a computer-based
 automated workstation.
 
 The above viewpoints emerged from a recently completed automation
 study (Anderson, R. H., and N. M. Kamrany, "Advanced Computer-
 Based Manufacturing Systems for Defense Needs," ISI/RR-73-10,
 USC Information Sciences Institute, September 1973); case studies
 of product manufacturing described in this study will be used
 to discuss relative priorities for R&D tasks in computer-based
 manufacturing automation.
 
 
 
Application of AI toward productivity concerns sensor devices,
control of assembly using sensory feedback, and programming assembly
devices ranging from special purpose "hard" automation to general
purpose manipulators.  For programmable devices to make an impact,
they must be conveniently programmed in user-oriented high level
language systems.  But more is required: to program at the level of
instruction manuals, with the machine performing necessary
bookkeeping; to simplify setup of assemblies using sensory
capabilities.  A principal area of AI is the study of representation
and descriptor systems.  As applied to productivity technology,
descriptor systems enable machines to understand assembly primitives. 
Setup and planning operations can be carried out using small amounts
of time on large systems with advanced facilities.  Repetitive
execution of assembly can be carried out on small dedicated machines
with optimal, special-purpose programs with only needed facilities,
compiled by planning systems on large machines.