00100	applications of robotics research
00200		automated assembly
00300		computerization of industry
00400			group technology
00500			scheduling and routing
00600	
00700	
00800	
00900	
01000	benefits of robotics research
01100		productivity increases
01200		social need for productivity
01300			social service needs
01400			more leisure
01500			unmet material needs: housing,public transit, poverty
01600	
01700	areas of robotics most likely to have short term payoff:
01800		manipulation w/ touch
01900		representation of shape and space
02000		perceptual systems-integrated systems
02100	
02200	necessity for programmibility in robotics
02300	
02400	structure of models
02500	
02600	knowability of things
02700		permanence
02800		contiguity
02900		computational requirements
03000		separability of objects
03100		independence of variables
03200			interactions separately described
03300		classification:objects cluster  can be classified into similarity classes
03400		structure		part/whole
03500		form=function
03600	
03700	mechanistic interpretation of affective behavior
03800			eg force someone to do something
03900	symbolic models
04000	
04100	
04200	information science in production
04300		systematizing control of production
04400		modularization and commonality of parts and operations
04500		description of parts and objects
04600		design automation
04700		analysis of tasks
04800		communication and file storage
04900		scheduling
05000		routing
05100		sensing, monitoring
     

00100	REPRESENTATION OF SHAPE
00200	
00300	what tasks required of a representation
00400		choice of parts
00500		manipulation
00600		avoidance
00700		display
00800		recognition
00900		visual feedback
01000		perceptual strategies
01100		learning
01200		3-d cues
01300	
01400	what operations required
01500		display-2 dim proximity
01600		manipulation-center of gravity,center of rotation,length, thickness
01700		avoidance-3 dim proximity, find free space
01800		visual memory-encode shape into memory reference descriptors
01900		recognition-features,visual memory,hypotheses,prediction,verification
02000			geometric relations among parts
02100		
02200		simplicity criteria for cross section function
02300			constant
02400			monotonic
02500			continuous
02600			number of maxima
02700	
02800		curve representation is topologically inadequate in 2d
02900		surface representation is topologically inadequate in 3d
03000			a)curve is one-dimensional
03100			  surface is two-dimensional
03200			b)when do we need higher dimensions-proximity
03300			  choice of parts
03400				volume-parts vs surface-parts
03500				area-parts vs curve-parts
03600			c)local realizability
03700			d)dimension arguments
03800	
03900		proximity
04000	
04100		fabrication operations
04200		form=function heuristics
04300		ease man-machine communication		technology transfer
04400			input devices -3d
04500			symbolic input
04600			high-level manipulation
04700		predictive results for perception
04800		initial organization
04900		
05000		level of representation
05100		symbolic descriptions--iconic descriptions
05200		example of gravity discussed at coffee
05300		what are we going to do with descriptions
05400		descriptions relative to what we are going to do with them
05500	
05600		important properties are not in 2d image: 3d interpretation
05700		
05800		system should be a programming language
05900			modular
06000			heterarchical
06100			redundancy, consistency, inconsistency
06200			problem-solving system
06300			approximation of decision trees by statistical estimates
06400	
06500		desiderata for representation
06600			generative
06700			levels of detail
06800			3d interpretation
06900			high level
07000			mechanism for similarity
07100			part/whole segmentation structured description
07200	
07300		schema, additions, deletions, transformations
07400	
07500		volume representation
07600		similarity as a paradigm problem
07700		choice of paradigms
07800		principle of redundancy, hypothesis and verification
07900	
08000		significance of shape representation
08100			fabrication operations
08200			form=function
08300				basis for extending Winston's learning paradigm
08400			ease of man-machine communication
08500				input 3d
08600				display
08700				primitives for programming
08800				common vocabulary
08900			useful in initial organization stages
09000			useful for monocular vision
09100	
09200			group technology and overall computer control of production
09300				analysis of assembly tasks
09400				description of parts and objects
09500				modularization of process
09600				commonality
09700				automated parts description
09800				task programming for robots
09900				machining descriptions,selective accuracy and monitoring
10000			graphics
10100			visual feedback
10200			
10300	
10400		partial knowledge--backs of objects invisible in single scene
10500	
10600		visual memory
10700	
10800		3d representation valuable since accurate predictions can always
10900			be made; not true with 2d representation
11000	
11100		for man-machine communication, machine should have same
11200			concepts as man
11300			understands objects in terms of pieces joined
11400			or in terms of fabrication operations (lathe,drill)
11500			pasting
11600		relationships
11700			near
11800			touch
11900			inside
12000			join
12100			cross
12200			overlap
12300			parallel
12400			corresponding
12500			adjacent
12600			above
12700			left,right,in front,behind
12800	
12900		see many problems as AI problems
13000			graphics
13100			computer languages
13200	
13300		experimentation and learning
13400		three models for learning
13500			statistical inference
13600			deductive/inductive
13700			experimental paradigm of construction of models
13800				independence of variables, specified interactions
13900	
14000		depth information as a way of building up the map
14100	
14200		bilateral symmetry
14300	
14400		three-dimensional cues
14500			
14600		joints
14700			sleeves
14800			parallel joints-flat
14900			conservation at joints
15000		intersections as cues to shape
15100		depth measurement techniques
15200			laser
15300			stereo
15400			phase modulation
15500			time of flight
     

00100	description
00200		computational aspects
     

00100	Many  of  us now believe that robotics will ultimately transform much
00200	of industry; we cannot estimate how rapidly this change will  happen,
00300	but  we  expect  a beginning within a decade, and considerable effect
00400	over a few decades.   Initially, touch, simple  vision  and  computer
00500	control  can  incrementally improve industrial manipulators which now
00600	have no sensing and no  contingency  actions.     In  the  industrial
00700	context,  visual  feedback  functions in constrained situations, with
00800	known objects and background , and approximately known  position  and
00900	orientation,  for  alignment  of  screws with holes, mating surfaces,
01000	etc.   Not only is this problem area of technological importance, but
01100	visual  feedback is a useful paradigm for research.   Current efforts
01200	in  artificial  intelligence  concentrate  on  goal-oriented   visual
01300	systems which take advantage of knowledge of the environment.
01400	
01500	For  robots  to  be highly effective in assembly and other industrial
01600	tasks, their programming for new tasks must be made simple and rapid.
01700	Of  course,  even  if  programming  is  difficult, there will be many
01800	applications  for  which  it  is  useful  to  stockpile  programs  of
01900	production  operations  rather than stockpile parts. But for ultimate
02000	impact, the programming of assembly devices is of great importance. A
02100	mini-computer   to   control  an  assembly  device  is  adequate  for
02200	manipulation and simple visual operations.  Programming tasks for the
02300	assembly  robot  can  be  done  on  a  central  time-sharing computer
02400	facility with large memory,  file  storage,  advanced  languages  and
02500	graphics.  That system could be an internal system for scheduling and
02600	production control, or an external  service.       For  this  primary
02700	problem  in  robotics,  the  whole  of  AI  research is available for
02800	application,  and  the  most  sophisticated  robotics   research   is
02900	relevant.
03000	
03100	But  on  a broader scale, the computerization of industry will affect
03200	production on a shorter time  scale  and  equally  profoundly.    The
03300	problems  under  consideration in AI and the representations of shape
03400	and space which are evolving, have great relevance to the  impact  of
03500	information  science  on  production.           The most far-reaching
03600	consequences will come from intellectual  economies  of  scale  which
03700	follow from systematizing production control, from modularization and
03800	commonality  of  parts  and  operations,  typified  by   the   "group
03900	technology"  approach of Opits.  In this analysis, parts are coded by
04000	descriptors; facilities are reorganized to produce  similar  elements
04100	in  the same facilities.   More generally, the availability of design
04200	records makes possible identification of common parts and  processes,
04300	and  possibility for redesign.   These abilities rest upon facilities
04400	for  design   automation,   description   of   parts   and   objects,
04500	communication  and  filing,  scheduling,  routing,  and  sensing  and
04600	monitoring of production.    Implicit in all this is the  ability  to
04700	represent objects and operations symbolically.
04800	
04900	Minsky and Papert [Minsky 1972] have described the central issues and
05000	the relevance of robotics in artificial intelligence.   These  issues
05100	in intelligence are emphasized:
05200		symbolic descriptions of facts and computational processes
05300		manipulation of descriptions
05400		heterarchical control structures--problem-solving  controlled
05500			by heuristics which depend on the meanings of events.
05600	Our concern is implementation of symbolic descriptions of  shape  and
05700	space  which  are  adequate for integrated robot systems.   We do not
05800	think a single representation is adequate throughout a system, but we
05900	have  attempted to analyze tasks in terms of computational primitives
06000	(originating from topological and geometrical primitives  as  in  the
06100	computational geometry scheme of Minsky and Papert [Minsky 1969]) and
06200	design representation abilities accordingly.     Representative tasks
06300	are manipulation, avoidance, display, and recognition.   Manipulation
06400	requires mass and  moments  of  articulated  parts,  and  ability  to
06500	calculate  overall  moments  in  various  positions and orientations.
06600	Collision  avoidance  in  manipulation   requires   three-dimensional
06700	proximity,  to find objects near a trajectory or near an object to be
06800	grasped.   Display requires description of surface properties and two
06900	dimensional  proximity,  to  find  overlapping surfaces in the image.
07000	Recognition requires visual  memory  for  generation  of  hypotheses,
07100	prediction  and  verification  facilities,  geometric relations among
07200	parts, and a good choice of parts.
07300	
07400	We haven't space for a survey of all research  in  integrated  visual
07500	systems,  nor to discuss modules for three-dimensional interpretation
07600	of line  drawings,  edge  and  region  description,  corner  finding,
07700	prediction,  etc.  We  will  concentrate  on  relating  progress in
07800	description of complex objects and its relevance to  complete  visual
07900	systems.   Figures 1 through 3 show experimental results for computer
08000	description of some objects, from  a  program  by  Agin  at  Stanford
08100	[Agin].    The  program  creates  intuitively natural descriptions of
08200	parts for objects such  as  those  shown.    The  description  system
08300	depends  on  representations  designed  by  te author, and on a laser
08400	triangulation system described by the author and implemented by Agin.
08500	
08600	Although many would disagree about criteria for  representation,  and
08700	although  other  types  of  representations  have  various uses in an
08800	integrated system, the desiderata for representations were:
08900	
09000	a)	Level of Representation
09100	Representations  sh∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈∧␈ders, cones, toroids, and
17400	helices, all with arbitrary cross sections, are typical examples.
17500	
17600	The process of segmentation allows non-unique representations.   This
17700	permits a choice of simple representations; we can regard  it  as  an
17800	advantage  that we have such a choice. We have a simplicity criterion
17900	of shapes based on the transformation function on the  cross  section
18000	(constant=cylinder,  linear=cone,  monotonic, convex, single maximum,
18100	shape changing...).  As an example:
18200	
18300	
18400	
18500				MAN
18600				
18700			       [HEAD]
18800				 |
18900				NECK
19000				 |
19100		      [ARM]--<SHOULDERS>--[ARM]
19200				 |
19300			       TRUNK
19400				 |
19500			      <HIPS>
19600			      |    |
19700			  [LEG]    [LEG]
19800	
19900	
20000	
20100	
20200	
20300	where brackets indicate complex parts and <> indicate joints.
20400	ARM is:		ARM
20500	
20600			     UPPER ARM
20700				|
20800			      <ELBOW>
20900				|
21000			     FOREARM
21100				|
21200			     <WRIST>
21300				|
21400			      <HAND>
21500		  |	   |	    |	      |	      |
21600		[FINGER] [FINGER] [FINGER] [FINGER] [THUMB]
21700	
21800	
21900	
22000	A primitive has the representation:
22100	
22200			FOREARM
22300		    |	  |      |
22400		CROSS	AXIS	TRANSFORMATION
22500	       SECTION
22600	
22700	The  primary  considerations  for  the  utility of the representation
22800	concern the tasks we will need to  perform.     These  tasks  can  be
22900	described in various ways; the most enlightening description of tasks
23000	is in terms of their computational structure.   For  these  problems,
23100	topological  and  geometric  considerations  are  important.  Surface
23200	representations have been used for display purposes.    For  robotics
23300	applications,  the representation for manipulation would need to make
23400	explicit how to pick up an object, i.e. where is the center of  mass,
23500	the  center  of rotation, and what are the moments.  How thick is the
23600	piece, and how long?  These are more than  just  surface  properties,
23700	and  the  surface  representation  is  no  help  in determining these
23800	properties. These properties are the sort  of  properties  which  are
23900	natural  for  our  volume representations; however surface properties
24000	are quite naturally obtained in this representation.
24100	
24200	The cross sections  are  represented  in  the  same  fashion  in  two
24300	dimensions  as  objects  are  in  three  dimensions.   The part/whole
24400	segmentation is crucial.    The parts which are useful are  not  just
24500	curves,  but  arm-like  areas.    Thus  we  are  interested  in  area
24600	representations, not curve representations. The descriptions in terms
24700	of  a two-dimensional axis and normal width are similar to the volume
24800	representations described by a space curve and normal cross section.
24900	
25000			CROSS SECTION
25100		[AREA-PART]--<JOINT>--[AREA-PART]
25200	
25300	
25400	where a part of a cross section is an area defined by  subparts.    A
25500	primitive area-part is defined by:
25600			    PRIMITIVE AREA-PART
25700			|	|	|	|
25800		     WIDTH    AXIS   TRANS   SURFACE
25900					     PROPERTIES
26000	
26100	
26200	The  coordinate  system is local to each element.  Each link connects
26300	two joints.  Each link contains a transformation required to go  from
26400	the  coordinate  system of an element at one joint, to the coordinate
26500	system of an element at  the  other  joint.  This  is  a  translation
26600	combined  with a rotation.  It is most convenient to define the local
26700	coordinates in terms of actual vectors rather than with an  arbitrary
26800	basis.
26900	
27000	DEPTH MEASUREMENT TECHNIQUES
27100	
27200	The human visual system uses both stereo and motion parallax to  gain
27300	depth  information.   The two function in approximately the same way.
27400	With stereo, the images seen by the two eyes are nearly alike.   With
27500	motion  parallax,  successive  images  are  nearly  alike.   Thus, to
27600	correspond a point in one image with a point in the other,  a  search
27700	over a small range is necessary.  With motion parallax, corresponding
27800	points can be tracked through successive images to  achieve  a  large
27900	baseline,  and  depth  resolution at large distances.  Relative depth
28000	information (ie one point nearer by two  inches)  requires  only  the
28100	shift  between  correspondences.  Motion parallax was used by Nevatia
28200	[Nevatia] to measure depth of points on samples of rocks.  With a  50
28300	mm  lens and a 5 degree baseline at a meter, depth accuracy was about
28400	2 mm. This was measured for a cup with circular cross section.   That
28500	correspondence   error   is  about  1/8  th  tv  picture  cell.   The
28600	corresponding phenomenon is noted  in  human  vision  in  stereo  and
28700	vernier  acuity.    It  has a straightforward explanation in terms of
28800	expected errors based on the correlation process.
28900	
29000	Work is underway at several laboratories on stereo correlation. [Quam
29100	1972].    Current  programs  are  inadequate  conceptually  at  depth
29200	discontinuities and with respect to ambiguities; they  require  about
29300	20  minutes with PDP-10 class machines.  There are major improvements
29400	which can be made in the strategies for correspondence,  and  factors
29500	of  approximately  100  in  speed  can  be gained by low cost special
29600	purpose visual processors.
29700	
29800	Laser depth ranging hardware has been  operating  at  Stanford  since
29900	January,  1971[Agin].   As  in stereo, we know the angles of the rays
30000	from the laser to the object, and from the object to the camera. Thus
30100	we  know  the baseline and two angles of a triangle, which completely
30200	determines the position of the object illuminated by the  laser.   As
30300	shown  in  figure  4,  the  laser  beam  is diverged into a plane, to
30400	simplify the deflection sytem to one dimension, and speed data taking
30500	without  introducing  ambiguity. Each ray from the laser appears at a
30600	unique angle in the video image. The approximate depth resolution  is
30700	about 1 mm at 1 meter.  The limit on the speed of reading is about 20
30800	scans per second, with about 100  scans  required  for  a  reasonable
30900	resolution.    That  scan  time  is  shorter  than  the corresponding
31000	computation time.  The laser  required  to  operate  with  a  silicon
31100	vidicon tube costs only about $250.
31200	
31300	Other  techniques  are  described  in  a sensor study [Binford 1972].
31400	Ultrasonics have too slow a sample rate.  Most other techniques are a
31500	form  of  triangulation  as  in  the  laser system just described, or
31600	measure the time of flight or phase change of modulated signals  over
31700	the  flight  time.    Since light travels at 1 ft per nanosecond, the
31800	measurement of 1 mm resolution requires 2 picosecond time resolution.
31900	Presently,  most  optical  depth-measuring  techniques have speed and
32000	resolution more appropriate for surveying or vehicle navigation  than
32100	computer-controlled  assembly,  but  techniques exist for substantial
32200	improvement.
32300	
32400		TECHNIQUES OF DESCRIPTION
32500	
32600	We describe briefly some of the work from Agin's thesis  [Agin].  The
32700	plane  of  the laser beam intersects objects along space curves which
32800	project as curves from the viewpoint of the camera. These  are  shown
32900	in  figure  5.     Depth discontinuities in space appear as breaks in
33000	the curve in the image and slope discontinuities in space  appear  as
33100	slope discontinuities in the image.  In the representation, parts are
33200	defined by continuity of axis and normal cross  section.    The  path
33300	through  corresponding points (midpoints) of cross sections should be
33400	everywhere normal to the cross sections. That calculation  is  stable
33500	for successive approximation. Some spurious local axes are removed by
33600	extending local  axis  descriptions  to  impose  continuity.    There
33700	remain  some  parts  which  are  described  in  several ways.   These
33800	alternate descriptions must be evaluated by a more global choice.
33900	
34000	To find good starting points for local axis descriptions, portions of
34100	objects  are  found  for  which  the scan cross sections (usually not
34200	normal to the eventual axis) are smoothly changing and  overlap.   An
34300	estimate  of  the  axis  for  the  group  is  made  by connecting the
34400	midpoints of  scan  cross  sections.     The  midpoints  are  a  good
34500	approximation  to  the axis even when the scan is relatively far from
34600	normal to the axis.   Axis estimates from preliminary groupings for a
34700	doll  are shown in figure 6.   The approximate axis gives an estimate
34800	for the directions of normal cross sections, which in  turn  give  an
34900	improved axis estimate.  The process quickly gives an axis determined
35000	by normal cross sections.   The pieces are thus  described  as  local
35100	"generalized  cones".   An  estimate  is  made  of  the cross section
35200	function, and the cone is extended by another  iterative  description
35300	process.    Eventually,  when  the  description is extended as far as
35400	possible, it is broken off.   The condition for breaking off is  that
35500	the cross section function must fit a parabola well over the whole of
35600	the axis.  Figures 1 through 3 show the  descriptions  of  parts  for
35700	some objects.  Because the program does not have good representations
35800	of the boundaries,  the  crude  tests  for  cross  sections  crossing
35900	boundaries  are  sometimes  inadequate,  and  two parts may be linked
36000	together, causing some gross errors in description of parts.    These
36100	operations  are  performed  in  three-dimensions,  and circular cross
36200	sections are assumed.   The assumpton of circular cross  sections  is
36300	only an expedient and not at all fundamental to the method.
36400	
36500	For  objects  consisting  of  one  part, such as a ring or snake, the
36600	descriptions are quite natural.  A ring is described as an element of
36700	constant  cross  section  along  a  closed,  circular  axis.     More
36800	complicated objects with several parts have  reasonable  descriptions
36900	of the parts, but the parts are not connected together, so that there
37000	is no sense of complete structure.
37100	
37200		CURRENT DIRECTIONS
37300	
37400	The  description  techniques  have  exemplified   the   idea   of   a
37500	computational    equivalent   to   a   representation.    Parts   are
37600	characterized by  continuous  axis  direction  and  continuous  cross
37700	sections.   In  generalizing  the  class  of  cross sections, another
37800	direct computational realization  emerged:  the  realization  of  the
37900	representation  in  terms  of  "generalized translational invariance"
38000	[Binford].
38100	
38200	The descriptions are intended for recognition,  among  other  things.
38300	To  that  end,  the  current  work  is  involved  in forming complete
38400	descriptions of objects, linking parts together at  joints,  choosing
38500	among  alternative descriptions. The paradigm is recognition from the
38600	system's own descriptions  among  a  potentially  very  large  visual
38700	memory.  The techniques of visual memory are closely coupled to these
38800	symbolic descriptions and will be described in forthcoming work.
38900	
39000	
39100	We  have  discussed representation and description without mentioning
39200	statistical learning.     We think that structured learning  is  more
39300	relevant,  and  that  the  best examples of learning schema have been
39400	Winston's [Winston].     Instead of many trial learning, that  system
39500	learned significant new structures on a single trial learning.    But
39600	the  essential  difference  from  statistical  schemes  is  that  the
39700	learning consisted in building up symbolic descriptions, manipulating
39800	them and comparing them.  Instead of examples  contributing  to  some
39900	diffuse  evaluation  function,  comparison  of an example description
40000	with the concept  description  causes  modification  of  the  concept
40100	description  specific  to the differences. What types of descriptions
40200	are useful?     Winston's work was successful because his  primitives
40300	were    geometric    relations    like    left/right,    above/below,
40400	supporting/supported-by, etc, and he was able to express  interesting
40500	concepts  of  geometric  structures  as  simple combinations of these
40600	geometric relations.  We aim for an extension of these ideas with  an
40700	understanding  of  form  and  its semantics, as expressed in a richer
40800	description of form and relations. The ability to  make  descriptions
40900	and  modify them enables the system to remember objects.  There is no
41000	description of procedures and ability to  modify  such  descriptions;
41100	consequently,  the  system cannot learn about how to perceive complex
41200	objects.
     

00100	[Anderson 1972] R. Anderson,"Programmable Automation, The Future
00200	of Computers in Manufacturing", Datamation, Dec 1972.
00300	
00400	[Agin] G. Agin, "Description and Representation of  Curved  Objects",
00500	Ph.D.  Thesis, Computer Science Dept., Stanford University, September
00600	1972.
00700	
00800	
00900	[Binford 1972]"Sensor Systems for  Manipulation",  Conference  on  Remotely  Manned
01000	Systems, JPL, Pasadena, Calif, Sept 1972.
01100	
01200	[Binford 1971] "Visual  Perception  by  Computer",  Invited  paper  at  IEEE Systems
01300	Science and Cybernetics, Miami, Dec 1971.
01400	
01500	
01600	
01700	[Minsky 1972] "Artificial Intelligence Progress Report", AI Memo 252,
01800	MIT AI Lab, (1972);
01900	
02000	[Minsky 1969] M.L.Minsky and S.Papert, "Perceptrons", MIT Press,
02100	Cambridge, Mass, 1969.
02200	
02300	
02400	[Quam 1972] L. H. Quam, S. Liebes, Jr., R. B. Tucker, M.  J.  Hannah,
02500	B.  G.  Eross,  "Computer  Interactive  Picture Processing", Stanford
02600	Artificial Intelligence Laboratory AIM-166, April 1972.
02700	
02800	
02900	[Winston 1970] Winston, P.H.,"Learning Structural Descriptions from
03000	Examples", AI Technical Report 231, Artificial Intelligence Laboratory,
03100	MIT, September 1970.