00100	MINOR IMPROVEMENTS TO TRAJECTORY CONTROL
00200		INCREMENTAL SHAFT ENCODERS
00300		IMPROVED CALIBRATION
00400		VELOCITY CONTROL SERVO LOOP
00500		SAMPLED DATA SYSTEM ANALYSIS
00600	CONCENTRATE ON  INTERACTIVE MANIPULATION
00700		JOINT FORCE SENSING
00800			PLACE SENSES MOTOR TORQUES TO STACK LIGHTLY
00900			EXECUTE MOTIONS WHICH EXERT CONTROLLED FORCES
01000			BUILDING A WRIST WHICH MEASURES FORCE AND TORQUE
01100		TOUCH SENSING
01200			WE HAVE BUILT AND USED A 4X4 ARRAY AND SIMPLE SWITHCH
01300			TOUCH INTERRUPT, EXECUTE FUNARG, RESUME
01400			WILL GRASP IRREGULAR OBJECTS USING TOUCH
01500			WILL BUILD BETTER TOUCH SENSORS
01600		COMPLETE CONSTRUCTION OF SECOND ARM
01700			TWO-HAND ASSEMBLY TASKS
01800			ALLOWS EXPERIMENTATION WITH HANDS
01900		COMPLETE CONSTRUCTION OF FIVE-FINGER HAND
02000			ELECTRONICS AND SENSORS QUITE EXTENSIVE
02100			TO CONTROL,  MAKE A GRASPING MOTION IN SOFTWARE
02200			BY CLOSING EACH FINGER UNTIL IT ENCMAKES
02300			CONTACT, LIKE TOMOVIC'S LINKAGE.
02400	
02500	CENTERING ON TOUCH
02600		CLOSE UNTIL ONE JAW TOUCHES
02700		BACK UP THAT JAW VERY SMALL DISTANCE
02800		KEEP JAW WHICH TOUCHED FIXED, CLOSE JAWS
02900			MOTION OF CENTER = - HALF CLOSING DISTANCE
03000	
03100	PERFORM ASSEMBLY OPERATIONS
03200		PULL PIN FROM HOLE
03300		PULL SLEEVE FROM PIN
03400		PUT SLEEVE OVER PIN
03500	MOTION STUDY OF ASSEMBLY OPERATIONS
     

00100				ARM PRIORITIES-HARDWARE
00200	
00300	FORCE-SENSING WRIST
00400	TOUCH SENSORS
00500	TOOL HOLDER
00600	EXTRA DEGREES OF FREEDOM
00700		NEW HAND WITH IN/OUT AND INDEPENDENT FINGERS
00800		ANOTHER WITH LONGER FOREARM, EXTRA WRIST, INDEPENDENT FINGERS
00900	FORCE-SENSING ON OTHER JOINTS
01000	5 FINGER HAND
01100	2ND ARM
01200	SHAFT ENCODERS W/ SECOND TRACK (FREQUENT ZERO REF) AND POTS (STARTUP)
01300	DEBUG A/D OR POT NOISE
01400	OUR OWN A/D 14 BITS
01500	ALTERNATIVE DESIGN
01600		TORQUE MOTOR DRIVE (COMPLIANCE)
01700		DIFFERENTIAL WRIST
02200	TACHS ON JOINTS
02300	MORE SLIP RINGS ON FIRST ARM
02400	
02500	
02600			ARM PRIORITIES-SOFTWARE
02700	AVOIDANCE
02800	FORCE-CONTROLLED MOTIONS
02900	TASKS WITH TOUCH
03000	SAMPLED DATA SYSTEM ANALYSIS
03100	TOOL OPERATION
03200	PICK UP IRREGULAR OBJECTS
03300	SPEED UP TRAJECTORY CALCULATIONS
03400	IMPROVE CALIBRATION
03500	TWO-HANDED TASKS
03600	CONVERT TO PDP-11?
03700	
     

00100	ARM SOFTWARE-LOU PAUL
00200	
00300	REFERENCES
00400		PAUL IJCAI 1971 P. 385
00500		INOUE
00600		GREENE
00700		KULAKOV
00800		WHITNEY
00900	
01000	ARM SERVO OPERATIONAL JAN 72
01100		A 4K ASSEMBLY CODE SPACEWAR PROGRAM
01200		USING NEWTONIAN DYNAMIC MODEL OF THE ARM WHICH TAKES ACCOUNT
01300		OF GRAVITATIONAL AND INERTIAL LOADING
01400	TRAJECTORY PLANNING
01500		RANGE OF GRASP, UNGRASP
01600		DEPARTURE AND APPROACH VECTORS
01700			(USUALLY VERTICAL, BUT CAN BE SPECIFIED)
01800		MIDSEGMENT MATCHING POSITION AND VELOCITY
01900			5TH ORDER INTERPOLATION
02000		CHECK LIMIT STOPS, ADD INTERMEDIATE SEGMENT IF NECESSARY
02100	
02200	USER PACKAGE IN HAND/EYE SYSTEM (**GENERATE A WRITEUP)
02300		WAVE INTERPRETER
02400		USES FAST BANDS
02500		MANY MESSAGE PROCEDURES, CAUSES SOME CLANKING, BUT CLEAN
02600	
02700	PLACE
02800		IMPROVED JT 4
02900		LATER IMPROVE JT 5
03000		DEPENDS ON FORCE SENSING WRIST
03100	
03200	NON-LINEAR CALIBRATION
03300		ERROR IS ABOUT .1 INCH IN EACH COORD
03400		NOW READ 3 TIMES
03500	**	APPARENTLY BAD NOISE IN THE A/D OR POTS
03600	
03700	INCREMENTAL MOTION
03800	
03900	TRAJECTORY MODIFICATION AND STORED TRAJECTORIES
04000		LINEAR REMOVAL OF ERROR IN MIDPART OF TRAJECTORY
04100	
04200	SPEEDED UP TRAJECTORY CALCULATION BY A FACTOR OF 5
04300		2 SEC FOR A TRAJECTORY
04400	
04500	DISPLAYS FOR SERVO
04600		RUNS IN 7K WITH DISPLAYS
04700	
04800	INTERRUPT ON TOU
04900		GOES INTO WAIT STATE
05000		CAN RESUME TRAJECTORY
05100		GO MUCH SLOWER TO BE ABLE TO STOP WITHOUT DISTURBING
05200		COULD EXECUTE A FUNARG
05300	MODEL MAKER 
05400		GENERATES PROTOTYPES AND INSTANCES
05500		UPDATES DISPLAYS
05600	
05700	AVOIDANCE NOW UNDERWAY
05800	
05900	
06000	
06100	AVOIDANCE
06200	FORCE-CONTROLLED MOTIONS
06300	
06400	TASKS WITH TOUCH
06500		HANOI ON POSTS, AHARON
06600	PICK UP IRREGULAR OBJECTS (USE 3 OR MORE FINGERS)
06700	SPEED UP TRAJECTORY CALCULATIONS,
06900		STORED TRAJECTORIES
07000		NATURAL MOTIONS
07100		MACHINE LANGUAGE
07200		SPLIT UP PROGRAM
07300		USE INVERSE JACOBIANS TO FIND GRASP LIMITS
07400	
07500	SAMPLED DATA SYSTEM ANALYSIS
07600		EFFECT OF IGNORING VELOCITY
07700		EFFECT OF VIBRATIONS OF HARMONIC DRIVES
07800	
07900	IMPROVE CALIBRATION
08000	SHAFT ENCODERS
08100		INCREMENTAL
08200		2ND TRACK TO PROVIDE FREQUENT ZERO REFERENCE
08300		POTS FOR STARTUP
08400	
08500	TWO-HANDED TASKS AND COORDINATED ARMS
08600	
08700	USE ARM MODEL FROM BRUCE
08800	
08900	CONVERT TO PDP-11
09000		PREFER TO DO RESEARCH
09100	
09200	TOOL OPERATION
09300	
09400	
09500	DEMOS
09600		INSTANT INSANITY AND FILM
09700		TOWER OF HANOI
09800		WITH WINOGRAD'S BLOCKS PROGRAM
09900		
     

00100	ARM HARDWARE --- VIC SCHEINMAN
00200	ARM OPERATIVE DEC 1970 REFER AI 92
00300	
00400	5 FINGER HAND   17 DEGREES OF FREEDOM
00500		SEMI-ASSEMBLED
00600		TO DO:	LACKS IN/OUT MOTION FOR THUMB, REQUIRES SOME MACHINING
00700		WIRING, ENCODERS
00800		OPERATIONAL: MID 1972 OR LATER, NOT HIGH PRIORITY
00900	
01000	TOUCH
01100		ONE VERSION OPERATIONAL NOV 1971
01200		SENSITIVITY NOT GOOD ENOUGH, NO PERIPHERAL COVERAGE
01300		HIGHEST PRIORITY TO IMPROVE
01400	
01500	HAND WITH REDUNDANT DEGREES OF FREEDOM
01600		EXTENSION AND INDEPENDENT FINGERS PROVIDE TWO EXTRA MOTIONS
01700		TO BE OPERATIONAL WITH SECOND SCHEINMAN ARM
01800		TO DO:	MORE CONTROLLER ELECTRONICS NEEDED
01900		OPERATIONAL:	APRIL 1972 
02000	
02100	SECOND SCHEINMAN ARM
02200		MACHINING AND ASSEMBLY FINISHED
02300		TO DO: POTS, SWITCH BETWEEN ARMS, ELECTRONICS, INSTRUMENT THE HAND
02400		OPERATIONAL:	APRIL 1972
02500	
02600	FORCE SENSING ON WRIST
02700		MACHINING DONE
02800		TO DO:	MOUNT STRAIN GAGES, ELECTRONICS
02900		OPERATIONAL:	END JAN 1972
03000		HIGHEST PRIORITY
03100	
03200	SHAFT ENCODERS
03300		THE SECOND SCHEINMAN ARM WILL BE OPERATIONAL WITH POTS
03400		NOW LOCATING A SUITABLE DISK
03500	
03600	IMPROVE POTS ON FIRST SCHEINMAN ARM
03700		JOINT 1 POT WAS CHANGED.
03750	DEBUG A/D AND OUR OWN A/D
03800	
03900	TOOL HOLDER
04000		PNEUMATIC IMPACT WRENCH
04100	
04200	
04300	
04400	
04500	
04600	
04700	OTHER SCHEINMAN PROJECTS
04800	MIRROR ASSEMBLY FOR LASER EXPT
04900	
05000	LASER MOVED TO HAND/EYE TABLE
05100	
05200	INTERFERENCE FILTER TO FIT COLOR WHEEL OR CHANGE COLOR WHEEL TO FIT
05300	
05400	SERVO THE WHEEL ON THE HAND/EYE TABLE
05405	
05410	FORCE CONTROL OF ALL JOINTS
05415	
05420	BETTER RANGE AND INCREMENTAL MOTION BY LONGER FOREARM
05425		AND EXTRA WRIST, WITH INDEPENDENT FINGERS
05430	
05435	MORE SLIP RINGS FOR FIRST SCHEINMAN ARM
05440	
05500	
05600	
05700	LONG RANGE
05800	ALTERNATIVE DESIGN
05900	TORQUE MOTORS
06000		DISK BRAKES, LOW BACKLASH
06100		NOT USABLE FOR LAST JOINTS, NOT ENOUGH TORQUE
06200		SIGNIFICANCE:	TORQUE MOTORS ARE EXPENSIVE NOW,
06300		BUT THEY ARE POTENTIALLY COMPETITIVE IN PERFORMANCE WITH
06400		THE HARMONIC DRIVE POWER TRAIN.  AN ADVANTAGE IS HIGH
06500		COMPLIANCE WHICH ALLOWS SELF-PROTECTION AND PERMITS SIMPLE
06600		FORCE MEASUREMENT AT JOINTS.
06700		LOW PRIORITY
06800	
06900	DIFFERENTIAL WRIST JOINTS
07000		ALLOWS A SHORT 6TH JOINT, LOWER MASS AT THE TIP
07100		A DEFECT IS BACKLASH
07200		THERE IS USUALLY SOME COUPLING AMONG MOTIONS IN A DIFFERENTIAL
07300	
07400	MOVABLE TV CAMERAS
07500	XY TABLE
07600	BETTER GRID FOR TABLE
07700	
07800	
07900	SLIDES, ETC
08000		II FILM
08100		SCHEINMAN HAND
08200		NEW ARM
08300		TOUCH
08400		FORCE SENSING
08500	
08600	1972-1973
08700	TOUCH  SENSITIVITY AND PERIPHERAL COVERAGE
08800	FORCE CONTROL OF JOINTS
08900	5 FINGER HAND
09000		SHAFT ENCODER
09100	ALTERNATIVE DESIGN
09200	TOOL USING
09300	
09400	
     

00100	SOME LONG RANGE HAND/EYE PROGRAMS
00200	
00300		CURRENTLY, WE HAVE A GOOD PROGRAM IN MANIPULATORS.
00400	WE HAVE A STABLE, LOW COST ARM WITH A USER SOFTWARE PACKAGE.
00500	IN THE IMMEDIATE FUTURE, WE WILL BE DOING:
00600			AVOIDANCE OF OBJECTS
00700			TASKS REQUIRING JOINT FORCE SENSING
00800			TASKS REQUIRING TOUCH SENSING
00900	FOR THE LONG RANGE, THERE ARE SEVERAL THINGS THAT WE WOULD
01000	LIKE TO CONSIDER SERIOUSLY.  
01100		1  SELF PROTECTION REQUIRES THAT THE ARM STRUCTURE
01200		HAVE LOW JOINT STIFFNESS.  WITH A STIFF STRUCTURE,
01300		EVEN WITH TOUCH SENSING, THERE IS ONLY A FEW MILS
01400		TRAVEL IN WHICH TO STOP THE ARM BEFORE DAMAGE, SO 
01500		THAT PROTECTION IS IMPOSSIBLE.  WITH A COMPLIANT
01600		STRUCTURE, THERE MAY BE A HALF-INCH OR MORE OF
01700		TRAVEL AS THE STRUCTURE FOLDS, BEFORE DAMAGE. AS
01800		HANDS BECOME MORE COMPLICATED, THE RISK OF DAMAGE
01900		INCREASES.
02000		AT THE SAME TIME, JOINT FORCE SENSING BECOMES VERY
02100		EASY WITH COMPLIANT JOINTS.
02200			HOW TO MAKE LOW STIFFNESS ACTUATORS
02300			FOR ARMS
02400	
02500		2 TOUCH SENSORS ARE VERY POORLY DEVELOPED.  SENSITIVITY
02600		SPATIAL RESOLUTION, AREA COVERAGE FOR PERIPHERAL
02700		SENSORS AND MINITURIZATION OF ASSOCIATED ELECTRONICS
02800		AND MULTIPLEXORS ALL REQUIRE ABOUT AN ORDER OF MAGNITUDE
02900		IMPROVEMENT AS A FIRST TARGET.
03000		IMPROVED SENSORS WOULD ALSO BE VALUABLE FOR 
03100		PROSTHETIC DEVICES FOR AMPUTEES.
03200	
03300		WE SUGGEST THAT DR. KEN WISE IN DR. ANGELL'S LAB
03400		HAS INDICATED INTEREST IN THIS PROBLEM, AND
03500		THAT A RESEARCH EFFORT BE FUNDED THERE.
03600	
03700	
03800	
03900	
04000		PROGRESS WITH MANIPULATION
04100		THE NEW ARM BECAME OPERATIVE IN THE LAST FEW DAYS OF 
04200	1970.  BY THAT TIME, A SERVO ROUTINE WAS WORKING WHICH USED
04300	TRAJECTORY SERVO WITH A NEWTONIAN MODEL OF THE ARM.  INSTEAD
04400	OF POINT BY POINT PROPORTIONAL SERVO, THIS CONTROL ROUTINE
04500	FOLLOWS SMOOTH PREPLANNED TRAJECTORIES, TAKING INTO ACCOUNT
04600	GRAVITY AND INERTIAL LOADING, EQUIVALENT TO VARYING SERVO
04700	CONSTANTS THROUGH WIDE RANGES.
04800		A TRAJECTORY PLANNING PROGRAM DESIGNS TRAJECTORIES
04900	WHICH ARE USEFUL FOR MANIPULATIONS, AND WHICH ALLOW SPECIFYING
05000	APPROACH AND DEPARTURE ANGLES.  A USER PACKAGE
05100	PLANS FOR HANDS
05200	
05300	SOFTWARE			HARDWARE
05400	
05500	USER PACKAGE
05600	INSTANT INSANITY
05700	NON-LINEAR CALIBATION
05800	AVOIDANCE
05900	TASKS WITH THREE OR MORE FINGERS
06000		PICK UP IRREGULAR OBJECTS
06100	TASKS W/ FORCE FEEDBACK		JOINT FORCES
06200	  PUT PIN IN HOLE		  MOTOR TORQUES
06300	  TURN CRANK			  DIRECT SENSORS
06400	TASKS W/TOUCH FEEDBACK		TOUCH SENSORS
06500	  GROPING			  FINGERS
06600	  SHAPE OF OBJECTS		  PERIPHERAL
06700	  IRREGULAR OBJECTS
06800	TASKS W/ VISUAL FEEDBACK
06900	  STACKING
07000	  COORDINATION FOR GRASPING
07100	TASKS WITH VELOCITY CONTROL
07200	  THROWING
07300					5 FINGER HAND
07400					WRIST WITH INCREMENTAL MOTION
07500	
07600	THERE ARE A NUMBER OF PROBLEMS ASSOCIATED WITH THE INSTANT
07700	INSANITY TASK OF TURNING OVER OBJECTS.  THE PRESENT SOLUTION
07800	PROCEDURE CHOOSES THE FIRST POSSIBLE SOLUTION, LARGELY BECAUSE
07900	OF TIME CONSTRAINTS.  THE RESULTING SOLUTIONS TEND TO SPRAWL
08000	NEAR THE TABLE MORE THAN IS CONVENIENT FOR AVOIDANCE.  IT IS
08100	CLEAR THAT FOR AVOIDANCE, WE WANT TO TEND TOWARD SOLUTIONS
08200	WITH THE ARM OPERATING FROM ABOVE.  THE SIMPLEST FIX IS
08300	TO CHOOSE THE SOLUTION WHICH CAUSES APPROACH AND DEPARTURE
08400	ANGLES TO BE AS NEAR VERTICAL AS POSSIBLE.  FOR TURNING OVER
08500	A CUBE, THIS MEANS APPROACHING AT -45 DEGREES AND DEPARTING
08600	AT +45 DEGREES.  THOSE ANGLES ARE NOT ALWAYS POSSIBLE WITH
08700	THE ARM IN CERTAIN POSITIONS.
08800	IF WE TRIED ALL FOUR POSSIBILITIES FOR PICKING UP A BLOCK BY
08900	ITS FACES, WE MIGHT FIND MUCH BETTER SOLUTIONS FROM THE
09000	STANDPOINT OF VERTICALITY, YET WE CANNOT AFFORD 4 SEARCHES
09100	THROUGH LIMITS OF APPROACH AND DEPARTURE ANGLE.  WE MIGHT INSTEAD
09200	SUGGEST THAT FOR THAT TASK, WE ARE ESSENTIALLY ORIENTING THE
09300	WRIST WITH RESPECT TO THE TABLE, AND FOR ORIENTING THE WRIST,
09400	THE WRIST ROTATIONS ARE RELEVANT.  THEN WE WANT TO MOVE
09500	THE BLOCK TO AN ANGLE WHERE THE ROTATIONS ARE EASY, AND MAKE
09600	THE RIGHT WRIST ROTATIONS.  BEYOND THAT, WE CAN THINK OF
09700	HAVING A MODEL OF THE ARM, AND PIECE TOGETHER TRAJECTORIES
09800	FROM THE NATURAL MOTIONS OF THE ARM.  FOR THE TIME
09900	WHEN WE GET REDUNDANT DEGREES OF FREEDOM, THAT WILL BE
10000	MORE OR LESS NECESSARY.  WITH AN ARM WHICH HAS ONLY THE CANONICAL
10100	SIX DEGREES OF FREEDOM, WE WOULD INVESTIGATE AN ARM MODEL
10200	AS A WAY OF FINDING NOT ONLY THE FIRST SOLUTION, BUT SOME
10300	GOOD SOLUTIONS AND PERMIT CUTTING COMPUTATION OF TRAJECTORIES
10400	SOME LONG RANGE HAND/EYE PROGRAMS
     

00100	LOU PAUL
00200		1  INCREMENTAL MOTION
00300			DONE
00400		   IJC PAPER
00500		   INSTANT INSANITY CODE
00600		2  NON-LINEAR CALIBRATION
00700		3  OBSTACE AVOIDANCE
00800		4  FORCE FEEDBACK (CONSTRAINED MOTIONS AS DONE BY INOUE)
00900		5  TACTILE FEEDBACK (GROPING AND SHAPE)
01000		6  FREQUENCY RESPONSE OF SERVO
01100	
01200	
01300	
01400		THESE REQUIRE THE FOLLOWING:
01500		DEBUG HARDWARE (FOR NON-LINEAR CALIBRATION)
01600			A/D NOISE
01700			A/D DONE FLAG
01800			DELAY FOR SETTLING ON FIRST SAMPLE
01900			136 PROBLEMS
02000				VALUES SHIFTED BY 6
02100				ZERO IN PLACE OF "DATA MISSED"
02200	
02300		3D MULTI-ENTRY (FOR OBSTACLE AVOIDANCE)
02400		TOUCH SENSORS (FOR TACTILE FEEDBACK)
02500		CURRENT CONTROL OF ARM
02600		FORCE SENSING ON JOINTS (PROPRIOCEPTORS)
02700			WOULD LIKE A WRIST LIKE THE MIT OR SCHEINMAN
02800			WRIST WITH FORCE AND TORQUE SENSING.  FOR NOW,
02900			THE CURRENT CONTROL OF THE ARM WOULD GIVE ACCURATE
03000			MOTOR TORQUES AND PERMIT USE OF THAT INFORMATION.
03100	
     

00100		The following hardware plans require technician help:
00200			instrument new TV camera
00300			debug TV digitizer
00400			modify target voltage A/D to read actual voltage
00500			build new log difference digitizer (Fairbairn)
00600	done?		debug multiplexer and A/D for arm
00700			touch sensors
00800			update kluge bay drawings
00900			current controller for arms
01000			build Quam arithmetic unit
01100			underflow mod to PDP-10
01200			new arm safety device
01300			second arm
01400			movie making
01500			new hand and wrist with incremental motion
01600				and independent fingers
01700			new 5 finger hand 
01800	Estimate time required for these things:
01900			instrument TV  3weeks
02000			debug TV digitizer 1 week
02100			log difference digitizer 6 weeks
02200			debug multiplexer and A/D 2 weeks
02300			touch sensors 4 weeks
02400			update kluge bay drawings
02500			current controller for arm 4 weeks
02600			Quam arithmetic unit 16 weeks
02700			underflow mod    
02800			new arm safety device 2 days
02900			second arm 12 weeks
     

00100		We discuss the intended development of touch sensors
00200	for manipulation tasks of the hand/eye system.  For our
00300	discussion, consider an artificial skin with force sensors
00400	embedded in a deformable material, in an
00500	array of about 10x10, with sensitivity to light
00600	touch.  The issues we will discuss are:
00700			1)specifications
00800			2)feasibility with existing techniques
00900			3)incremental utility; what can we do with a 
01000			small amount of additional software along current lines
01100			of effort
01200			4)long range utility; what will we do toward complex tasks
01300			which will be greatly aided by touch sensing.
01400	
     

00100	SPECIFICATIONS
00200		To be able to sense mis-alignments 
00300	in assembly without disturbing structures, a sensitivity
00400	at least as good (and hopefully 10 times better) as to 
00500	sense a penny on a fingertip --1 gm per sq cm pressure
00600	is a useful target.
00700	
00800		we require only crude proportionality and linearity,
00900	and we can tolerate drift (or so it seems now) because
01000	we can compensate for level shifts by modifying 
01100	thresholds on a one second basis
01200	
01300		We can use the computer to do simple processing on the
01400	direct output from each sensor, so that we can tolerate
01500	considerable variation in response characteristics among
01600	the sensors and their calibration.
01700	
01800		The sensors must be intrinsically self-protected
01900	from large loads by some force-limiting mechanism or by an intrinsic
02000	very high maximum force limit.  Self-protection is a problem that we
02100	can deal with in only simple terms now, and there are obviously problems
02200	with sharp objects cutting through a soft finger and severing individual
02300	sensors or connections of several sensors.  For the moment, we guess 
02400	that simple design considerations to protect large nerves, and easy 
02500	replacement with low individual cost is a satisfactory solution.
02600	
02700		The sensors must be small, nominally smaller than 100 mil 
02800	centers.  Sensor densities such as this allow good positional location
02900	and description of objects.
03000	
03100		To manipulate complex objects, the fingers themselves should
03200	be deformable to provide gripping surfaces in contact.  This implies
03300	requirements on the sensors and their location.  If the sensors are
03400	anchored on a firm substrate deep under the skin, the intervening
03500	soft material will blurr force location considerably.
03600	That will degrade sensitivity and positional resolution.  A reasonable
03700	requirement is that the sensors be on the surface of the skin.
     

00100	TECHNICAL FEASIBILITY
00200		We have discussed the development of sensors with
00300	Dr. Ken Wise of Dr. Angell's laboratory .  Their laboratory
00400	has developed for medical use a small pressure sensor for
00500	catheter implant.  This sensor is a 60 mil diameter diaphragm
00600	etched from a silicon wafer, with a bridge of etched 
00700	piezoresistive elements to detect strains.  The bridge 
00800	design gives good stability and sensitivity, partially 
00900	compensating for temperature shifts.  These integrated circuit
01000	devices are fabricated in their laboratory.
01100		We had originally discussed using a flat array of sensors,
01200	with a somewhat simpler design  
01300	than the diaphragms used for pressure sensing.  However,
01400	it seems clear that what we really want are deformable fingers
01500	with sensors in the surface, and we must come up with a different
01600	design geometry.
01700		The integrated circuit technique seems to have adequate 
01800	sensitivity (.5 mv for 1 gm/sq cm which we feel can be considerably
01900	increased by larger currents.  Also we can detect with narrow band 
02000	phase sensitive AC detection).  The elements are sufficiently small
02100	and we hope replacement costs to be reasonable with integrated 
02200	circuit fabrication.  
02300		We do not yet have a sensor which would function in the 
02400	surface of skin.  Perhaps the pressure sensing diaphragms themselves
02500	would be adequate.
02600		We have examined other techniques and find none with promise
02700	of acceptability.  These include piezoelectric devices (with ac 
02800	detection, these are interesting but difficult technology),
02900	magnetostrictive devices (sensitivity is marginal and cost is high)
03000	and semiconductor strain gauges.
03100		A useful design consideration for any of these techniques
03200	is analog multiplexing, such as developed by Fairchild for their
03300	self-scanning photodiode arrays.
     

00100	IMMEDIATE UTILITY
00200		In discussing utility, we could take an extreme stand and
00300	say that touch sensors are unnecessary and that we could do a great 
00400	deal with just visual feedback, as can be easily demonstrated with
00500	prosthetic devices.  Vision has been difficult and we reasonably show
00600	in the following discussion that interesting tasks will require only
00700	simple use of touch sensing.  We could as well say that we could do 
00800	many of the current tasks without vision.  Instead, we want to use 
00900	each independent mode of information.  In some cases, only one mode
01000	such as touch, will be capable of giving needed information, and 
01100	in that sense will be necessary.  In the usual case, there will be
01200	redundant information which will be useful for confirmation and 
01300	discovering errors.
01400		To consider an immediately useful task, we want to use 
01500	touch sensing for centering in grasping.  The ability to do so 
01600	would lessen our need for absolute visual and arm calibration 
01700	and increase reliability and adaptability.  The basic operations
01800	are to sense touch on one side, use incremental motion to move 
01900	to touch the other side while holding the first side steady.
02000	Incremental motion already
02100	exists on both arms; however, it would be much more useful with
02200	our planned wrist with redundant degrees of freedom.
02300		Locating edges of objects requires existing techniques
02400	for edge finding.
02500	These are reasonably simple if high sensitivity is not needed 
02600	and if the pattern on the finger is simple.  In general, processing
02700	the pattern on the array of touch sensors raises difficult problems.
02800	The task above required only asking am I touching anything?  
02900	This task requires asking is there an obvious edge?
03000		Coordinating these two simple operations, yes/no and edge
03100	detection, with incremental motion allows feeling along a surface
03200	and feeling along an edge.  As mentioned above, incremental motion
03300	already 
03400	exists.  This might be called the one finger paradigm by which we
03500	can tell whether a surface is flat or curved, or an edge is straight
03600	or curved by moving along it.  
03700		The same techniques can be used in simple object description
03800	by the two-finger paradigm.  For this, we use two fingers to get cross
03900	section thickness along the length of an object.  This is very close 
04000	to the representation for three-dimensional objects as used by 
04100	Binford and Agin, and provides a considerable amount of descriptive 
04200	information on shape, with obviously simple and well-defined
04300	manipulations.
04400		The edge detection and incremental motion can be used for
04500	aligning stacks of blocks.
04600		Some simple self-alignment tasks for reliability can be done
04700	with simple touch sensing.  We would like very much to sense when we
04800	touch the table, and to refer positions relative to temporarily
04900	stationary objects.
05000	
05100	
     

00100	LONG-RANGE UTILITY
00200		We have dealt with simple objects, blocks and wedges;
00300	now we aim for complex objects with curved surfaces.  Shape
00400	description by touch is a useful extension of our abilities
00500	which would provide independent criticism of our visual judgment.
00600	We have shown some simple shape description techniques which
00700	would be useful with touch sensing.  Now we would like to
00800	mention groping and exploration by touch.  At first, it 
00900	seems conservative to class simple groping with long-range
01000	developments, but groping requires sensors on the external
01100	parts of the hand and, further, on the arm. An exploration 
01200	sequence, such as a spiral motion with diversion around any found
01300	object would be adequate for simple spatial sense and not difficult
01400	to program. It is not clear that such simple procedures by 
01500	themselves are adequate, but coupled with vision to be sure
01600	that exploration is done for very simple environments, can be
01700	useful.
01800