COMMENT ⊗   VALID 00002 PAGES 
RECORD PAGE   DESCRIPTION
 00001 00001
 00002 00002	CORRECTIONS AND ADDITIONS TO AI PAPER.   NUMBERS  CORRESPOND  TO  THE
 00019 ENDMK
⊗;
CORRECTIONS AND ADDITIONS TO AI PAPER.   NUMBERS  CORRESPOND  TO  THE
NOTED LINES ON THE PAPER.

	1) Hydraulic power is characterized by high power levels in a
small volume and continuous high stall torque capability with  little
actuator  heating.    However,  the  interface  hardware, such as the
motor driven pump, servo valves, and hoses are  bulky,  difficult  to
easily  integrate  into  a compact configuration, and also expensive.
Electric power is characterized by a simple ,all  electric,  computer
and  power  source  interface at the expense of lower power to weight
ratios and motor stall torque duty cycles limited by heating  .   The
wide variety and availability of electric motors make them attractive
for rotary joints.  The size and loading of the JPL manipulator  does
not  require  high  power,  suggesting the use of electric motors for
this application.

	2) To properly employ  this  type  of  motor,  it  should  be
matched  to  a low driven load inertia.      In addition, specialized
electronics are required  to  obtain  acceptable  performance  levels
(power/weight).

	3)  but, as the arm moment of inertia and gravity torque vary
considerably with the configuration of the joints, it is necessary to
make  an  estimate  of  the  average  inertia and gravity torque load
handled by each joint when selecting the proper reduction ratios.


	This  next section is to be added as a description.   You can
put it under Introduction, or maybe another section  like  background
or  furthur work.Feel free to change things a little to make them fit
in better.


	The  JPL  manipulator  is  an  outgrowth  of  the manipulator
designed at  Stanford  University  and  used  exclusively  for  their
Hand-Eye  research  work. The original manipulator has been operating
reliably for about 3 years with few maintainance or design  problems.
Several  new  versions  of the original design have been built. These
reflect new developments in the areas  of  improved  electronics  and
mechanical  hardware.     A brief description of some of the research
and development work being carried out at Stanford University follows
under 4 main categories: Drive trains, Structure, Feedback, Control.

	Drive Trains:

	The Stanford arm was one  of  the  first  small  manipulators
designed  to use the USM Corporation's Harmonic Drive.   This compact
high reduction unit exhibits high torque capability, high efficiency,
low  backlash and a simple in-line mounting arrangement. The original
arm featured a low inertia DC motor directly coupled to the  harmonic
drive  with  a slip clutch mounted on the output shaft. This was done
because calculations indicated that should the arm run into  a  solid
object  at high speed, the angular acceleration reflected through the
harmonic drive could cause damaging overloads.  In  actual  practice,
this  condition  was rarely encountered.  At the expense of increased
inertia, the slip clutches have been replaced  with  larger  Harmonic
Drives giving an adeguete margin of safety.   Besides backlash, which
controls the maximum gain and dead  band  amplitude  on  an  unloaded
joint,  the  joint  stiffness determines the natural frequency of the
arm.   The lowest spring constant element in each link  is  generally
the  Harmonic  Drive.  Thus, there is a trade-off between the inertia
contribution of a larger harmonic drive and reduced windup (increased
stiffness).	Much  thought has been given to motor selection.  The
original design used conventiional,  permanent  magnet  d.c.   motors
characterized  by  high  speeds  and  low  continuous  stall torques.
Presently, these motors are being  replaced  by  high  torque,  lower
speed  "torque  motors"  which  give better performance but require a
controller with current limiting to prevent demagnetization of  their
strong  permanent  magnet fields.  These torque motors permit the use
of high efficiency, low reduction ratio (<100/1) transmissions  which
means  that  the  joint  is  easily  back drivablle.  This allows for
torque control of the joint enabling sensing of loads and forces.

Present studies center in the area of ultra high performance electric
motors  and  very high torque capability precision transmissions with
the goal of approaching the performance of  hydraulic  actuatorswhile
maintaining the simplicity of an all electric manipulator.

		The  arm  design is characterized by the placement of
the drive elements at or  physically  in  the  joint  that  is  being
operated.  This places a gravity load and increased moment of inertia
on the shoulder joints in a manipulator of this configuration.  Other
alternatives  such  as  cable  or  band  operated  joints  are  valid
possibilities, but the increased  complexity  created  by  the  cable
guides  and  tensioning  elements  seems  to  more  than  offset  the
advantages of their potentially high performance

	Structure:

	In   the  original  design,  care  was  taken  to  produce  a
configuration  which  could  be  easily  manufactured  in   prototype
quantities  using the machine shop facilities at Stanford University.
Most of the components are lathe  turned  parts  with  a  few  simple
milling  operations required.  No castings, or complicated setups are
required by the design.  Large diameter widely spaced bearings on all
rotary joints ease the machining tolerance requirements and provide a
stiff and rugged structure by allowing large shaft  sections.     All
components  are  machined  from aluminium with the execption of a few
small motor related components which are  made  of  stainless  steel.
This  speeds  the  machining  process, produces a corrosion resistant
manipulator, and allows the use of a colored anodize finish.  As  all
the  sections  are  holllow  tubular  members,  little  shrouding  is
necessary to hide or protect components.   The prismatic joint is  an
unmachined drawn square section aluminium tube carefully selected for
straightness and freedom of flaws.  This boom rolls  on  16  phenolic
faced  bearings  which guide it while preventing rotation or play. As
mass in the shoulder results in relatively small increases in overall
arm  inertia,  shoulder  sections are heavy and sturdy to provide the
necessary rigidity to  eliminate  the  need  for  structural  bending
calculations in the solution programs.

	Feedback:

	Traditionally, arms of this type have been  designed  without
much  consideration  to  the  feedback elements.  While it may not be
much of an inconvenience to mount a pot on a shoulder  joint,  it  is
not  an  easy matter to incorporate a pot into an outer joint without
it getting in the way, or else being vunerable to damage.

	For this reason,the original arm was designed to use integral
pots  mounted  directly  on  each  joint.  There are no set screws,no
gears, no mounting clamps, or limit stops; just a  resistive  element
and  a  wiper cemented to the joint.  The resolution is high, and the
repeatability good as the coupling between pot and  joint  is  rigid.
It  was  thought  that  by having large diameter pots, high linearity
could easily be obtained.   This is true, but experience  shows  that
it  is  a  much  simpler  matter to plot a curve of of pot output vs.
displacement, store it in the computer and then use  this  correction
table  when reading the pots to get the actual values. This procedure
has been successfully used with the Stanford Arm.

	Recent experiments with optical encoder  components  indicate
that  it  is  now  practical  to  build  reliable compact incremental
encoders into each joint as an integral unit.   For a long  time  the
price  of  such devices, when coupled with suitable read and counting
electronics, was high.   Now, with improved electronics  and  a  more
competitive  encoder  industry  the  use  of integrated encoders is a
reality.   The latest Stanford arm will have such encoders shortly.

	For a  long  time,  velocity  information  was  developed  by
differentiating the position feedback signal.  This resulted in jerky
motion, especially when very slow trajectories were commanded. Adding
analog tachometers to each joint has greatly improved the performance
of the arm.  Not only does it produce greater  smoothness,  but  also
allows  an  increased  gain,  with  resulting  improved  accuracy and
shorter servo time.   Here  too,  these  analog  tachometers  can  be
replaced by digital type optical tachometers.


		Control:

	The Stanford arm operates in a trajectory control mode.  This
means that every motion is precalculated to produce a terminal device
motion which is controlled in position and velocity vs. time.    This
desired trajectory produces acorresponding trajectory requirement for
each  of  the  6 arm joints. The performance properties of each joint
are known and predicted for each arm configuration,  with  individual
joint  velocity  or  torque  limits  dictating  the  maximum  overall
performance. Deviations from the  predicted  drive  requirements  are
interpreted  as  external  force imputs on the arm.  This information
can be used to perform tasks such as tightening  screws  to  a  given
torque,  pushing  objects  along a table surface, turning a crank, or
stacking blocks where the stack height is  not  known.  More  precise
force sensors are being developed at the present time.  This includes
a 6 component wrist  mouunted  force  balance  and  individual  force
sensors  in  the  fingers.  Tactile feedback in the form of a one bit
signal has been looked at, but present work is centered around  using
discrete  force sensors for both tactile and local force feedback. As
an example:  a finger with 8 discrete analog force sensors  has  been
developed which allows a detailed description of the contact surfaces
of a grasped object and the forces exerted on it.