RateHardness : A New Performance Metric for Haptic Interfaces

1
Rate-Hardness A New Performance Metric for
Haptic Interfaces

• Jaeyoung Cheon
• icejae02_at_postech.ac.kr
• VR Lab, POSTECH
• 2006. 5. 1

2
Outline

• Introduction
• Hypothesis
• Experiment Description
• Apparatus
• Testing Protocol
• Analysis of Results
• Main Results
• Additional Issues
• Conclusion

3
Introduction

• Question of Interest
• What makes a virtual surface feel hard?
• Performance Measurement
• Lowest Level Technical Quantities
• Easy to quantify in technical terms
• Hard to establish perceptual value to a user
• Highest Level User Performance
• Easy to compare between existing systems
• Hard to correlate the result with the underlying
  technical capabilities

4
Introduction

• Specific Question of Interest
• Which physical attributes of surfaces can users
  perceive as causing differences in apparent
  hardness?
• Applications
• Optimizing haptic interface components to improve
  perceptual performance
• Reducing the cost of haptic interfaces

5
Hypothesis - Observation

• Experiment Environment
• 3-DOF Haptic interface
• Tapping Task
• Stiffness 15003000 N/m
• Measurement
• Force Strain gauge sensors (a resolution of
  0.03N)
• Position Optical encoder (a resolution of 7 um)
• How can people haptically distinguish high levels
  of stiffness?

6
Hypothesis - Observation

• Comparison of Surface 1 4
• Stiffness
• Surface 1 gt Surface 4
• Perceptual Hardness
• Surface 1 gt Surface 4
• Its easy to expect that difference of
  penetration depths makes the human discriminate
  surfaces.
• But the human has low position sensing resolution.

7
Hypothesis - Observation

• Comparison of Surface 1 3
• Stiffness
• Surface 1 ? Surface 3
• Perceptual Hardness
• Surface 1 gt Surface 3

Surface3
Surface1
8
Hypothesis - Observation

• Rates of force change after penetration are quite
  different.
• There is ample evidence that fast rate of change
  in skin deflection or force can be sensed.
• Can the rate of force change be discriminated?
• Does the rate-of-change characteristic dominate
  stiffness in the psychophysical judgment of
  surface hardness?

9
Hypothesis Rate-Hardness

• Perceptual hardness of the virtual surfaces is
  more closely correlated with the surface
  rate-hardness HR than with the surface stiffness
  K, where HR is defined by
• HR initial force rate of change (N/s) initial
  penetration velocity (m/s)

10
Hypothesis Rate-Hardness

• Measurement
• Rate-hardness by recording force and position
• Stiffness by static relation between measured
  position and commanded force
• Surface Comparison of 12, and that of 34
• Similar rate-hardness
• Different stiffness

11
Experiment Description Experiment Apparatus

• Virtual surface is vertically oriented.
• Surface properties are separately specified in
  the right and left halves of the vertical virtual
  plane.

12
Experiment Description Experiment Apparatus

• No torques are produced.
• Operators elbow resting on a padded table top.

13
Experiment Description Experiment Apparatus

• The system is partitioned into
• Human operator dynamics
• Haptic interface mechanical and electrical
  dynamics
• Digital control dynamics
• Za The impedance at the point where the fingers
  touch the interface grip
• Fs The force provided by the haptic interface at
  the point where the fingers grip
• Zh The equivalent impedance of the human
  hand/arm
• Fhe The equivalent force provided by the
  operator to cause motion
• Zm The impedance which describe mechanical and
  electrical dynamics of haptic interface
• Zc The impedance specified by the digital
  controller
• Fce An external force provided by the control
  system
• T A sample period ( T1.2 ms, 833Hz)

14
Experiment Description Testing Protocol

• Surface 13, and 24 provide the same
  low-frequency gain.
• Surface 12, and 34 provide the same
  high-frequency gain.
• There is close correspondence between
  rate-hardness and the peak impedance magnitude
  achieved by the control laws shown in Fig. 6.

15
Experiment Description Testing Protocol

• Instructions to subjects
• To probe the pairs of surfaces at will
• To avoid changing finger grip or arm pose
• To respond with the words Left, Right, or
  Same
• A test suite of 16 distinct side-by-side
  combinations
• Presented randomized order
• Constituted one block of trials
• Each subject was given three consecutive blocks
  of trials.
• 49 volunteer subjects.
• 33 males and 16 females.
• Ranging in age from 18 to 37.
• Each subject was given a standardized training
  session.
• All subjects wore ear protectors.

16
Analysis of Result

• Score of responses
• Different surface
• 0 (none), 1 (one), 2(both)
• Same surface
• 0 (unable to discriminate), 1 (correct)
• Three Groups
• (a, b) surface a on the left, surface b on
  the right
• Group 1 contains pairs that differ in
  rate-hardness.
• (24), (13), (14), (23)
• Group 2 contains pairs which differ in stiffness.
• (34), (12)
• Group 3 contains pairs of identical walls.
• (11), (22), (33), (44)

17
Analysis of Result Raw Data
18
Analysis of Result - Normalized Personal Score

• Wall Group1 It is reliable to distinguish.
• Wall Group2 It is difficult to distinguish.
• Wall Group3 It is not reliable rather difficult.

19
Analysis of Result - Response Time

• Wall group 1 responses are consistently faster
  than other groups.
• This indicates that Wall group 1 surfaces were
  easier to distinguish.

20
Analysis of Result- Percentage of Correct
Response

• Percent Correct versus Rate-Hardness Difference
• Positive correlation
• Percent Correct versus Stiffness
• Negative correlation

21
Analysis of Result - Additional Issues

• There are vary wide variations in individual
  subject personal scores.
• 10 of the subject had persistent difficulty
  distinguishing between wall combinations with the
  largest difference.
• Considerations
• Gender Differences
• Right-Handed Versus Left-Handed Users
• Left/Right Bias
• Time to Respond

22
Analysis of Result - Gender Differences

• Wall groups 1 and 3, men had higher personal
  scores.
• Success rates between men and women would
  converge.
• Womens scores show continued improvement through
  the third trial block.

23
Analysis of Result - Left/Right Bias

• Top figure is the result of overall 49 subjects.
• Bottom figure is the result of the 10 subjects
  who had 50 or fewer correct responses in Wall
  group 1 in at least one of the three trial block.
• There is a bias that left surface is harder.
• Top figure shows a slight bias.
• Bottom figures shows a strong bias.

24
Analysis of Result - Time to Respond

• The faster responses tend to come from those
  subjects who were more correct.
• Some subjects seem to be genuinely unable to
  discriminates surfaces which are quite distinct
  for the majority of the test population.

25
Conclusion

• Hard virtual walls are not necessarily the result
  of surfaces with high stiffness.
• Rate-hardness can effectively substitute for
  large stiffnesses in peoples perception of wall
  hardness.

26
Supplement

• From Bolanowski (1988)

• Sensory Adaptation

PC
NP I
27
References

• D. A. Lawrence, L. Y. Pao, A. M. Dougherty, M. A.
  Salada, and Y. Pavlou, "Rate-Hardness A New
  Performance Metric for Haptic Interfaces," IEEE
  Transactions on Robotics and Automation, vol. 16,
  pp. 357-371, 2000
• S. J. Bolanowski, Jr., G. A. Gesheider, R. T.
  Verrillo, and C. M. Checkosky, "Four Channels
  Mediate the Mechanical Aspects of Touch," Journal
  of Acoustical Society of America, vol. 84, pp.
  1680-694, 1988.