13/06/2008

1. Using vibration patterns to provide impact position information in haptic manipulation of virtual objects 13/06/2008 Jean Sreng Anatole Lécuyer Claude Andriot jean.sreng@inria.fr

2. Outline • Introduction • Vibration patterns • Evaluation • Conclusion OUTLINE

3. Introduction • The impact between “rigid” objects : • A low-frequency reaction force of contact (simplified model) • A high-frequency force transient (more complex model) • This high-frequency force transient can be used in haptic rendering of impact (open-loop haptic) : • Improve the realism • Convey some information about the object’s material • Is-it possible to convey position information using this transient ?

4. Vibrations and impact position • Chosen approach : • The impact generates vibrations depending on the impact position • Vibration model, a first approximation • Euler-Bernouilli cantilever beam

5. Vibrations and impact position • General solution : Impact position

6. Simplified vibration patterns • Simplified patterns based on the physical behavior • Simplified computation • Maybe easier perception ? • Chosen model : exponentially damped sinusoid • Amplitude changes with impact position • Frequency changes with impact position • Both amplitude and frequency changes

7. Simplified vibration patterns Am Fr AmFr (Consistent) AmCFr (Conflicting) Far impact Near impact

8. Evaluation • Preliminary evaluation : • Quantitative : “Between these two impacts which one was the closest one from the hand ?” • Qualitative : Subjective rating of the impact realism • Conducted on 15 subjects • Apparatus : • Virtuose6D device • Sound blocking noise headphones

9. Quantitative evaluation ERRATUM Page 7 § 4.3 EB1: 10Hz, 64Hz EB2: 5Hz, 25Hz Other: 15Hz to 45Hz • Quantitative evaluation : “Between these two impacts which one was the closest one from the hand ?” • 6 models • 2 realistic models (Euler-Bernoulli) (EB1, EB2) • 4 simplified models (Am, Fr, AmFr, AmCFr) • 4 locations • 8 random repetitions • Total of 576 trials (40 min)

10. Quantitative evaluation : results • The ratio of “good answers” was evaluated : “How well the user was able to associate a vibration and an impact location ?” 0.25 0.5 0.75 No association association Inverted association • The value was used to represent the overall performance 0.25 0 0.5 No association association

11. Quantitative evaluation : results • Overall performance : • ANOVA Significant (p < 0.007) • Paired t-tests (p < 0.05) : • Am - • EB1 • EB2 • AmCFr • Fr - • EB1 • EB2

12. Quantitative evaluation : results • Inversion ratio : The ratio of users who inverted the association between the vibration pattern and the position

13. Qualitative evaluation : results • Rate the impact realism : • Paired t-tests : • Am • EB1 • EB2 • Fr • AmFr • EB2 • Fr • AmCFr

14. Discussion • Global weak inter – subject correlation : • Each subject seems to have his/her own interpretation (inversion or not) • Strong intra – subject consistency : • Subjects seem to be very consistent within his/her interpretation • Several strong inter – subject correlation between models : • Fr and AmCFr strongly correlated

15. Conclusion • We proposed several vibration patterns to convey impact position information • We conducted a preliminary evaluation : • The user is able to use the vibration pattern to get an impact position information • The simplified models Am and Fr seems to perform better • Future work : • Evaluate the role of the hand impedance and the haptic device bandwidth to elaborated more effective feedback • Deeper analysis of the data • New models

16. ? Questions ?