Notice:
This poster is associated
with a Hands-On Demo
Haptic Teaching
using Opposite Force Presentation
Satoshi Saga* Naoki Kawakami* Susumu Tachi*
(*) The University of Tokyo, Japan
E-mail: [email protected],{kawakami,tachi}@star.t.u-tokyo.ac.jp
1. Abstract
We propose a haptic teaching system that uses haptic devices
devices to teach hand skills.
Specifically, we chose some tasks with pencilpencil-like devices and examined how to
teach an expert’
expert’s hand skills. We proposed a new haptic teaching method, in which
the haptic device produces force that is in the opposite in direction. The trainee tries
tries
to cancel the force and consequently, necessary force is ‘‘proactively
’’ generated.
‘‘proactively’’
Our hypothesis is that this ‘‘proactiveness
’’ is essential for haptic teaching. We
‘‘proactiveness’’
made a prototype system and compared our methods with existing teaching
teaching
methods.
Fig.1: Existing methods
2. Description
Expert’s Phase
Let FE and FT be output force of the expert and the trainee. ‘‘Proactive
’’
‘‘Proactive’’
operation means that a trainee outputs the force that is necessary
necessary for operation and
equal to the force FE. We propose the following methods to realize this situation.
Trainee’s Phase
Haptic Device
Haptic Device
Expert’
expert’s model motion and force FE are recorded. The
Expert’s Phase: The expert’
Freact
haptic device with a tool attached at the tip is used as a measuring
measuring machine.
Inverse
Internal
Force
Expert
Reaction
Force
FE Manipulate
Trainee
Finv
FT
Force
Trainee’
Trainee’s Phase: Following two methods are combined.
Virtual Trajectory
Trajectory
Fvf
Virtual Fixtures
A) Opposite presentation of expert’s force along trajectory (Finv)
• The haptic device generates the force that has the same amplitude
amplitude of the expert,
but in opposite direction.
direction. The trainee tries
tries to cancel the force and
consequently, necessary internal force is “proactively
“proactively”” generated.
Manipulate
Force
Fig.2: Proposed method
Finv = FE
FT = -F
Normal & Vertical
to the trajectory
B) The utilization of Virtual Fixtures [2] orthogonal to the trajectory (Fvf )
• Information of the expert’s motion is given (figure2
(figure2), and the pathway is used
as a ruler by using Virtual Fixtures technique.
Normal & Horizontal
to the trajectory
Fvf = - k (xT - xE) 3
Tangential
to the trajectory
Trajectory
Fdisp = Finv + Fvf
Correlation of Force History
0.7
Existing Method
Proposed Method
0.6
0.5
0.4
0.3
We recorded force history data while subjects wrote “mirror reversed”
reversed” Greek
letters with both the proposed and existing methods. First, to estimate
estimate the effect of
0.2
these methods, we calculate these correlations between the subject
subject’’s and expert’
expert’s
0.1
history data. Three components of the force (tangential, normal
normal vertical, normal
0
horizontal part(Fig.4)) were separately evaluated. The result is shown in Fig.5.
Through comparison, we found our method can transmit tangential and normal
horizontal force history to the trainee better than the existing method.
Fig.5:
4. References
[1] Gibson J.J.: Observations on active touch. Psychol.
Psychol. Rev, 69: pages 477477-491, 1962.
[2] L. Rosenberg. Virtual fixtures: Perceptual tools for telerobotic manipulation. In
Proc. of IEEE Virtual Reality Int’
Int’l. Symposium,
Symposium, pages 76–
76–82, 1993.
[3] Y. Kuroda, M. Nakao,
Nakao, T. Kuroda, H. Oyama,
Oyama, M. Komori and T. Matsuda,
“Interaction Model between Elastic Objects for Accurate Haptic Display
Display””,
International Conference of Artificial Reality and TeleTele-Existence (ICAT),
(ICAT), pages.
148148-153, 2003
[4] C. L. Teo, E. Burdet,
Burdet, and H. P. Lim. A robotic teacher of Chinese handwriting. In
Proc. of 10th Symposium on Haptic Interfaces for Virtual Environment
Environment and
Teleoperator Systems,
Systems, pages 335–
335–341, 2002.
proposed
exisiting
0.7
Tangent
Normal & V
Normal & H
Each Force Direction
Correlations of force history
0.36
0.35
Force Error(N/frame)
Second, we examine the learning effect of our proposed method and
and existing
method by recording force history data between the iteration of practices. The result
is shown in Fig6. We found our method achieves some result in learning
learning tangential
force and normal horizontal force.
NormalVerticalForce Error
0.75
Force Error(N/frame)
• [Existing method] (SlavedSlaved-Tracking Mode)
Mode)
Passively
Passively following the expert’
expert’s positional information as presented by the
haptic device.
Fig.4: Separating force
Fig.3: System overview
TangentForce Error
0.65
0.6
0.55
0.5
0.45
0
proposed
exisiting
1
0.34
0.33
0.44
0.32
0.42
0.31
0.3
0.29
0.28
0.27
0
1
2
3
4
5
6
7
Number of Trial(times)
2
3
4
5
Number of Trial(times)
6
7
Force Error(N/frame)
The proposed method was compared with the following existing method,
method,
through an experiment of letter writing tasks.
Correlation of Force History
3. Experiment / Result
NormalHorizontalForce Error
proposed
exisiting
0.4
0.38
0.36
0.34
0.32
0.3
0
1
2
3
4
5
Number of Trial(times)
Fig.6: Learning Curves of Each Force
6
7