[VRST 2017] Computational Design of Hand-Held VR Controllers Using Haptic Shape Illusion
1. Eisuke Fujinawa1
Yuki Koyama2
Tomohiro Tanikawa1
Shigeo Yoshida1
Takuji Narumi1
Michitaka Hirose1
1The University of Tokyo
2National Institute of Advance Industrial
Science and Technology (AIST)
Computational Design of Hand-Held VR Controllers
Using Haptic Shape Illusion
4. ‣ cannot provide proper haptic perception
‣ leads to lack of immersion inappropriate object handling
Typical Controllers
5. ‣may not fit into limited room space
‣can be dangerous against surrounding people
Naively Designed Controllers
6. Previous Work: Nonlinearity of Human Haptic Perception
Force direction
[Amemiya et al. 2005]
Haptic Perception
Stimulus
Perception is nonlinear to Stimulus
7. Previous Work: Shape Perception Through Haptic Cue
Shape Perception
[Tuvey et al. 1996]
Haptic shape perception consists of moments
the moment of inertia
I = ∫ mr2
9. Target VR Controller
A hand-held VR controller that
‣gives haptic shape perception of VR object
‣is smaller than visualized
‣is customized for each VR application
‣is limited to thin symmetric planar for simplicity
12. Shape Perception Model
f : (moments) → (perceived length,width)
moments perceived shape
f -1: (perceived length,width) → (moments)
13. Data Driven Approach
[Lau et al. 2016][Umetani et al. 2014]
Physical Property Shape Semantics
We take data-driven approach to predict shape perception model
17. Contribution
(1) a novel design concept that uses a haptic shape illusion
(2) a data-driven representation of a perceived shape based on
the mass properties of a wielded object
(3) an interactive optimization-guided hand-held VR controller
design tool
19. Overview of Data Collection Experiment
sample controllers
(w/ various moments)
perceived shape
(length/width)
wield and estimate in VR
21. Condition of Experiment
controllers (moments)
16 (x 4 times each)
participants
10
perceived shape
640=
we obtained 640 evaluation data pairs (moments, perceived shape)
x
22. Regression of Shape Perception Model
Linear
Regression
Quadratic
Regression
Gaussian
Process
Regression
Error in L
[mm]
117.5 116 115.9
Error in W
[mm]
99.2 98.5 97.9
we use linear regression model for simplicity
30. Target virtual model
size
constraint
handle
Deformation Weight Carving
CAD data of designed
VR controller
Optimization of Perceived Shape
Target Shape
(target length,width)
Current Perceived Shape
f (moments)
Target virtual model
size
constraint
handle
Deformation Weight Carving
E = argmin || f (moments) - (target length,width) ||2
minimization for moments of the controller model
31. Optimization of moments
moments of inertia
[Beacher et al. 2012]
center of gravity
[Prevost et al. 2012]
optimize moments of 3d model by weight placement and inner carving
Target virtual model
size
constraint
handle
Deformation Weight Carving
CAD
V
32. 2. Weight Placement Optimization
cf. [Baecher et al. 2012] [Prevost et al. 2012]
49. Limitation & Future Work
Limited to symmetric 2D planner
→ Extension to 3d objects
Ignored air resistance and deflectance
→ Incorporating into perception model
Cannot represent multiple shapes
→Dynamic Shape Change
Weight movable device
[Zenner et al. 2017]
51. Computational Design of Hand-Held VR Controllers
Using Haptic Shape Illusion
‣a novel design concept using Haptic Shape Illusion
users perceive as if they handle the object though the actual
appearance of the controller differs from that of the object in VR
‣a data-driven representation of a perceived shape
based on the moments of a wielded object
we aggregate perceived shape data against hand-held VR
controllers with different mass properties through experiments
and derive the mapping using regression techniques.
‣an interactive optimization-guided hand-held VR
controller design tool
we implement a design system which enables automatic
design of hand-held VR controllers which are smaller than
target shapes while maintaining their haptic shape perception.