University of Pennsylvania

ScholarlyCommons

Department of Mechanical Engineering &

Departmental Papers (MEAM) Applied Mechanics

2-12-2014

The Penn Haptic Texture Toolkit for Modeling, Rendering, and

Evaluating Haptic Virtual Textures
Heather Culbertson
University of Pennsylvania, [email protected]

Juan Jose Lopez Delgado

University of Pennsylvania, [email protected]

Katherine J. Kuchenbecker
University of Pennsylvania, [email protected]

Follow this and additional works at: https://repository.upenn.edu/meam_papers

Part of the Applied Mechanics Commons, and the Electro-Mechanical Systems Commons

Recommended Citation
Culbertson, Heather; Lopez Delgado, Juan Jose; and Kuchenbecker, Katherine J., "The Penn Haptic
Texture Toolkit for Modeling, Rendering, and Evaluating Haptic Virtual Textures" (2014). Departmental
Papers (MEAM). 299.
https://repository.upenn.edu/meam_papers/299

Please fill in your information on this form to be contacted about updates or bugs.
Data files updated 12/17/2013.
Rendering files updated and support for Windows application added 2/12/2014.

This paper is posted at ScholarlyCommons. https://repository.upenn.edu/meam_papers/299

For more information, please contact [email protected].
The Penn Haptic Texture Toolkit for Modeling, Rendering, and Evaluating Haptic
Virtual Textures

Abstract
The Penn Haptic Texture Toolkit (HaTT) is a collection of 100 haptic texture and friction models, the
recorded data from which the models were made, images of the textures, and the code and methods
necessary to render these textures using an impedance-type haptic device such as a SensAble Phantom
Omni. This toolkit was developed to provide haptics researchers with a method by which to compare and
validate their texture modeling and rendering methods. The included rendering code has the additional
benefit of allowing others, both researchers and designers, to incorporate our textures into their virtual
environments, which will lead to a richer experience for the user.

Keywords
haptic texture rendering, haptics, virtual reality

Disciplines
Applied Mechanics | Electro-Mechanical Systems | Engineering | Mechanical Engineering

Comments
Please fill in your information on this form to be contacted about updates or bugs.

Data files updated 12/17/2013.

Rendering files updated and support for Windows application added 2/12/2014.

This other is available at ScholarlyCommons: https://repository.upenn.edu/meam_papers/299

 The Penn Haptic Texture Toolkit for Modeling,
Rendering, and Evaluating Haptic Virtual Textures
Heather Culbertson∗, Juan José López Delgado†,
and Katherine J. Kuchenbecker‡

Haptics Group, GRASP Laboratory

Mechanical Engineering and Applied Mechanics
University of Pennsylvania, USA
February 12, 2014

Abstract
The Penn Haptic Texture Toolkit (HaTT) is a collection of 100 haptic texture and friction models,
the recorded data from which the models were made, images of the textures, and the code and methods
necessary to render these textures using an impedance-type haptic device such as a SensAble Phantom
Omni. This toolkit was developed to provide haptics researchers with a method by which to compare and
validate their texture modeling and rendering methods. The included rendering code has the additional
benefit of allowing others, both researchers and designers, to incorporate our textures into their virtual
environments, which will lead to a richer experience for the user.

1 License
This tookit is made publicly available under copyright from the University of Pennsylvania. The toolkit may
be used freely for non-commercial purposes such as research. You are free to alter, transform, or build upon
the work included in this toolkit. However, if you use any of the included data, models, or rendering code,
you must attribute them to the Penn Haptic Texture Toolkit [1]. Please see the attached license document
for full copyright and permission information. Also see [1] for a full description of our recording, modeling,
and rendering methods. The rendering code is distributed with permission from Geomagic, the maker of the
Omni and OpenHaptics1 .

2 Texture Samples
This toolkit includes information for 100 isotropic and homogeneous textures. As shown in Table 1, these tex-
tures are divided across ten material categories (paper, plastic, fabric, tile, carpet, foam, metal, stone, carbon
fiber, and wood). With the exception of metal, all materials were mounted on acrylic using double-sided tape.
Images of all 100 textures are included in the toolkit. These images were taken with a Nikon D40 digital
camera and are stored as square bitmaps with 1024 pixels on each edge. The physical scale of the images is
15 pixels/mm. The images shown in Table 1 are zoomed in from the images included in the toolkit to show
detail.
∗ e-mail: [email protected]
† e-mail: [email protected]
‡ e-mail: [email protected]
1 http://www.geomagic.com/en/products-landing-pages/haptic

1
 Table 1: Texture Samples
.
Paper .
. . .
Book Bubble Envelope Cardboard Coffee Filter Dot Paper

.. .. .. .. .
Folder Gift Box Glitter Paper Greeting Card Masking Tape

.. .. . . .
Paper Bag Paper Plate 1 Paper Plate 2 Playing Card Resume Paper

. . . . .
. . . .
Sandpaper 100 Sandpaper 220 Sandpaper 320 Sandpaper Textured Paper
Aluminum Oxide

. . . . .
.
Tissue Paper Wax Paper

. .
Plastic .
. . . .
ABS Plastic Binder Candle File Portfolio Frosted Acrylic

. . . . .

2
 . . . . .
Nitrile Glove Plastic Mesh 1 Plastic Mesh 2 Tarp Wavy Acrylic

. . . . .
. .
Fabric
. . .
Athletic Shirt Blanket CD Sleeve Canvas 1 Canvas 2

. . . . .
Canvas 3 Cotton Denim Felt Flannel

. . . . .
Fleece Leather 1 Back Leather 1 Front Leather 2 Back Leather 2 Front

.. . . .. ..
Microfiber Cloth Nylon Bag Nylon Mesh Pleather Portfolio Cover

.. . . .. ..
Silk 1 Silk 2 Textured Cloth Towel Velcro Hooks

. . . . .

3
 . . . . .
Velcro Loops Velvet Vinyl 1 Vinyl 2 Whiteboard Eraser

. . . . .

. .
Tile
. . .
Floortile 1 Floortile 2 Floortile 3 Floortile 4 Floortile 5

. . . . .
Floortile 6 Floortile 7

. .
.
Carpet
. . .
.
Artificial Grass Carpet 1 Carpet 2 Carpet 3 Carpet 4

. . . . .
Foam
. . .
. .
EPDM Foam Pink Foam Polyethylene Foam Scouring Pad Styrofoam

. . . . .
Textured Rubber

4
 Metal .
. . . .
Aluminum Foil Aluminum Metal Mesh Metal Shelving Textured Metal

. . .. . .
Whiteboard

.
. .
Stone
. . .
Brick 1 Brick 2 Ceramic Painted Brick Stone Tile 1

. . . . .
Stone Tile 2 Terra Cotta

. .
.
Carbon Fiber .
Carbon Fiber Resin Carbon Fiber

. .
. .
Wood
. . .
Cork MDF Painted Wood Stained Wood Wood

. . . . .

5
3 Recorded Data
The toolkit includes two recorded data files for each texture. Three axes of acceleration, force, and position
data were recorded while the experimenter explored each texture with a custom recording device using
natural and unconstrained motions. The recording device was fit with a 3.175 mm diameter stainless steel
hemispherical tooltip. Each data file is 10 seconds long and is stored at a sample rate of 10 kHz. Table 2
shows the information included in the recorded data files, which are stored in XML format. These data files
were used to create the texture and friction models presented in this toolkit.

Table 2: Recorded Data Files

Field Description
Accel MATERIAL

<AccelUnits> Units of acceleration

<SampleRate> Sample rate in Hz
<Accel> Combined acceleration signal using DFT321
Files

. algorithm
<Accel x> Acceleration in x-direction
<Accel y> Acceleration in y-direction
<Accel z> Acceleration in z-direction

. .
Position MATERIAL

. .
<SpeedUnits> Units of speed
<PositionUnits> Units of position
<SampleRate> Sample rate in Hz
Files

<Speed> Absolute speed value

<Position x> Position in x-direction
<Position y> Position in y-direction
<Position z> Position in z-direction
Force MATERIAL

<ForceUnits> Units of force

<SampleRate> Sample rate in Hz
<ForceNormal> Force normal to surface
Files

<ForceTangential> Force tangential to direction of motion

<Force x> Force in x-direction
<Force y> Force in y-direction
<Force z> Force in z-direction

4 Texture Models
Each texture’s recorded acceleration signal is modeled as a piecewise autoregressive (AR) process. The
models are stored in a Delaunay triangulation and are labeled with the normal force and scanning speed
used when recording the data.
To increase the flexibility and utility of HaTT, the toolkit includes a method for resampling the texture
models so they can be used to render textures at a sampling rate lower than the 10 kHz used when recording
data. Table 3 provides a summary of the files included, which resample the models and write the XML files
necessary for the rendering code.

6
 Table 3: Model Resampling Files
File Description
. .

function
Main
CreateResampledModels.m Main resampling function. Takes in
. new sampling rate in Hz as argument.
. .
. .

readValueXML.m Reads in single value from XML file

Helper functions

readArrayXML.m Reads in array of values from XML

. file
resampleModels.m Takes in the parsed model infor-
. mation, the original sampling rate
. (10,000 Hz), and the desired sampling
. rate. Returns the model information
. of the downsampled models.
modelXSL createHTML.xsl XSL stylesheet to create an HTML file
. for each new model XML filep

The toolkit includes both the original models at 10 kHz and the resampled models at 1 kHz. Table 4
shows the information included in the model XML files. These model XML files are used in the rendering
code included in the toolkit. HTML files are also included for visualization of the model sets.

5 Rendering Code
The code presented in the toolkit is based on OpenHaptics 3.0 and is for the implementation of our texture
rendering methods with a SensAble Phantom Omni. The rendering code may be adapted to run on other
hardware, but this has not yet been tested. The rendering methods presented in the toolkit are available for
Linux and Windows computers. The Linux version of the rendering code was implemented on a computer
running Ubuntu version 12.04 LTS with a GeForce FTX 570/PCIe/SSE2 graphics card. The Windows version
of the rendering code was implemented in Visual Studio 2008 on a computer running 64-bit Windows 7 with
an Intel HD Graphics 2000 graphics card.
Table 5 provides a description of the files included in the toolkit to run the sample rendering code. Third
party code is explicitly labeled. In addition, the Boost Random Number Library2 is needed to compile and
run the code.

6 Acknowledgements
The code was developed from the original TexturePad haptic rendering system designed by Joseph M.
Romano. This material is based upon work supported by the National Science Foundation under Grant No.
0845670. The first author was supported by a research fellowship from the National Science Foundation
Graduate Research Fellowship Program under Grant No. DGE-0822.
2 http://www.boost.org/

7
 Table 4: Model Files
File Description

<Material> Texture name

General Information
<htmlPicture> File path to image of texture for HTML
<renderPicture> File path to image of texture for rendering
<mu> Kinetic friction coefficient
<numARCoeff> Number of AR coefficients
<numMACoeff> Number of MA coefficients
<numMod> Number of models
<maxSpeed> Maximum modeled speed
<maxForce> Maximum modeled force
<speedList> Array of all modeled speeds
<forceList>p Array of all modeled forces
<AccelUnits> Units of acceleration
Units

<SpeedUnits> Units of speed

<ForceUnits> Units of force
<SampleRate> Model sample rate in Hz
. .
Triangulation

. .
Delaunay

<numTri> Number of triangles in Delaunay triangulation

<DTpicture> File path to image of Delaunay triangulation
<tri>p Model vertices for one triangle
. .
. .
<model> Information for one ARMA model, including:
Model Information

.....<modNum> .....Model number

.....<ARcoeff> .....Array of AR coefficients
.....<ARlsf> .....Array of AR line spectral frequencies
.....<MAcoeff> .....Array of MA coefficients
.....<MAlsf> .....Array of MA line spectral frequencies
.....<var> .....Model variance
.....<gain> .....Gain of MA coefficients
.....<speedMod> .....Model speed
.....<forceMod> .....Model force

References
[1] H. Culbertson, J. J. López Delgado, and K. J. Kuchenbecker. One hundred data-driven haptic tex-
ture models and open-source methods for rendering on 3d objects. In Proc. IEEE Haptics Symposium,
February 2014.

8
 Table 5: Rendering Files
Field Description
main.cpp Runs graphics loop and haptics loop;
src folder

. calculates texture vibrations

AccSynthHashMatrix.cpp Constructs hash entries to store models;
. interpolates between models
ContactModel.cpp? Defines equation of the force fields
helper.cpp? Sets the graphics state
pugixml.cpp§ XML parser
AccSynthHashMatrix.h Hash table class for texture generation
autoGenHashMatrix.h Parses model parameters from xml file
. and stores models in memory
shared.h Shared memory file for texture genera-
include folder

. tion
sharedInit.h Initializes variables in shared memory
. for texture generation
ContactModel.h? Defines constants for contact in simula-
. tion
Constants.h? Defines physical constants of simulation
pugixml.hpp§ XML parser
pugiconfig.hpp§ Configuration file for XML parser
foreach.hpp§ Boost.Foreach support for pugixml
classesp
. .
build folder

. .
. .
Models MATERIAL.xml. Model XML files
. .
. .
.
. .
images folder

. .
. .
MATERIAL square.bmp Texture image for display in rendering

? Adapted from OpenHaptics sample code, § Copyright 2006-2012 by Arseny Kapoulkine