EventTouch Dataset

1. Overview

The EventTouch dataset belongs to Haptic Sensing Lab at the School of Advanced Manufacturing and Robotics, Peking University. The EventTouch dataset contains data from tactile-sensor contact experiments and finite element method (FEM) simulations. It includes sensor outputs, contact forces, nodal displacements, and experimental parameters collected for different objects, contact poses, and press depths.

The dataset can be used for:

• tactile perception research;
• contact-force prediction;
• FEM simulation validation;
• comparison of simulated and real-world data; and
• tactile-sensor calibration.

2. Dataset Structure

data/
└── sensor/
    └── object/
        └── pose/
            └── press_depth/
                ├── fem_sensor.csv
                ├── fem_node.csv
                ├── real_sensor.csv
                └── camera_frames/

Directory descriptions:

• sensor identifies different types of tactile sensor.
• object identifies the geometry, material, and radius of the contact object(mm).
• pose identifies the contact pose of the object.
• press_depth identifies the press depth.
• Each lowest-level directory contains the data from one complete experiment.

3. File Descriptions

fem_sensor.csv

Contains the simulated six-axis force/torque data. Each row corresponds to one simulation frame. The number of data rows, excluding the header, is the number of valid frames in the experiment.

fem_node.csv

Contains the displacements and contact forces of the tactile sensor's surface nodes.

real_sensor.csv

Contains the raw and processed measurements from the physical sensor experiment.

camera_frames

A folder contains camera frames of vision_based tactile sensor.

4. Field Descriptions

Sensor types

• digit: digit sensor;
• s1: sphere vision-based tactile sensor;
• s2: ellipsoid vision-based tactile sensor;
• s3: large area resister-based tactle sensor;

Contact poses

For spherical objects:

• pose1: centered pressing;
• pose2: gripper offset upward by 2 mm;
• pose3: gripper offset upward by 4 mm;
• pose4: gripper offset downward by 2 mm; and
• pose5: gripper offset downward by 4 mm.

For cylindrical objects:

• pose1: centered pressing;
• pose2: gripper offset forward by 2 mm;
• pose3: gripper offset forward by 4 mm;
• pose4: gripper offset backward by 2 mm; and
• pose5: gripper offset backward by 4 mm.

Material codes

Code	Material
0010	Ecoflex 00-10
0030	Ecoflex 00-30
0050	Ecoflex 00-50
15A	Dragon Skin 15
PLA	PLA

fem_sensor.csv

Field	Description	Unit
frame	Frame index, starting from 0	Dimensionless
Fx	Force along the x-axis	N
Fy	Force along the y-axis	N
Fz	Force along the z-axis	N
Mx	Moment about the x-axis	N·m
My	Moment about the y-axis	N·m
Mz	Moment about the z-axis	N·m

fem_node.csv

Field	Description	Unit
frame	Frame index, starting from 0	Dimensionless
time_s	Simulation time	s
node_id	Node identifier	Dimensionless
x_m	x-coordinate of the node	m
y_m	y-coordinate of the node	m
z_m	z-coordinate of the node	m
ux_m	Node displacement along the x-axis	m
uy_m	Node displacement along the y-axis	m
uz_m	Node displacement along the z-axis	m
fx_N	Contact force on the node along the x-axis	N
fy_N	Contact force on the node along the y-axis	N
fz_N	Contact force on the node along the z-axis	N

real_sensor.csv

Field	Description	Unit
frame	Frame index, starting from 0	Dimensionless
sensor_ros_time_ns	Sensor timestamp	ns
phase	Gripper operating phase	Dimensionless
device_raw_fx	Raw x-axis force measured by the six-axis force/torque sensor	N
device_raw_fy	Raw y-axis force measured by the six-axis force/torque sensor	N
device_raw_fz	Raw z-axis force measured by the six-axis force/torque sensor	N
device_raw_mx	Raw x-axis moment measured by the six-axis force/torque sensor	N·m
device_raw_my	Raw y-axis moment measured by the six-axis force/torque sensor	N·m
device_raw_mz	Raw z-axis moment measured by the six-axis force/torque sensor	N·m
corrected_fx_F_N	Zero-offset-corrected x-axis force	N
corrected_fy_F_N	Zero-offset-corrected y-axis force	N
corrected_fz_F_N	Zero-offset-corrected z-axis force	N
corrected_mx_F_Nm	Zero-offset-corrected x-axis moment	N·m
corrected_my_F_Nm	Zero-offset-corrected y-axis moment	N·m
corrected_mz_F_Nm	Zero-offset-corrected z-axis moment	N·m
camera_image_path	Path to the visual-tactile sensor image associated with the current frame	N/A

5. Notes

1. The raw 1,000 Hz signals from the physical six-axis force/torque sensor are downsampled to 60 Hz.
2. Images from the visual-tactile sensor are captured at 30 frames per second.