license: apache-2.0
task_categories:
- tabular-classification
tags:
- tactile
- anyskin
- robotics
annotations_creators:
- expert-generated
size_categories:
- n<1K
dataset_info:
features:
- name: log_time
dtype: string
- name: mag1_x
dtype: float32
- name: mag1_y
dtype: float32
- name: mag1_z
dtype: float32
- name: mag2_x
dtype: float32
- name: mag2_y
dtype: float32
- name: mag2_z
dtype: float32
- name: mag3_x
dtype: float32
- name: mag3_y
dtype: float32
- name: mag3_z
dtype: float32
- name: mag4_x
dtype: float32
- name: mag4_y
dtype: float32
- name: mag4_z
dtype: float32
- name: mag5_x
dtype: float32
- name: mag5_y
dtype: float32
- name: mag5_z
dtype: float32
- name: slip
dtype: bool
Slip Detection Dataset
A Reachy 2's gripper has been equipped with an Anyskin tactile sensor, composed of 5 magnetic sensors. This dataset aims to reproduce the slip detection task described in the paper.
Capture protocol
A list of 16 objects with various hardness levels has been selected (see pictures
folder).
The gripper closure was manually triggered. Then the operator pulled the object without removing it from the gripper. Four forces were applied in different directions: up, down, left, and right. The gripper was opened between each force application, and the sensor was reset.
The data was recorded using the rerun software and then exported to a CSV file. The operator manually annotated the slip events. An additional capture with no slip events on various objects was recorded to provide only static contact examples.
Data format
Each row of the CSV file contains the following columns:
log_time
: the time of the capturemag1_x
,mag1_y
,mag1_z
: the magnetic field values of the first sensormag2_x
,mag2_y
,mag2_z
: the magnetic field values of the second sensormag3_x
,mag3_y
,mag3_z
: the magnetic field values of the third sensormag4_x
,mag4_y
,mag4_z
: the magnetic field values of the fourth sensormag5_x
,mag5_y
,mag5_z
: the magnetic field values of the fifth sensorslip
: a boolean indicating if a slip event occurred