A multimodal tactile dataset for texture classification on uneven surfaces using a 6-DoF robotic arm
4 affiliations
- 1Department of Electrical and Computer Engineering, University of Waterloo, Waterloo, ON, Canada
- 2Department of Computer Science, Memorial University of Newfoundland, St. John’s, NL, Canada
- 3Department of Computer Science, Memorial University of Newfoundland, St. John’s, NL, Canada; Department of Computer Science and Media Technology, Linnaeus University, Växjö, G, 352 52, Sweden
- 4Department of Computer Science, Lakehead University, Thunder Bay, ON, Canada; Corresponding author.
Abstract
This article presents a comprehensive tactile dataset designed to facilitate the perception and classification of textures on uneven surfaces. The data was collected using a 6-DoF Lite6 robotic manipulator equipped with a BioIn-Tacto multimodal sensing module. The data acquisition process consists of two main phases. First, an initial probing procedure is conducted to acquire 12 reference points on the curved surface. During this phase, the sensor's Inertial Measurement Unit (IMU) orientation is recorded precisely when the embedded barometer registers contact pressure, accurately capturing the surface normal. Subsequently, the robotic arm performs dynamic exploration by sliding the sensor along the generated trajectory, maintaining consistent contact with the surface. The dataset comprises multi-modal measurements from 12 distinct materials (including various fabrics, wood, leather, and plastics) mounted on curved geometries, with 25 exploratory episodes recorded per texture. The hardware captures nine-degrees-of-freedom Magnetic, Angular Rate, and Gravity (MARG) data alongside deep-pressure readings from a barometer encased in polyurethane rubber. This dataset has substantial reuse potential for researchers developing machine learning models and AI-driven sensorimotor policies for tactile texture classification in realistic environments. Furthermore, as it complements a previous dataset collected with a 4-DoF robot on identical surfaces, it can be used for cross-embodiment research, enabling the community to benchmark sensor performance and evaluate the impact of kinematic constraints on robotic tactile perception.
Related
- CBAM-enhanced convolutional networks for diabetic retinopathy classification in the ODIR dataset
- Multimodal patch-based deep learning for early Alzheimer’s disease classification using MRI and PET
- Multimodal fusion of SpO₂ time series and clinical data for explainable classification of obstructive sleep apnea severity
- An uncertainty-aware multimodal deep learning framework for Alzheimer’s disease classification using MRI and electronic health records
- SiNFluD: A novel benchmark dataset for sindhi figurative language classification
BibTeX
@article{khatibi2026multimodal,
title = {{A multimodal tactile dataset for texture classification on uneven surfaces using a 6-DoF robotic arm}},
author = {Soheil Khatibi and Frederico Luiz Martins de Sousa and Mateus Nazario Coelho and Vinicius Prado da Fonseca and Thiago Eustaquio Alves de Oliveira},
journal = {Data in Brief},
year = {2026},
month = aug,
doi = {10.1016/j.dib.2026.113334},
url = {https://doi.org/10.1016/j.dib.2026.113334},
note = {CC BY},
}
RIS
TY - JOUR
TI - A multimodal tactile dataset for texture classification on uneven surfaces using a 6-DoF robotic arm
AU - Soheil Khatibi
AU - Frederico Luiz Martins de Sousa
AU - Mateus Nazario Coelho
AU - Vinicius Prado da Fonseca
AU - Thiago Eustaquio Alves de Oliveira
PY - 2026
DA - 2026/08/01
JO - Data in Brief
DO - 10.1016/j.dib.2026.113334
UR - https://doi.org/10.1016/j.dib.2026.113334
AB - This article presents a comprehensive tactile dataset designed to facilitate the perception and classification of textures on uneven surfaces. The data was collected using a 6-DoF Lite6 robotic manipulator equipped with a BioIn-Tacto multimodal sensing module. The data acquisition process consists of two main phases. First, an initial probing procedure is conducted to acquire 12 reference points on the curved surface. During this phase, the sensor's Inertial Measurement Unit (IMU) orientation is recorded precisely when the embedded barometer registers contact pressure, accurately capturing the surface normal. Subsequently, the robotic arm performs dynamic exploration by sliding the sensor along the generated trajectory, maintaining consistent contact with the surface. The dataset comprises multi-modal measurements from 12 distinct materials (including various fabrics, wood, leather, and plastics) mounted on curved geometries, with 25 exploratory episodes recorded per texture. The hardware captures nine-degrees-of-freedom Magnetic, Angular Rate, and Gravity (MARG) data alongside deep-pressure readings from a barometer encased in polyurethane rubber. This dataset has substantial reuse potential for researchers developing machine learning models and AI-driven sensorimotor policies for tactile texture classification in realistic environments. Furthermore, as it complements a previous dataset collected with a 4-DoF robot on identical surfaces, it can be used for cross-embodiment research, enabling the community to benchmark sensor performance and evaluate the impact of kinematic constraints on robotic tactile perception.
SN - 2352-3409
ER -