AutoLab-Co-Manipulation-Dataset

A dataset of Robot recordings of 2D symbols.

Welcome to the AutoLab Co-Manipulation Dataset page!
This is a public available dataset and comprises of 21 different 2D symbols. Each symbol was printed in A4 format, then traced back with the help of a Franka Emika Panda robot, equipped with a Schunk FT-AXIA force/torque sensor.
Six recordings were made for each symbols, by fixing the start and end point to be always the same. In the provided files.pdf related to the symbols, start and end points are represented, respectively, by a circle and an asterisk.

A key feature of the dataset is that symbol's recordings of the same group do not differ in path, but instead vary significantly in time, including speed variations and pauses.
The recordings are provided both as MATLAB and Python structures in every symbol folder, named respectively as symbol_data.mat and symbol_data.npy . The structure is the following:

symbol_data: structure of 6 elements with fields
--- symbol_data[i].pos : end-effector position recordings [3xT]
--- symbol_data[i].vel : end-effector velocity recordings [3xT]
--- symbol_data[i].F : end-effector force recordings [3xT]

where T defines the number of samples for each recording.

Load MATLAB file example:

load('symbol_data.mat');
i = 2; % 2nd recording
pos = symbol_data(i).pos;
vel = symbol_data(i).vel;
F   = symbol_data(i).F;

Load Python file example:

import numpy as np
symbol_data = np.load('symbol_data.npy')
i = 2 # 2nd recording
pos = symbol_data[i]['pos']
vel = symbol_data[i]['vel']
F   = symbol_data[i]['F']

The AutoLab Co-Manipulation Dataset was introducted as part of the paper entitled Arc-Length-Based Warping for Robot Skill Synthesis from Multiple Demonstrations by G.Braglia, D.Tebaldi, A.E.Lazzaretti and L.Biagiotti, from University of Modena and Reggio Emilia and Federal Technological University of Paranà.
If you find this dataset useful for your work/research, please cite:

@INPROCEEDINGS{11245947,
  author={Braglia, Giovanni and Tebaldi, Davide and Lazzaretti, André E. and Biagiotti, Luigi},
  booktitle={2025 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)}, 
  title={Arc-Length-Based Warping for Robot Skill Synthesis from Multiple Demonstrations}, 
  year={2025},
  volume={},
  number={},
  pages={8111-8118},
  keywords={Motion planning;Accuracy;Tracking;Heuristic algorithms;Robustness;Trajectory;Recording;Synchronization;Robots;Intelligent robots;Learning from Demonstration;Motion and Path Planning;Datasets for Human Motion},
  doi={10.1109/IROS60139.2025.11245947}}

Folders

• AutoLab_CoManipulation_Data/ : dataset files in .mat and .npy format, for each symbol a .pdf file depicting the respective symbol is included;
• load_example.m : a MATLAB code to plot symbol recordings;
• load_example.py : a Python code to plot symbol recordings;

Questions & Suggestions

For any doubt, question or suggestion, please feel free to email at: [email protected]