Datasets for classification and regression in human-robot interaction with parallel robots

This repository contains three datasets for the analysis of contact events in robots. The data was recorded using a parallel robot and is suitable for regression and classification tasks in the field of machine learning.

The datasets are divided into the following three tasks:

1. **Regression of Contact Location and Force**: The goal is to predict the location and force of a contact on the robot.
2. **Classification of the Collided Body**: This task aims to identify the robot's body that is in contact with the environment (one of the six links or the end-effector platform).
3. **Classification of the Contact Type**: This task deals with distinguishing between different types of contact, such as collisions and clamping.

More details on the first two datasets can be found in the publication:
[https://doi.org/10.1109/IROS55552.2023.10342345](https://doi.org/10.1109/IROS55552.2023.10342345)

Information on the third dataset can be found here:
[https://doi.org/10.1109/IROS55552.2023.10341581](https://doi.org/10.1109/IROS55552.2023.10341581)

## Data Description

The data is available in `.csv` format and contains time-series data from the robot's sensors and a force-torque sensor. More information on the datasets can be found in their readme-files. The following variables are included in the datasets:

- **`t_s`**: Time (s)
- **`q_des_deg_[1-9]`**: Target joint angle for joints 1-9 (deg)
- **`q_deg_[1-9]`**: Actual joint angle for joints 1-9 (deg)
- **`x_des_m_rad_[1-3]`**: Target end-effector pose (x, y, orientation) (m, deg)
- **`x_m_rad_[1-3]`**: Actual end-effector pose (x, y, orientation) (m, deg)
- **`xd_ms_rads_[1-3]`**: Actual end-effector velocity (x, y, orientation) (m/s, deg/s)
- **`tau_qa_Nm_[1-3]`**: Actual motor torque for motors 1-3 (Nm)
- **`tau_ext_fts_Nm_[1-3]`**: External torque projected from force-torque sensor (Nm)
- **`tau_ext_est_Nm_[1-3]`**: Estimated external torque (Nm)
- **`F_ext_fts_N_Nm_[1-6]`**: External forces (1-3) and moments (4-6) from force-torque sensor (N, Nm)
- **`F_ext_est_mobPlat_CS0_N_Nm_[1-3]`**: Estimated external force and moment on the mobile end-effector platform (N, Nm)
- **`F_ext_fts_proj_mobPlat_CS0_N_Nm_[1-3]`**: Measured and projected external force and moment on the mobile end-effector platform (N, Nm)
- **`distances_m_[1-3]`**: Distances for classification (m)
- **`angles_deg_[1-3]`**: Angles for classification (deg)
- **`collided_body`**: Identifier for the body in contact (1-6=links, 7=platform) (-)
- **`chain`**: Collided chain (-)
- **`link`**: Collided link of the chain (-)
- **`location`**: Normalized location of the contact point on the link (-)
- **`clamping_collision`**: Identifier for the type of contact (0=collision, 1=clamping) (-)

### Target Variables

- **Regression**: `location` and the third component of `F_ext_fts_N_Nm_[1-6]` (link-orthogonal force).
- **Classification of the Collided Body**: `collided_body`.
- **Classification of the Contact Type**: `clamping_collision`.

## Usage

The datasets are available as `train.csv` and `test.csv` (for the classification tasks) and `data.csv` (for the regression task). They can be loaded using common libraries like Pandas in Python to train and evaluate machine learning models.

## Citation

If you use these datasets in your research, please cite the corresponding publications.