Replication data for the "Visual Goal-Conditioned Reinforcement Learning for Multi-Target Industrial Operations" paper

GENERAL INFORMATION Title of Dataset: Generated dataset for the article "Visual Goal-Conditioned Reinforcement Learning for Multi-Target Industrial Operations" Abstract: This dataset contains data generated during the experiments and validation of the Goal-conditioned Reinforcment Learning Control architecture, gathered in the control loop of the RL Manager node. Author Name: Enaitz Berrocal Molano ORCID: 0009-0002-7904-0007 Institution: Tecnalia Address: Mikeletegi Pasealekua 7, Donosti-San Sebastian, 20009, Basque Country, Spain Email: [e-mail] Author Name: Aitor Ibarguren Soldevilla ORCID: 0000-0002-1601-3645 Institution: Tecnalia Address: Mikeletegi Pasealekua 7, Donosti-San Sebastian, 20009, Basque Country, Spain Email: [e-mail] Additional Information Date of data collection: 2026-09-15 Geographic location of data collection: Basque Country, Spain Information about funding sources that supported the collection of the data: Funder name: European Union Horizon Europe Project identifier: RENEE (Grant agreement 101138415) SHARING/ACCESS INFORMATION Licenses/restrictions placed on the data: CC BY 4.0 Dataset doi: 10.5281/zenodo.23014735 DATA & FILE OVERVIEW dataset/ ├── README.md │ ├── ordered_screw/ | ├── pos1/ | | ├── screw1.csv | | ├── screw2.csv | | ├── screw3.csv | | └── screw4.csv | ├── pos2/ | | ├── screw1.csv | | ├── screw2.csv | | ├── screw3.csv | | └── screw4.csv | └── ... │ └── episode_randomised_screw/ | ├── pos1/ | | ├── screw1.csv | | ├── screw2.csv | | ├── screw3.csv | | └── screw4.csv | ├── pos2/ | | ├── screw1.csv | | ├── screw2.csv | | ├── screw3.csv | | └── screw4.csv | └── ... | └── timestep_randomised_screw/ | ├── pos1/ | | ├── screw1.csv | | ├── screw2.csv | | ├── screw3.csv | | └── screw4.csv | ├── pos2/ | | ├── screw1.csv | | ├── screw2.csv | | ├── screw3.csv | | └── screw4.csv | └── ... └── METHODOLOGICAL INFORMATION Description of methods used for collection/generation of data: Generated on the control loop of the RL Manager node, developed on ROS2 as a Python node, with a frequency of 30Hz. In each ROS2 call, the node generates a CSV file containing the selected pixels, Z heigh, action and velocity of the robot. The pixel error is calculated using the distance between the selected pixel obtained from the CSV and the desired pixel (475, 245). Specific Information for tabular data DATA-SPECIFIC INFORMATION FOR: screw*.csv All \*.csv/ data files use comma(,) as the delimiter* Description: These files contains the trajectory executed by the robot using a RL model. Number of variables: 15 Number of cases/rows: 1000 Variable List: U: Selected Screw pixel U relative to the camera Units: pixel Range: [0-480] V: Selected Screw V relative to the camera Units: pixel Range: [0-360] Z: Robot height Units: m a₁–a₆: Joint actions (movement per joint) Units: rad v₁–v₆: Joint velocities Units: rad / s