FAIR Data Fund use case: The impact-aware robotics database

We inter­view Maarten Jon­ge­neel (Eind­hoven Uni­ver­si­ty of Tech­nol­o­gy) – one of the win­ners of the FAIR DATA Fund 2023 edi­tion and win­ner of the Dutch Data Prize 2024. Full team: Maarten Jon­ge­neel, Sander Dinge­mans, Alessan­dro Sac­con.

What is your project about?

In this project we present a new data­base sup­port­ing the devel­op­ment of impact-aware robot­ics, an emerg­ing field of research focused on enabling robots to exploit phys­i­cal impacts with objects and envi­ron­ments to allow for dynam­ic manip­u­la­tion and loco­mo­tion. 

The data­base can store a wide vari­ety of datasets con­tain­ing record­ed impact exper­i­ments where robots, objects, and envi­ron­ments expe­ri­ence inten­tion­al col­li­sions while per­form­ing dynam­ic robot­ic tasks. Exam­ples are object toss­ing with robot­ic arms to speed up through­put or grab­bing swift­ly a heavy object. This open data­base based on FAIR prin­ci­ples pro­vides access to data that sup­ports research relat­ed to, e.g., mod­el­ing, con­trol, para­me­ter iden­ti­fi­ca­tion, and object track­ing. 

Instead of focus­ing on a sin­gle dataset, our project revolves around the cre­ation of a data­base to share exper­i­men­tal data relat­ed to the field of impact-aware robot­ics. We cre­at­ed a cus­tom web inter­face where peo­ple can search for datasets using var­i­ous key­words among which those relat­ed to the used object, robots, and envi­ron­ments. The datasets itself are sin­gle HDF5 files, con­tain­ing all data and meta­da­ta of the exper­i­ments. These files are stored on the 4TU.ResearchData servers. More specif­i­cal­ly, we use a col­lec­tion — where the datasets are put togeth­er. 

This way, any per­son can upload their datasets to the 4TU servers, and can be added to the data­base after eval­u­a­tion. The soft­ware tools we pro­vide help in cre­at­ing and stor­ing data in a FAIR way, fol­low­ing the struc­ture we use for the data­base. A preprint of the sci­en­tif­ic pub­li­ca­tion relat­ed to the data­base is also avail­able.

What are some key results that you can share?

Stor­ing data in a FAIR way allows for effi­cient reuse of the data. With­in our project, the EU-fund­ed project on Impact-Aware Manip­u­la­tion (I.AM.), we have col­lect­ed a vast amount of exper­i­men­tal data with the objec­tive to use this data for learn­ing and val­i­da­tion of mod­els. We observed that data is often stored with­out con­sid­er­ing its reuse. In prac­tice, this often means datasets are lack­ing a prop­er struc­ture and addi­tion of meta­da­ta for human and machine inter­pretabil­i­ty. By cre­at­ing a struc­ture and a tax­on­o­my that can clas­si­fy the exper­i­ments relat­ed to our research field, we make the datasets eas­i­er to under­stand, enhanc­ing the pos­si­bil­i­ties for reuse. 

This not only helps oth­er researchers in the field to test and bench­mark their meth­ods, but also enhances the qual­i­ty of any asso­ci­at­ed sci­en­tif­ic pub­li­ca­tion, as it allows for repro­ducibil­i­ty of the results. So far, we have stored over 25 datasets from var­i­ous insti­tu­tions includ­ing those out­side our own project. The inter­est in the data­base is grow­ing, and we are cur­rent­ly talk­ing with var­i­ous insti­tu­tions to sup­port them in upload­ing their datasets to the data­base. At the time of writ­ing, there are six sci­en­tif­ic pub­li­ca­tions using datasets that are stored in the data­base, show­ing the added val­ue of our work to the research com­mu­ni­ty. 

How has the FAIR DATA Fund helped your work?

Set­ting up and main­tain­ing a data­base that is grow­ing in size and inter­est by var­i­ous insti­tu­tions is a demand­ing job. The FAIR Data fund has helped to cre­ate the pos­si­bil­i­ty to main­tain and fur­ther devel­op the data­base. More specif­i­cal­ly, it has allowed us to focus on stor­ing addi­tion­al datasets, both from with­in the I.AM. project and out­side this project, updat­ing the web­site front-end, and the back­end inter­ac­tion with the 4TU.ResearchData servers. At the same time, the FAIR Data fund has pro­vid­ed us with pos­si­bil­i­ties to present our work to a wide audi­ence, fur­ther increas­ing the vis­i­bil­i­ty of our work and the ben­e­fits of FAIR and open data. Besides focus­ing on the data­base itself, we hope to inspire oth­er researchers and stim­u­late them to take a sim­i­lar approach for their research field.