Periodic Reporting for period 2 - AIFORS (ERA Chair in Artificial Intelligence for Robotics)
Reporting period: 2022-07-01 to 2024-09-30
Despite serious efforts at national and European level, there are still significant differences in research and innovation performance in Europe. With adequate institutional support, outstanding researchers can have a decisive and positive impact on the culture and performance of research institutions, especially promising institutions in countries with low R&I performance. In this context, the six-year AIFORS project will establish an ERA Chair in Artificial Intelligence for Robotics at the University of Zagreb Faculty of Electrical Engineering and Computing (UNIZG-FER), which is the leading research and educational institution for robotics and artificial intelligence in Croatia. The AIFORS project and the opening of the ERA Chair in Artificial Intelligence for Robotics will significantly increase the attractiveness of UNIZG-FER on an international level.
The overall objective of the AIFORS project is to make structural changes at UNIZG-FER in order to attract excellent researchers and research managers who can have a significant impact on the Faculty’s research and innovation performance. The following structural changes are planned:
- The remuneration of researchers must be increased in order to offer positions to international applicants;
- International and intersectoral mobility of researchers needs to be systematized to increase exposure to different working environments, improve career prospects and attract international talent;
- A plan to upgrade research equipment or access to it in other institutions must be implemented to provide researchers with the resources to conduct world-class research;
- Human resources policies, particularly with regard to gender balance, must be in line with the principles of the European Research Area to ensure good employment conditions;
- The UNIZG-FER Research Support Center needs to be upgraded to strengthen research management and administration skills, as the Center currently does not provide support in proposal preparation and project management. A Horizon 2020/Horizon Europe specialist needs to be recruited.
- Dissemination, communication and exploitation capacities need to be strengthened within the research teams to increase the impact of their research.
The overall objective of the AIFORS project is to make structural changes at UNIZG-FER in order to attract excellent researchers and research managers who can have a significant impact on the Faculty’s research and innovation performance. The following structural changes are planned:
- The remuneration of researchers must be increased in order to offer positions to international applicants;
- International and intersectoral mobility of researchers needs to be systematized to increase exposure to different working environments, improve career prospects and attract international talent;
- A plan to upgrade research equipment or access to it in other institutions must be implemented to provide researchers with the resources to conduct world-class research;
- Human resources policies, particularly with regard to gender balance, must be in line with the principles of the European Research Area to ensure good employment conditions;
- The UNIZG-FER Research Support Center needs to be upgraded to strengthen research management and administration skills, as the Center currently does not provide support in proposal preparation and project management. A Horizon 2020/Horizon Europe specialist needs to be recruited.
- Dissemination, communication and exploitation capacities need to be strengthened within the research teams to increase the impact of their research.
We hired the ERA Chair holder in M16, while the first three PhD researchers and the grant proposal writing and project management specialist were hired shortly after the start of the 2nd reporting period. At the same time, we hired a postdoctoral researcher and a PhD student with the support of the Ministry of Science and Education. Later, we recruited a few more researchers so that the team became fully operational and could carry out the research work envisaged in the ERA Chair holder's strategy. We have already published a significant number of papers (11 published and 9 submitted papers), and with the 5 new researchers we have recently recruited, we expect a substantial increase in high-impact research outputs and corresponding publications by the end of the project.
To ensure the sustainability of the research activities beyond the project duration, we have organized a training course on research grant application and project management for the members of AIFORS, LAMOR and the Research Support Center of UNIZG-FER. We also submitted 13 new project proposals.
We sent several of our members abroad for short-term study visits (3-5 days), one PhD student for a 6-month research visit and one for a 3-month internship. The ERA Chair also spent a month as a visiting professor at the University of Basilicata in Potenza, Italy. We also organized five incoming mobilities, two of which resulted in joint conference publications.
We purchased a new server with graphics processors to be used exclusively by the members of the ERA Chair for training deep neural networks and secured free access to the necessary equipment and infrastructure as well as to UNIZG-FER and some partner institutions.
We also made an important contribution to the optimization of HR and equality policies at UNIZG-FER by organizing the creation of documents that will form the basis for UNIZG-FER's application for the "HR Excellence in Research Award". We organized a workshop to share our findings with UNIZG-FER employees.
An important contribution is also the creation of a new unit for grant proposal preparation and project management at UNIZG-FER, led by the specialist recruited by AIFROS, which has already supported the preparation of 17 major project proposals.
We have organized/attended many national and international dissemination events, but here we mention only the two most important events that we have established as annual events. The first is the AIFORS Colloquia on Natural and Artificial Intelligence, a series of small, invitation-only, face-to-face group meetings that focus on a small number of important research questions in the field of AI and robotics. The second event is the AIFORS Workshops on Industry–Academia Collaboration, a half-day event with panel discussions on innovation topics in the field of intelligent robotics. We also organized two training courses to strengthen the ability of our researchers to disseminate, communicate and exploit research results.
To ensure the sustainability of the research activities beyond the project duration, we have organized a training course on research grant application and project management for the members of AIFORS, LAMOR and the Research Support Center of UNIZG-FER. We also submitted 13 new project proposals.
We sent several of our members abroad for short-term study visits (3-5 days), one PhD student for a 6-month research visit and one for a 3-month internship. The ERA Chair also spent a month as a visiting professor at the University of Basilicata in Potenza, Italy. We also organized five incoming mobilities, two of which resulted in joint conference publications.
We purchased a new server with graphics processors to be used exclusively by the members of the ERA Chair for training deep neural networks and secured free access to the necessary equipment and infrastructure as well as to UNIZG-FER and some partner institutions.
We also made an important contribution to the optimization of HR and equality policies at UNIZG-FER by organizing the creation of documents that will form the basis for UNIZG-FER's application for the "HR Excellence in Research Award". We organized a workshop to share our findings with UNIZG-FER employees.
An important contribution is also the creation of a new unit for grant proposal preparation and project management at UNIZG-FER, led by the specialist recruited by AIFROS, which has already supported the preparation of 17 major project proposals.
We have organized/attended many national and international dissemination events, but here we mention only the two most important events that we have established as annual events. The first is the AIFORS Colloquia on Natural and Artificial Intelligence, a series of small, invitation-only, face-to-face group meetings that focus on a small number of important research questions in the field of AI and robotics. The second event is the AIFORS Workshops on Industry–Academia Collaboration, a half-day event with panel discussions on innovation topics in the field of intelligent robotics. We also organized two training courses to strengthen the ability of our researchers to disseminate, communicate and exploit research results.
Even the most advanced robots are still a long way from the ones we see in science fiction movies, or from their natural counterparts, animals and plants.
The biggest challenge is to embed AI and ML in physical systems with their mass and mechanical properties, with sensors and actuators that constantly interact with the environment in real time and require specific and fundamental adaptations. The learning methods have already proven to be extremely useful.
In the short term, we are working in three areas:
- Soft bimanual mobile manipulation.
- Network robots for multisensory perception.
- Fundamentals of new robotics.
Methods of probabilistic robotics, machine learning and deep learning have been used extensively in many areas of AI and robotics, including SLAM and navigation, the "bread and butter" of LAMOR, but not so much in these areas. On the other hand, very powerful and promising deep learning technologies such as "Transformers" have not yet been widely used in soft robotics.
Soft BiManual Mobile Manipulation is one of our main research areas. We deal with the fundamentals of new robotics to solve the technical challenges, especially with a focus on Physically Informed Neural Networks (PINNS). To this end, we are exploring new methods that exploit the most advanced results of contemporary theoretical mathematics (especially in complex analysis and differential geometry) to eliminate existing bottlenecks.
In the field of network robots for multisensory perception, we have promising collaborations in seismology and even gravitational wave detection.
In addition, we are working on deep learning for vision and mobile navigation of robots or for the control of collaborative robots, which promotes synergies between AIFORS and LAMOR research. We also want to jointly tackle "deep" research problems that have a large potential impact on application (in terms of technology transfer and potential to create spin-offs).
The biggest challenge is to embed AI and ML in physical systems with their mass and mechanical properties, with sensors and actuators that constantly interact with the environment in real time and require specific and fundamental adaptations. The learning methods have already proven to be extremely useful.
In the short term, we are working in three areas:
- Soft bimanual mobile manipulation.
- Network robots for multisensory perception.
- Fundamentals of new robotics.
Methods of probabilistic robotics, machine learning and deep learning have been used extensively in many areas of AI and robotics, including SLAM and navigation, the "bread and butter" of LAMOR, but not so much in these areas. On the other hand, very powerful and promising deep learning technologies such as "Transformers" have not yet been widely used in soft robotics.
Soft BiManual Mobile Manipulation is one of our main research areas. We deal with the fundamentals of new robotics to solve the technical challenges, especially with a focus on Physically Informed Neural Networks (PINNS). To this end, we are exploring new methods that exploit the most advanced results of contemporary theoretical mathematics (especially in complex analysis and differential geometry) to eliminate existing bottlenecks.
In the field of network robots for multisensory perception, we have promising collaborations in seismology and even gravitational wave detection.
In addition, we are working on deep learning for vision and mobile navigation of robots or for the control of collaborative robots, which promotes synergies between AIFORS and LAMOR research. We also want to jointly tackle "deep" research problems that have a large potential impact on application (in terms of technology transfer and potential to create spin-offs).