Last week, from 14-18 October 2024, the INVERSE project was represented at the prestigious IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2024) in Abu Dhabi, UAE. Our INVERSE coordinator, Matteo Saveriano from University of Trento (UNITN), Emre Ugur and his team from Boğaziçi University (BOUN) and Alberto Finzi, Giuseppe Rauso and Riccardo Caccavale from C.R.E.A.T.E. were active participants, connecting with experts in the field and discussing the future of robotics.
BOUN team at IROS
During the conference, the BOUN team (Alper Ahmetoglu, Emre Ugur, had a chance to meet with fellow INVERSE partners Riccardo Caccavale and Alberto Finzi from C.R.E.A.T.E., where they shared thoughts on promising directions for robotics research. They also reconnected with well-known scholars like Prof. Tadahiro Taniguchi from Kyoto University, Prof. Takamitsu Matsubara from NAIST, and had insightful discussions over dinner with Justus Piater from Innsbruck University, Tamim Asfour from Karlsruher Institute for Technology (KIT), and his lab members. Additionally, they exchanged ideas with Gregory Stein from George Mason University, who works on task and motion planning (TAMP), and networked with other researchers in the poster sessions.
Despite the growing buzz around large language models (LLMs) and visual language models (VLMs), a recurring theme in many discussions was skepticism. Emre Ugur noted that while these models have potential, they are often overhyped. He emphasized that the robotics community largely views LLMs as useful for linking agent knowledge, but not as standalone reasoning tools.
C.R.E.A.T.E and their work
In other highlights, Alberto Finzi, Giuseppe Rauso, and Riccardo Caccavale from C.R.E.A.T.E. presented their work titled “Incremental Learning of Robotic Manipulation Tasks through Virtual Reality Demonstrations,” an exciting exploration of using virtual reality to facilitate the learning of manipulation tasks.
The research, authored by Giuseppe Rauso, Riccardo Caccavale, and Alberto Finzi, was well-received and sparked insightful discussions about the future applications of VR in robotic learning.
UniTN involved in workshops and presentations
Matteo Saveriano had a particularly busy week, chairing two sessions on robot safety:
- Robot Safety I (Wednesday), where the team presented their IEEE RAL paper, “A Novel Safety-Aware Energy Tank Formulation Based on Control Barrier Functions,” featuring the INVERSE project logo.
- Robot Safety II (Friday), which continued to explore safety advancements in robotics.
Saveriano was also selected as one of 25 researchers for the NOKOV New Generation Star Project, an honor that was highlighted in a presentation video shown during the social dinner.
Additionally, on Tuesday, the INVERSE team held a highly attended workshop titled “Variable Impedance Learning and Control: Navigating Challenges, Exploring Opportunities, and Shaping the Future,” drawing over 50 participants. The workshop explored cutting-edge approaches in variable impedance control, an essential aspect of safe and adaptive robotics.
This year’s IROS saw record participation, with over 29,000 visitors, averaging around 6,000 attendees each day. The conference provided a valuable opportunity for the INVERSE team to exchange ideas, showcase their research, and contribute to shaping the future of robotics.
