My ICRA 2026 Experience - Robots, Posters, Humanoids, and Research Overload

I visited ICRA 2026, from 1 June to 5 June. ICRA stands for International Conference on Robotics and Automation. It is one of the biggest and most prestigious robotics conferences in the world. Every year, world-renowned researchers, scientists, students, and companies attend ICRA to share the latest advancements in different areas of robotics. This year, ICRA was held in Vienna, Austria. I was attending the conference because my paper was accepted there. I had to present a poster for this work at the conference.

The title of the paper was:

COMPAct: Computational Optimization and Automated Design of Modular Planetary Actuators. [Paper] [Video]

I was not alone at the conference. My three other lab mates were also there, Vamshi, Aditya Shirwatkar, and Deepak Kapa. Vamshi and Shirwatkar were presenting their own respective papers, and Deepak was my co-author. So it was not just a conference trip. It was also a small lab trip. We enjoyed both the conference and the city. The city part of the trip deserves a separate blog, but this one is about the conference.

The Research Side of ICRA

One thing became very obvious at ICRA: a lot of robotics research is moving towards end-to-end neural network-based control. There were many papers on VLAs and world models. After a point, many paper titles started sounding like slightly different versions of the same sentence. Since this is not exactly my primary area of interest, I often found myself a little lost in the sea of posters.

And it was literally a sea of posters. Each poster session had around 400 posters, which was honestly overwhelming. After every poster session, I felt mentally saturated. So instead of stopping at every poster and trying to discuss everything on the spot, I followed a simpler strategy: I took pictures of the posters that looked interesting and planned to go through them later after the conference.

The number of classical control papers felt smaller. I saw one poster where the presenter had written:

“This paper contains no learning, if you have difficulty in understanding this, please keep walking.”

I found that funny.

Mechanical design papers were also fewer in number. One area that had a lot of energy around it was multi-fingered hands. There was interest both in design and control. There was also a workshop on Dexterous Manipulation using Multi-Fingered Hands, which I attended. The number of papers on quadruped locomotion was lower than I expected, which was a little disheartening for me because I love quadrupeds. On the other hand, there were more papers on humanoid locomotion. There were also many papers on manipulator control. The general vibe I got was that locomotion is being treated as almost solved, while manipulation is still far from solved.

So basically: fewer quadrupeds than I wanted, more humanoids than expected, and manipulation everywhere.

Glimpses of Researchers discussing their research

The Industry Section

The industry section was probably the best part of the conference for me. There were many robotics companies from around the world showcasing their robots, products, and solutions. This section felt very different from the poster sessions. Posters are useful, but seeing actual robots moving around hits differently. Humanoids were everywhere here as well. Companies like Unitree, LimX Dynamics, and several others were showing humanoid and biped robots.

A lot of companies were showcasing upper-body humanoid robots performing tasks using VLA policies. Some robots were folding clothes. Some were putting test tubes precisely into small pockets. Many humanoid and non-humanoid robots were also being teleoperated by humans. So yes, there was a lot of “robot doing human-like task while everyone watches carefully” energy.

There were also many companies working on multi-fingered robotic hands, which matched the research trends at the conference. It felt like the whole conference was saying: hands are hard, but everyone wants to solve them. Several companies, including Unitree, Direct Drive Robotics, and Rainbow Robotics, were showing quadrupeds as well. That made me happy.

However, some big names were missing. There was no Boston Dynamics, RAI Lab, Agility Robotics, ANYbotics, or Ghost Robotics. Their absence was noticeable, and because of that, ICRA felt slightly lacking. Many companies were also there to recruit people. They were clearly looking for talented researchers, which was interesting to see. Since actuator design is one of my core areas of interest, I talked to several actuator companies about their products. For me, that was one of the most useful parts of the industry section.

Glimpses of the industry section

Presenting My Work

I had my poster presentation on the third day of the conference. Presenting a poster at ICRA is fun, slightly stressful, and useful at the same time. We had a few interesting conversations with fellow researchers, and even got a possible collaboration. One positive feedback we received was about the automated design framework in our paper. People appreciated that our framework could design the actuator in an automated way.

Overall, presenting your work is always interesting. You get feedback, you understand how people see your research, and that helps in future work.

Imperial Night

One of the main attractions of ICRA each year is the gala dinner. This year, it was called Imperial Night. Imperial Night was held at the Hofburg Palace and the Spanish Riding School. It was a historical building in the heart of Vienna. It started with a huge queue outside the palace, but the queue was moving swiftly. The venue was a huge complex with food, drinks, and several cultural activities like dance and music. It was a fun place to be.

Only Deepak and I went. We participated in a treasure hunt where we were given a card and had to explore the whole area to find six booths. At each booth, we had to get a stamp and then submit the completed card. Out of the completed entries, some people would get a monetary prize through a lucky draw. We both, like proper nerds, ate food and did the treasure hunt. We did not properly watch any of the performances and came back. Honestly, there was too much crowd to properly enjoy the dance performances or music, and we were too tired to dance ourselves. But we enjoyed the treasure hunt a lot. One of the stamps in the treasure hunt was done by a manipulator, which was cool.

There was also a Mark Rober-style movable dart target. Wherever you hit the dart, the board would move so that the dart landed in the bull’s eye. I thought of throwing the dart at the very corner of the wall, but I could not do it because of my lack of skill. It fell somewhere in the middle, and the dartboard caught it. It became a bull’s eye anyway.

A glimpse of Imperial Night can be found here: Video. This is a video uploaded by Prof. Michael Milford, QUT Professor. He also attended the conference.

People, Indian Researchers, and Possible Collaborations

A lot of labs from various institutes in India were present at ICRA. Labs from IIT Gandhinagar, IIIT Hyderabad, IIT Bombay, IIT Jodhpur, IIT Delhi, and many labs from our own institute, IISc, were there. This was encouraging to see. It shows that robotics research in India is growing at a steady pace.

I met several people from foreign institutes as well. I met Riccardo, a PhD student working in Italy on cycloidal actuators for exoskeletons. He was quite interested in our work. We will probably talk after the conference and try to collaborate, which I think is wonderful. I also met Dr. Wesley Roozing from the University of Twente. He also works on cycloidal actuators for legged robots. We discussed his work and possibly made a connection for future opportunities.

What Was Good and What Could Be Improved

There were a lot of good works at the conference that caught my eye, and I made a list of some interesting papers. You can find the list here. But one thing I want to complain about is the poster organization.

Back in ICRA 2023, the papers and posters were divided according to broad topics during the poster sessions. That was helpful because you could first go to the relevant section, find papers close to your area, and then explore other works. This time, all the posters were in one place with no distinct division. That made it much harder to find relevant papers. When there are around 400 posters in one session, some structure really helps.

The app was also not very useful this time. During ICRA 2023, the app was quite helpful. You could search the name of a paper or researcher, and it would tell you the time, date, and place where the work was being presented. This time, the search feature was not good. I searched for my friend Vamshi’s name, and it did not come up. It was already hard to find papers because of the lack of proper poster division, and the app not functioning properly made it even more difficult.

Key Takeaways

• VLAs and world models are dominating a lot of robotics research.
• The conference seems to be moving more towards humanoids and manipulation.
• Quadruped locomotion felt less dominant than I expected.
• Multi-fingered hands for dexterous manipulation stood out as an important topic.
• Reliable data collection pipelines, for humanoid manipulation, are still being figured out.
• Model-based control and mechanical design papers were fewer, but they were still present.

ICRA 2026 gave me a broad picture of where robotics research is currently heading. Presenting my own work at ICRA was also meaningful. It gave me useful feedback and helped me see my research from other people’s perspectives.