Robotics and Embodied AI Conferences 2027: IROS, ICRA and More

Robotics has a smaller conference calendar than the rest of AI, and a more physical one. People bring hardware, run live demos, and spend as much time in the exhibition hall as in the talks. This guide pulls dates from our AI events calendar, so the lists are current, and covers what the generic AI calendars skip: which event suits which team, and what it takes to make a robotics trip pay for itself.

Research conferences. The peer-reviewed venues, in date order. Each name opens a page with the venue and the organiser's site.

Industry shows with a large robotics track. Not robotics conferences in the academic sense, but where a lot of the commercial conversation about physical AI and automation now happens.

CoRL, RSS and Humanoids are not in the list because their next editions are not in our data yet. All three run every year, and will appear as soon as dates are confirmed.

ICRA is the flagship of the IEEE Robotics and Automation Society and usually meets in late spring. IROS is co-sponsored by the IEEE and the Robotics Society of Japan and usually meets in the autumn. Both are large, broad and multi-track, with an exhibition, competitions and a long list of workshops. Most of the field treats them as the two fixed points of the year.

CoRL, the Conference on Robot Learning, is where the machine learning approach to robotics concentrates: learned policies, manipulation from demonstration, foundation models for control. If your team works on embodied AI in the current sense of the term, it is often the most relevant single venue.

RSS, Robotics: Science and Systems, is a smaller, single-track conference. Everyone sees the same talks, which makes it unusually good for conversation and for getting to know the people whose work you follow.

Humanoids, the IEEE-RAS conference on humanoid robots, is narrower again. Worth knowing about if that is your area, and a good place to meet the small number of teams building legged and humanoid platforms.

Competitions and challenges deserve a mention of their own. The flagships host robot challenges alongside the main programme, and they are often the best thing to watch if you want to see what systems can do outside a lab video. They are also a good recruiting ground: the teams that compete are the people who have spent months making something work under pressure, which is the skill most robotics roles actually need.

Perception work also lands at the computer vision flagships, and learning work at NeurIPS, ICML and ICLR. The NLP and machine learning conference guide covers those.

You publish robotics research. ICRA and IROS for breadth, CoRL for learning, RSS for a smaller room. Choose by deadline and by fit, and take the workshops seriously: they are where the newer ideas tend to surface first.

You run a robotics startup. One research flagship a year for hiring, one industry show for customers and investors. Trying to do both jobs at the same event usually means doing neither well, because the people you need are in different halls.

You buy automation for a business. An industry show or a trade fair in your sector. A research conference will show you what might work in five years, and you need what works on your floor next quarter. Ask every vendor for a reference customer with your kind of environment.

You are hiring. ICRA or IROS, with a list. Read the accepted papers in your area, pick the authors, and message them before you go. The robotics job market is small enough that a handful of good conversations can fill a role.

You run a service business considering robots. Warehousing, cleaning, hospitality, food preparation. Go to the trade show for your own sector before any robotics event. The vendors who have real deployments in businesses like yours will be there, and the conversations will be about uptime, maintenance contracts and what happens when the robot gets stuck, which are the questions that decide whether it pays back.

You are curious about embodied AI but do not work in it. An industry keynote stream will tell you more per hour than a research week. The research venues assume you know the field.

Robotics conferences are the only AI events where a good share of the attendees are travelling with equipment, and that changes the planning.

None of this is glamorous, and all of it decides whether the demo you spent two months building is running when the person you most wanted to meet walks past.

Robotics trips are expensive, because of the equipment and because the flagships are long. The case has to be specific.

State the outcome. "Meet six candidates for the controls role", "run demos for three named prospects", "present the paper and recruit from the workshop". Each of those is a reason. "Visibility" on its own is not.

Separate the people from the hardware. Price the trip with and without a demo. Sometimes the right answer is to send two people with a video and booked meetings, and save the freight for the show where buyers are.

Promise the debrief. A short note within a week on who you met, what you saw, and what you recommend. It is the cheapest way to make the next trip easy to approve.

Before: book meetings two to three weeks out. Author lists, exhibitor lists and workshop organisers are all public, and people expect to be contacted.

During: write one line after every useful conversation: what they are working on, what they need, what you promised. At a show with live demos, conversations are short and interrupted, so the line matters even more.

After: follow up within three days. The guide to event lead follow-up explains why that window decides most of the outcome. If your team exhibits and comes back with more conversations than it can follow up by hand, Wysera is built for that: Wyse drafts each message from the note you captured, and nothing sends until you approve it.

For regional and business events, browse the full AI events calendar.