A historical past of RoboCup with Manuela Veloso

RoboCup is a global competitors that promotes and advances robotics and AI via the challenges offered by its varied leagues. We acquired the prospect to sit down down with Professor Manuela Veloso, one among RoboCup’s founders, to search out out extra about how it began, how the neighborhood has grown over time, and the imaginative and prescient for the longer term.

I feel it will be very attention-grabbing to go proper again to the start and listen to how RoboCup acquired began. What was the preliminary concept, and the way did it get arrange?

So we’re speaking concerning the mid-90s. By way of the analysis in these days, it was the start of the web and lots of AI and pc science researchers have been targeted on the web, first on refined search algorithms, on pure language understanding, on info retrieval, after which on software program brokers and machine studying utilized to digital info. From what I recall, there was a smaller group of researchers who have been thinking about precise, bodily robots, and specifically in AI and robotics. I personally was particularly thinking about the issue of making autonomous robots with notion (get info from the world), cognition (choose motion to attain targets), after which act (execute the deliberate actions). This mixture of notion, cognition, and motion is an excellent framework for autonomous robots, as a result of they need to get their info from their sensors, they need to cause about actions to attain their targets, after which execute them. So, through the 90s, I used to be at Carnegie Mellon with this AI analysis objective of integrating notion, cognition, and motion in autonomous robots.

Over in Canada there was Alan Mackworth, who collectively along with his great scholar, Michael Sahota, constructed a one-on-one little autonomous robotic soccer set-up. Two robots ran on a small subject that had a digicam overhead, and geared toward scoring every in one of many two little targets. This work confirmed that this process of kicking a ball and defending and aiming at a objective may very well be accomplished autonomously. So it was an amazing demonstration {that a} robotic soccer world might exist. Principally on the similar time, in Japan, Minoru Asada was exhibiting {that a} massive robotic might be taught with reinforcement studying tips on how to push a ball right into a objective. So you’ve got these one-on-one, absolutely autonomous little robotic automobiles that have been pushing balls round in Canada, after which there was this effort of studying to attain with a bigger robotic in Japan. The training robotic didn’t have a workforce, it was not an actual recreation, however it was exhibiting that reinforcement studying might be taught the talent of aiming right into a objective. After which there was additionally Hiroaki Kitano at Sony who was very thinking about little humanoids.

So that is very lovely as a result of all these items got here into play – all of us had totally different pursuits.

Alan Mackworth didn’t become involved with RoboCup, however he gave an illustration of those one-on-one robots at AAAI in 1994. And in these days, I had a PhD scholar who had simply joined – Peter Stone. And Peter was a critical and passionate soccer participant. He noticed this little recreation and he got here to me and stated, “that is what I wish to do for my thesis analysis, robotic soccer!” And for me, I used to be looking for a analysis setting the place autonomy was wanted within the robotic world. I had already a scholar, Karen Haigh, who was working with autonomous workplace robots, and studying to plan and execute. However with these soccer robots and Peter Stone’s curiosity the whole lot got here to play, and we began robotic soccer analysis in my lab.

In 1996, there was additionally a robotic soccer effort in South Korea, referred to as MIROSOT, and that’s the primary competitors we participated in. So Peter Stone, myself, and the workforce we constructed at Carnegie Mellon – Sorin Achim and Kwun Han – went to South Korea to take part. From South Korea, we flew to a robotic soccer workshop in Japan organized by Minoru Asada and Hiroaki Kitano. Additionally in attendance have been Dominique Duhaut, Itsuki Noda, Silvia Coradeschi, and Enrico Pagello. And that’s the place RoboCup actually began – we determined to do a contest. And the great factor was that Kitano was the chair of IJCAI, which was going to occur in Osaka in the summertime of 1997. So we’re there in Osaka and actually we got here up with this concept of getting a robotic soccer competitors, RoboCup. It was a giant second for us as researchers. We needed to give you the principles of this competitors so that folks would have the ability to take part seven months later. We got here up with the three leagues that we have been thinking about and had experience in.

The small dimension league, constructing upon our Carnegie Mellon pursuits, would have a subject with a digicam overhead related to a pc after which the pc would remotely management the robots via radio.

Then Minoru Asada had these greater robots with wheels, and we created a league that we name the middle-size league to incorporate the robotics analysis of Minoru Asada and others. After which Itsuki Noda was thinking about making a simulation setting. We thought that this is able to assist get extra individuals collaborating on this process of robotic soccer.

In order that’s how the three leagues began: the small dimension, the center dimension, and the simulation. Hiroaki Kitano, Dominique Duhaut and I have been in control of the small-size league, Minoru Asada was in control of the middle-size league, and Itsuki Noda ran the simulation league.

One of many challenges was to give you the principles and outline the robots and enjoying fields. I keep in mind my very own pragmatism in suggesting that we play on a ping-pong desk for the small-size, as a ping-pong desk is one thing that exists in the entire world. That meant that we might have a enjoying floor, with outlined dimension and texture, wherever on the planet. We determined one ping-pong desk for the small-size league, and 9 ping-pong tables for the middle-size league.

In the summertime of 1997, at IJCAI, once we all went to the precise first RoboCup competitors, the area was beautiful. Hiroaki Kitano had made these lovely fields and white bleachers across the fields. It was a really lovely area with an space with computer systems for the simulation league. There have been 80 groups that had joined the simulation on-line. We have been, I imagine, 5 groups for the small dimension league and about eight to 10 for the center dimension league.

That’s how it began. And 1997 it was in Osaka, in 1998 Dominique Duhaut organized RoboCup in Paris, at a contemporary Science Middle, La Villette. After which in 1999, RoboCup was organized by Silvia Coradeschi, in Stockholm, once more co-located with the IJCAI convention. In these three years 1997, ’98, and ’99, there have been solely these three leagues, small dimension, center dimension, and simulation. It was the muse of the whole lot. RoboCup grew persistently yearly when it comes to the variety of groups, the variety of members, and the variety of collaborating international locations.

Among the pitches in the principle soccer area at RoboCup 2024, held in Eindhoven, The Netherlands.

So how did RoboCup increase and will you discuss concerning the choices so as to add additional leagues?

Nicely, nothing was essentially deliberate, it was extra associated to the analysis pursuits that folks had. So once we have been in Melbourne in 2000, there have been two issues that have been added. One was RoboCupJunior. There was a professor, Elizabeth Sklar, who had an amazing curiosity in robotics training for youngsters. She proposed the RoboCupJunior competitions for youngsters Ok-12. The objective of RoboCupJunior was to coach all these younger individuals to do analysis in robotics. It was and it nonetheless is extraordinarily profitable. In Melbourne there have been in all probability the identical variety of kids as there have been individuals competing in what we now name the key leagues. In addition to the soccer leagues, Elizabeth additionally ran a dance competitors. This was very spectacular – the kids would dance on stage with their designed, constructed, and programmed cellular robots.

I imagine that the rescue league was additionally launched in 2000. The explanation why we got here up with rescue was as a result of in these days, there was loads of analysis on robots in catastrophe environments. So many individuals in our teams had an curiosity in creating robots that have been in a position to deal with catastrophe environments. So, we included that curiosity. And in a while, @residence was additionally launched as a result of individuals had an curiosity in service robots.

The logistics league got here later to cowl robots inside factories. The explanation for introducing new leagues was at all times to incorporate the neighborhood. And we didn’t need anybody to really feel they may not come to RoboCup as a result of they did totally different analysis with their robots. And so it was a really intellectually inclusive setting for AI and robotics researchers that might carry out duties autonomously – with notion, cognition, and motion principally in groups for specific duties.

Rescue league area at RoboCup 2024.

Might you discuss concerning the adjustments that have been introduced final 12 months, and the choice behind these?

Again in 1997, Hiroaki Kitano got here up with this objective of a robotic soccer workforce having the ability to beat the human World Cup winners by 2025. (I later tried, within the early 2000s, to rephrase that objective in order that it was not concerning the robots beating the people, however as an alternative having robots enjoying alongside human gamers by 2050.) Nonetheless, for this to occur, we now have began the humanoid league in order that the robots would have legs, not wheels. Our roadmap therefore strikes in the direction of humanoid robots. It additionally occurs that at present there are a number of robotic firms who’re producing humanoid robots which can be straightforward to amass, so researchers do not need the design and construct their very own humanoid robots, which was a troublesome process that was pursued solely by a restricted group of researchers. In parallel, occasions began aiming at together with humanoid robotic soccer. We believed that RoboCup must be seen and a reference for humanoid robotic soccer. We then proposed and altered that, within the worldwide RoboCup competitions, robotic soccer will focus solely on humanoid robots. That doesn’t imply that the opposite RoboCup leagues aren’t useful. But when they proceed at a regional stage, there’ll nonetheless be venues to foster these different varieties of pursuits. The RoboCup worldwide competitions could be solely targeted on humanoid robotic soccer.

Motion from the humanoid soccer league at RoboCup 2025. Picture courtesy of Alessandra Rossi.

RoboCup was tremendously profitable when it comes to together with a wide range of totally different autonomous robotic analysis pursuits from the neighborhood. However I additionally imagine that now we must always transfer into simply humanoid soccer on the worldwide stage. This may assist with visibility of the competitors and likewise for consolidation. We are able to now purchase humanoid robots that we didn’t have earlier than. Beforehand, you needed to construct them in your universities, and that was not a analysis path for many individuals. However there’s not that excuse anymore. You should buy humanoid robots, and more and more, you’ll be able to discuss with these humanoid robots utilizing GenAI. So it turned a unique problem, a way more accessible problem than earlier than. The problem will likely be: can these platforms play soccer within the presence of one other workforce? I feel it’s an interesting new path to attempt to give attention to this problem of the sport of soccer – a number of gamers, two groups in a big area, finally coordinating, collaborating, or enjoying with and in opposition to people. In order that’s the rationale.

When individuals don’t know the story, they have a tendency to marvel why we now have all these different leagues, resembling rescue and logistics. It was at all times pushed by the pursuits of the neighborhood and a giant coronary heart for a analysis strategy. And we at all times thought that as a result of we have been organizing an occasion with a venue, that we might embody these those that have pursuits outdoors of soccer. That has at all times been our drive. Nonetheless, it’s true that it diluted just a little bit the objective of 2050 – the humanoid soccer robots. Now I feel it’s time to return to soccer as a important focus, and nonetheless preserve supporting the opposite pursuits that we now have fostered, extra at a regional stage.

There are nonetheless loads of particulars to be sorted out. And, in fact, individuals could also be upset by leagues terminating on the worldwide stage. I’ve at all times participated within the small-size league myself, 20 years of collaborating, and I’m very hooked up to it. I used to be at all times the trustee representing the small-size league, and I’m not certain I’m pleased with the small-size not being within the worldwide occasion. However I have to suppose past what my private pursuits are and attempt to perceive what would have a much bigger influence for AI, robotics, and RoboCup. I feel that that’s demonstrating a gaggle of humanoids enjoying soccer. And if we don’t do it at RoboCup, possibly another person will.

The way in which I give it some thought is that we now have made this choice and we will reevaluate it in 5 or ten years. However not making a change and by no means having the braveness to deviate from what we now have been pursuing is much less thrilling. Issues have stalled a bit, with the identical variety of groups, the identical individuals, the identical guidelines, the identical sort of mental and analysis accomplishments. We have to excite individuals about one thing main once more. So I feel that the neighborhood will drastically embrace this new RoboCup worldwide. And we’re so effectively organized domestically, that we will help different leagues on the native occasions.

I do suppose that, from a scientific and analysis viewpoint, it’s the fitting second to focus on the soccer humanoid robots, due to their availability and our RoboCup final objective. I feel it will be superb if individuals knew RoboCup worldwide because the humanoid robotic soccer competitors. Are you able to think about in 2027 having 100 groups all enjoying soccer with humanoids? We’ll see.

About Manuela Veloso

Herbert A. Simon College Professor Emerita, Faculty of Pc Science, Carnegie Mellon College 2018-2026 Founder and Head, JPMorganChase AI Analysis, Managing Director

For the final eight years, Manuela Veloso has been the founder and Head of JPMorganChase AI Analysis and Herbert A. Simon College Professor Emerita at Carnegie Mellon College, the place she was college within the Pc Science Division after which Head of the Machine Studying Division.

At JPMorganChase, she constructed a workforce of 100 high gifted members with graduate training (PhD and Masters) in AI and associated disciplines. The workforce targeted on pillar areas of AI in finance, together with data-driven optimization, planning and search, doc evaluation, reliable AI, AI and mathematical reasoning, continuous studying, and multiagent programs. The workforce printed their analysis in educational venues and addressed and contributed to enterprise wants and imaginative and prescient.

Veloso has a licenciatura diploma in Electrical Engineering and an M.Sc. in Electrical and Pc Engineering from Instituto Superior Técnico, Lisbon, an M.A. in Pc Science from Boston College, and a Ph.D. in Pc Science from Carnegie Mellon College. Veloso has Doctorate Honoris Causa levels from the Örebro College, Sweden, the Instituto Universitário de Lisboa (ISCTE), Portugal, the Université de Bordeaux, France, and the Universidade Católica of Portugal.

She served as president of the Affiliation for the Development of Synthetic Intelligence (AAAI), and he or she is co-founder and a Previous President of the RoboCup Federation. She is a fellow of important skilled organizations in her space, specifically AAAI, IEEE, AAAS, and ACM. She is the recipient of the ACM/SIGART Autonomous Brokers Analysis Award, the Einstein Chair of the Chinese language Academy of Sciences, an NSF Profession Award, and the Allen Newell Medal for Excellence in Analysis. Veloso is a member of the Nationwide Academy of Engineering with a quotation “for contributions to synthetic intelligence and its purposes in robotics and the monetary providers trade.” She can be a member of the Academy of Sciences of Portugal.

Her analysis pursuits are in AI, together with Multiagent Programs, Autonomous Robots, Continuous Studying Brokers, and AI in Finance. For additional particulars, see Manuela’s webpage.

tags: RoboCup