At ICRA 2022, Competitions are a core a part of the convention. We shine a highlight on influential competitions in Robotics. On this episode, Dr Liam Paull talks concerning the Duckietown Competitors, the place robots drive round Rubber Ducky passengers in an autonomous driving observe.
Dr. Liam Paull
Liam Paull is an assistant professor at l’Université de Montréal and the pinnacle of the Montreal Robotics and Embodied AI Lab (REAL). His lab focuses on robotics issues together with constructing representations of the world (reminiscent of for simultaneous localization and mapping), modeling of uncertainty, and constructing higher workflows to show robotic brokers new duties (reminiscent of by way of simulation or demonstration). Earlier to this, Liam was a analysis scientist at CSAIL MIT the place he led the TRI funded autonomous automobile mission. He was additionally a postdoc within the marine robotics lab at MIT the place he labored on SLAM for underwater robots. He obtained his PhD from the College of New Brunswick in 2013 the place he labored on strong and adaptive planning for underwater autos. He’s a co-founder and director of the Duckietown Basis, which is devoted to creating partaking robotics studying experiences accessible to everybody. The Duckietown class was initially taught at MIT however now the platform is used at quite a few establishments worldwide.
Hyperlinks
transcript
Abate: [00:00:00] Hiya everyone. That is Abate. Subsequent week is ICRA and a core a part of this 12 months’s convention goes to be robotics competitions. So we’re going to deep dive into among the influential robotics competitions on the market. with a few quick spotlights on a number of totally different ones this week, we’ll be speaking to Dr. Liam Paul, the co-founder of the Duckietown competitors.
Hey Liam, welcome to Robohub. Might you give us a bit of little bit of background about your self?
Dr. Liam Paull: Positive. My identify’s Liam Paul. I’m a professor on the college of Montreal. I’m additionally the president of the Duckietown basis and one of many co-founders of that mission.
I did my PhD in in new Brunswick. After which I did a postdoc in MIT, which is the place this Duckietown factor began. And now I’ve been a proffer about 5 years or so.
Abate: Yeah. So right now really we actually wish to dive into the Duckietown competitors. Um, so might you give us a bit of little bit of [00:01:00] details about the way you began it, what your motivations had been?
Dr. Liam Paull: Yeah. So, I imply, the Duckietown factor is one thing that’s type of taken on a lifetime of its personal, for certain. It began as a category before everything, it was used for academic functions, however then sooner or later alongside the best way we thought that it could have additionally worth as, as a scientific benchmark. And so we began to see if we might reformulate and repurpose the platform to host these these competitions.
And the primary one was it NeurIPS. And I wish to say 2018 after which we’d achieved at the very least one at ICRA and some at NeurIPS and it’s kind of one thing that’s actually actually gathered the motivation, I feel actually is it’s all about making an attempt to carefully benchmark robotic algorithms. And this can be a fairly, it’s really a reasonably [00:02:00] exhausting activity.
A variety of robotic analysis is completed in some particular lab with a really particular setup and is kind of exhausting to breed. And so we wished to construct a really standardized however very accessible platform that individuals might simply get their fingers on, simply, put their algorithms on, and that we might someway like examine all kinds of algorithms in some.
Standardized and like honest, honest approach.
Abate: Yeah. So what’s the precise problem that they’re competing for and the way does it, how does it look?
Dr. Liam Paull: Yeah, so that is developed through the years, however the primary premise is, is, is. Principally the identical. In order a part of the Duckietown platform, now we have the automobiles, that are these little, little automobiles you could construct, however then there’s additionally an atmosphere wherein they function.
And the atmosphere is [00:03:00] made up of like yoga mats and duct tape and indicators that we’ve like printed and stuff. Um, however the concept is that it’s very standardized and really reproducible. To you or me, like, it seems like a small metropolis. Prefer it’s a really simplified view of a metropolis, nevertheless it’s one thing that approximates someway a small metropolis and the challenges are very in complexity, however largely concerned the robots navigating on this metropolis.
And we will. we will range the complexity by having totally different typologies of town intersections. We will have totally different obstacles, we will produce other autos. And so the complexity can actually develop. Um, however probably the most type of like primary, basic, like a PR factor that an agent ought to be capable to do is like drive down the street within the metropolis, keep away from obstacles and keep of their lane type of factor.
Abate: Yeah. So what was the motivation behind the identify Duckietown?
Dr. Liam Paull: That’s an [00:04:00] attention-grabbing, that’s an attention-grabbing one as nicely, really. So just like the ducky not too many individuals know this, however the ducky branding, not solely does it, it predates the Duckietown mission, nevertheless it additionally has an ICRA connection. So the opposite co-founder of the mission his identify’s Andrea Censi and now he’s at ETH Zurich.
And I feel the 12 months earlier than Duckietown began, he was… I overlook precisely what the title was, nevertheless it at present this push for everyone to submit movies they usually had been going to try to sew all of those movies collectively to make like a promo video for the, for the convention. And Andrea got here up with the concept that each video ought to have a rubber ducky in it kind of for numerous causes.
However I feel that partially, it was like for scale and in addition for like some type of coherence between the totally different movies. So they may do like enjoyable cuts and stuff in between the movies, however someway the branding of it similar to completely exploded. After which after we began this mission, [00:05:00] like earlier than anything, the one constraint was that it needed to have like rubber duckies concerned.
I… I don’t know… Simply kind of occurred that approach.
Abate: Yeah,
no, it’s nice. As a result of whenever you like grounded in one thing, that’s like a enjoyable idea it makes it way more partaking for individuals to, to wish to do it.
Dr. Liam Paull: Yeah. And there’s additionally a side of I imply, my view is that some, some robotics specifically is type of portrayed in a sure approach.
And I feel that like Hollywood has one thing to do with this. Scary, not like both it’s like Terminator are going to come back and kill you, or it’s scary within the sense that it’s going to take your jobs or no matter. And I feel, yeah, ultimately a part of, a part of the motivation behind this like type of enjoyable, playful type of factor was that we’d break this mildew a bit of little bit of making an attempt to make one thing that’s tremendous quick and tremendous scary and tremendous massive or no matter that perhaps this might enchantment to.
Totally different people who find themselves perhaps not [00:06:00] drawn to the, like, let’s construct an enormous, quick, scary factor, however as a substitute, you recognize, additionally need to have the ability to like categorical themselves someway by way of like by way of their work. And I feel yeah, I feel that’s additionally been, been a part of it and has been type of, type of profitable.
Yeah.
Abate: And so the competitors now it’s been operating for, is it a decade or two?
Dr. Liam Paull: It’s not, no, it’s not that lengthy. I feel it’s, I feel the primary iteration was in 2018. So I feel we’re at like, across the five-year mark. Um, however the five-year time. Yeah. The primary iteration of the category at MIT would have been one thing round 2016.
I feel. So the mission itself has in all probability been round for six or seven years, however the, the, the competitors itself perhaps solely 4. Hm. Yeah.
Abate: So what have been among the, the real-world advantages that that you just’ve seen out of the competitors?
Dr. Liam Paull: Yeah, that’s an ideal query. I imply, I feel with Roberta [00:07:00] robotics, I imply, a part of our you recognize, philosophy is that robotics ought to contain a robotic.
And I feel particularly in newer previous, there’s been this big development in direction of like machine studying and deep studying. Sort of algorithms. And I feel these algorithms actually have big potential, however whenever you try to put a few of these algorithms on robots, you see among the, among the type of nitty-gritty particulars that you just perhaps didn’t take into consideration actually have a big effect, you recognize, like how the latency of your system you recognize, the way it’s coping with.
asynchronous singles versus synchronous alerts, like treating time, you recognize, non-model defects and issues like friction and slippage and issues like this. And so for lots of the oldsters, I feel like the true, like the true world profit has been that, wow, they actually have gotten an appreciation for simply how, how powerful it’s [00:08:00] to, to construct these programs.
After which whenever you have a look at like what, though we’re not all the best way to having, you recognize business, autonomous autos. I feel you could get some type of an appreciation for simply how exceptional, what has already been achieved. , it truly is when you think about all of the totally different items that should work collectively and the way strong all of them should be.
Yeah.
Abate: And I can think about through the years, you recognize, totally different applied sciences have taken extra curiosity within the eyes of roboticists and that the method that the totally different individuals competing has modified fairly a bit as nicely.
Dr. Liam Paull: Oh, for certain. Yeah. Firstly, I imply, we very a lot noticed fairly conventional what I’d name like classical.
Not as a result of they’re outdated, however simply because it’s like the best way that issues was once achieved, type of like stacked that had the very customary abstractions of like, you recognize, notion and state estimation and planning and management, and now way more we’re seeing opponents [00:09:00] try to remedy this. And to finish machine studying sort of strategies, whether or not they’re primarily based on extra like imitation studying paradigm leveraging knowledge that we make accessible, or whether or not they’re utilizing the simulator primarily.
And simply making an attempt to do like reinforcement studying stuff. Type method after which switch their brokers that the true, the true robotic, these, I, I nonetheless assume it’s like stays to be seen at this level at this juncture, like which one is definitely higher at fixing the duty. However one factor that’s undoubtedly true is that the scholars within the opponents appear to be way more they discover the, like, I feel the machine studying type of method is extra interesting at this level.
It’s type of like this sizzling, sizzling matter, I assume.
Abate: Oh, that’s attention-grabbing. So it’s perhaps it’s extra interesting, however perhaps it’s not essentially as of proper now leading to a extra success for the opponents.
Dr. Liam Paull: Yeah. I imply, the best way that I view it, particularly like from a say a scientific standpoint is that [00:10:00] particularly on this atmosphere, all the things’s very well specified a very well engineered resolution with little or no studying goes to be very exhausting to be.
you recognize, the potential advantages of extra studying primarily based programs or that they need to be capable to be extra strong to various situations, be capable to generalize in kind of a extra, a easy, extra S simpler option to totally different environments. And so, yeah, it’s, it’s not, it’s not at all times straightforward. It’s not at all times straightforward to love now we have now we have to think twice about even simply what the metrics we’re going to make use of.
to match, you recognize, these totally different algorithms, like, is that simply the one which, you recognize, drives the quickest? I’m undecided that’s the perfect, you recognize, that’s the perfect metric. Um, there’s all these different parts about like robustness and talent to generalize, to totally different like situations and issues like that.
And in these circumstances, the [00:11:00] machine studying options perhaps do a bit.
Abate: Yeah, no, it’s an attention-grabbing level about overfitting your resolution to particularly the competitors atmosphere, aside from like whether or not or not that’s one thing that you just actually wish to do as a choose to say whether or not or not this can be a higher resolution, it could be higher on this competitors as a result of it was sooner… however ought to the impediment course change a bit, the topology change, now, perhaps it’s not so strong.
Dr. Liam Paull: I feel that is really the central problem in constructing robotic competitions. It’s very tough to construct a robotic competitors. That’s like not hackable in some sense you could’t win by simply actually overfitting to the specifics of that exact of that exact setup.
And so, yeah, I imply, I feel. You hit the nail on the pinnacle there it’s that is the massive problem for certain. And [00:12:00] making an attempt to construct like actually good robotic benchmarks.
Abate: Yeah. In order you, as you consider subsequent 12 months’s competitions have you ever guys ever thought-about perhaps doing a not releasing the map and having it’s a bit extra of a shock and have a bit of extra randomness related?
Dr. Liam Paull: Yeah. So we, now we have, now we have sometimes achieved that. Like, now we have a kind of a, like a, a validation set that individuals get the outcomes they usually can see all the things. After which what they’re really evaluated on as like a held out take a look at set that they don’t see. However what we’re fascinated about doing this 12 months, So sometimes what we’ve achieved is we’ve had kind of like perhaps two or three fundamental challenges, just like the lane following problem, the lane following with obstacles, problem, and the lane following with intersections problem or no matter.
And every certainly one of these challenges is, has its personal outlined metrics. Like how lengthy you survive for, or how far you’re touring in a sure period of time, kind of like customary stuff. What we’re going to do that [00:13:00] 12 months is we’re going to. Have a sequence of ranges successfully which can be simply more and more advanced and more and more tough.
And every certainly one of them perhaps has like some, some stage by way of the metrics that you need to obtain to ensure that it to be handed. However what we’re making an attempt to do is definitely alleviate the overfitting to any particular type of like particular activity and stage. You’re going to have an excessive amount of extra. B constructing a common function agent that’s capable of do moderately nicely in a, like a extremely like various like environments of various complexity and rising complexity.
And so I, that is our, that is our subsequent try, really at type of making an attempt to alleviate this, like over-fitting to the specifics of the, of the the particular like problem or no matter.
Abate: Thanks.
transcript
tags: Algorithm AI-Cognition, Algorithm Controls, Competitors-Problem, podcast
Abate De Mey
Robotics and Go-To-Market Skilled
