UBR-1 during a coffee break.
We’ve been following Unbounded Robotics, the final spin-off from Willow Garage, since we first learned of the company’s existence back in April. Unbounded has been working in stealth mode for the past year, but our best guess was that they were developing a low-cost mobile manipulator for research and education: something like a PR2, except (we were hoping) significantly cheaper. Today, Unbounded is unveiling UBR-1, a shiny new human-scale one-armed robot designed to completely revolutionize the market for research and education robotics and beyond, for just a tiny fraction of the cost of similar platforms.
UBR-1 and the PR2
As Unbounded Robotics says in their blog post this morning, “as Willow Garage alumni, we realize that UBR-1 will undoubtedly be compared to the PR2 robot from Willow Garage.” Yep, no kidding. It’s easy to just compare specs: the PR2 has two arms and costs $400,000 while UBR-1 has one arm and costs $35,000 sums it up in a nutshell. But obviously, there’s a lot more to it than that. To get some perspective on how the research community is reacting to UBR-1, we spoke with Pieter Abbeel from UC Berkeley, who was one of the participants in the PR2 Beta Program and has been working with PR2s just about as long as anyone. He got an early look at UBR-1 a few weeks ago, and here’s what he told us:
IEEE Spectrum: What were your thoughts when you first saw UBR-1?
Pieter Abbeel: First thoughts: super-cute! I’m hopeful that it might be as good as (or even better than) the PR2 thanks to lessons learned over the past few years, yet at 1/10 of the price, far more accessible and I accordingly anticipate wide adoption. At this price level one could even envision, if managing to program UBR-1 to solve a real-world task, that it could actually make sense to have a UBR-1 performing some tasks in the real world!
You’ve spent a lot of time using the PR2. How do you think UBR-1 will compare?
Research-wise, PR2 has been fantastic and UBR-1 looks like it could be as good, or even better. Of course, for bi-manual manipulation we might need two of them, but that’s still cheaper than one PR2. Visiting Unbounded two weeks ago, I learned that UBR-1 was designed to be easier to maintain and repair, which is important. UBR-1 also has a smaller footprint, which will be great for it to operate in human environments, which can be pretty tight space-wise.
If you get an UBR-1, what kinds of things will it enable you to do that you haven’t been able to do before?
Prolonged out-of-the-lab tests weren’t practical with PR2 per the monetary (and repair time) risks associated with such tests. I’m really looking forward to embarking on some projects with an UBR-1 outside of the lab environment!
Similarly, giving access to the PR2 for course projects was tricky because of the risk of breaking it, but with UBR-1 this seems more realistic. I think such wider access to a state-of-the-art mobile manipulator robot will have significant impact.
Willow Garage fostered a fantastic community around the PR2, and code sharing was abundant and this was really great for the robotics community. With UBR-1’s price tag (yet fairly similar capabilities) it has the potential to provide a very substantial boost to collaborations and sharing of code within the robotics community.
Why Affordable Mobile Manipulation Matters
Why could UBR-1 be so revolutionary for robotics right now? Really, it’s the general area of mobile manipulation that has a huge amount of potential for turning research robots into robots that might actually be able to do stuff that people like you and I actually care about. Humans spend a ridiculous amount of time moving things from one place to another, whether it’s shopping or cleaning the house or doing laundry or any number of other dull tasks.
It’s no coincidence that much of the research done with the Willow Garage PR2 focused on getting robots to do common household chores: the PR2 was one of the first (if not the first) truly capable universal mobile manipulator, with the necessary combination of sensors, hardware, and software to perform human-scale tasks in human-scale environments.
That combination of capabilities is very valuable, but it’s also very expensive, because so many different systems (sensing, computation, manipulation, and movement) have to be integrated with each other. Obviously, it’s not like you or I are going to run out and pick up a PR2 for $400,000 to do our laundry, but even though mobile manipulators aren’t quite ready for consumers yet, high cost is still a major issue, even for researchers. Part of the reason that the PR2 has been so successful is that Willow Garage just went and gave away 11 of them for free, effectively removing the enormously expensive barrier to entry for a lucky few. But that’s not sustainable, and as much as we love the PR2, the simple fact is that most researchers will never have access to one.
With UBR-1, just about every robotics research lab will be able to afford their own robot. And $35,000 is even potentially achievable for high schools with a little bit of grant money or fundraising. Heck, instead of asking their parents for a new car when they go to college, budding roboticists could just ask for an UBR-1 instead, and it would probably be more beneficial for them in the long run. But anyway, the point that we’re trying to make is that when you vastly expand the number of researchers to have access to your robot, you also vastly expand the amount of progress than can be made towards robotics in general, especially since UBR-1 is built on ROS, which is open source.
The other nice thing about having a robot that costs just $35,000 is that it suddenly becomes much easier to justify using it in a commercial or industrial context where getting a tangible return on your investment is important. This is the logic behind Rethink Robotics’ Baxter: the robot is inexpensive enough that using it to replace humans in certain situations starts to make sense.
While Unbounded’s first customers may be in research and education, the company is also aggressively emphasizing the potential the robot has for real-world applications by building apps for businesses. Since the robot can move around autonomously for extended periods, recharge itself, and pick objects up off of the floor and put them on a table (a skill that most robots find exceptionally difficult because of the range of motion required), tasks like cleaning, cooking, and even fetching drinks from the fridge are starting to become feasible. The PR2 has already done all of these things, of course, but if UBR-1 can do them as well (or better) for a tenth of the cost, we can start having a serious conversation about when we’ll finally be getting those robotic maids and butlers that we’ve been promised since The Jetsons.