It’s always exciting when a new robot arrives in your lab. Usually, the more expensive the robot is, the more exciting it is. With the possible exception of Boston Dynamics’ ATLAS, NASA’s Valkyrie has got to be one of the most expensive humanoid robots ever made, and last year, NASA promised to give away (or, at least, lend) three of them to universities in the hope that Valkyrie will learn some new skills.
Within the last few weeks, the University of Massachusetts Lowell, which teamed up with Northeastern University in Boston, Mass., took delivery of their fancy new robot, as did MIT and the University of Edinburgh in Scotland. We talked to Holly Yanco at UMass Lowell and Taskin Padir at Northeastern, along with Sethu Vijayakumar and Maurice Fallon at Edinburgh, and Russ Tedrake at MIT, about what it’s like to have a smokin’ hot space robot show up on your doorstep in a bunch of pieces. We also asked them what they’ve told NASA that they’re going to do with it, and what they actually plan to do with it. NASA, you will be happy to hear that these last two things are only slightly different.
Valkyrie arrives at UMass Lowell in Pelican cases. Photo: Eric Marcoux
When we first met Valkyrie at NASA’s Johnson Space Center in Houston back in 2013, we were told that it was designed to be easy to take apart and then reassemble. It’s pretty cool to see this modularity in action; here’s a video of how Valkyrie got put together in Massachusetts:
Loyal Valkyrie fans will have immediately noticed that this version of Valkyrie (a “Unit D”) has some upgrades. Most notably, the robot’s head has been redesigned in order to accommodate a Multisense SL camera and LIDAR array, the same kind of “head" that ATLAS has. Also, the cameras in Valkyrie’s legs have been removed, the range of motion of the pelvis has been increased, and the fabric leg covers have been replaced with a plastic shell that incorporates new fans to help keep the robot cool as it attempts more dynamic walking tasks. There are some other minor upgrades to improve Valkyrie’s modularity and make the batteries safer, but the big deal is the Multisense, since it will allow people with experience doing perception on ATLAS to translate much more easily to working with Valkyrie.
This particular Valkyrie will live at the NERVE (New England Robotics Validation and Experimentation) Center at UMass Lowell, which is basically a big playground for robots, by which I mean a test area for robots. This is ideal, since part of NASA’s grant involves providing access to teams participating in NASA’s Space Robotics Challenge.
Meanwhile, across the pond, another Valkyrie arrived in pieces at The Edinburgh Centre for Robotics at the University of Edinburgh, and was put back together again.
Sethu Vijayakumar (front right), Maurice Fallon (front left), and Zhibin Li (second from right) with the Edinburgh Valkyrie team Photo: Sethu Vijayakumar
At Edinburgh, this Valkyrie will be the centerpiece of the “Robotarium” at the Center for Robotics. Along with world-class robotics research and testing facilities, Edinburgh also has funding to train more than 80 robotics PhD students over the next decade. A Valkyrie ended up here after years of close collaboration with NASA, and it looks like it’s going to have a fabulous new home.
Last but certainly not least, a third Valkyrie was delivered to Russ Tedrake at MIT:
I love that “Critical Space Item” sticker. The other thing to notice in that video is, at about 0:45 seconds in, the inside of Valykrie’s chest. You can clearly see those big actuators that push the robot’s chest out; those boobs aren’t just for show.
For details about what Valkyrie will be up to, we spoke with Taskin Padir (Northeastern), Holly Yanco (UMass Lowell), Sethu Vijayakumar and Maurice Fallon (University of Edinburgh), and Russ Tedrake (MIT).
IEEE Spectrum: How did you get a Valkyrie from NASA, and what are you planning on doing with the robot?
Taskin Padir: The vision is that Valkyrie or a similar platform will play a role in future Mars and space exploration missions. The most common tasks that astronauts perform in space is repairs. So we will focus on validating relevant space tasks that will involve both manipulation and locomotion.
More specifically, we will work on tasks such as maintaining solar panels (wiping the sand off after a storm for example), repairing an electrical panel which could be the part of a habitat or spacecraft or rover, exploring an area and picking up samples. While working on these tasks, we will extend our research expertise from the DRC in constrained motion control and grasping of unknown objects. The overarching goal is to develop manipulation and control capabilities for reliable, rapid, and risk-averse robot autonomy.
Holly Yanco: At the DRC Finals Workshop, Taskin Padir and I first talked about submitting a proposal for Valkyrie. Taskin had led the WPI/CMU DRC team, and I had led a study of the human-robot interaction of the DRC teams. Rob Platt (from Northeastern) also joined our team, bringing his grasping expertise. In addition to our research capabilities, our team also has a 10,000 sq ft robot testing environment, the NERVE Center at UMass Lowell, in which Valkyrie could live.
One of our tasks is to help NASA design and validate the Space Robotics Challenge. We will be building test courses at NERVE, then trying Valkyrie on them. In the initial stages of this work, the robot will be much less autonomous completing the tasks than it would be in the competition.
Sethu Vijayakumar: My view is that our research on the Valkyrie need not be necessarily restricted to challenges in the space domain. While the SRC defines some interesting challenges, we have the freedom (compared to the other U.S. teams) to investigate several parallel strands of investigation. One thing I am keen on exploring is data driven dynamics learning to complement model based control schemes—this is particularly relevant if you think about robots operating in various gravitational fields.
It is early days and I would not want to speculate on exotic, fancy things we want to try—I am under no illusion about the challenge that a platform such as this poses and the speed of progress that can be achieved given the limitations of the university research team size and funding. However, I hope that some of this research translates to closer human robot interaction scenarios.
Maurice Fallon: I'm interested in improving the dexterity of Valkyrie - bringing advanced visual tracking and mapping to the robot. Within the DRC competition time-frame, we really had to focus on making traditional approaches reliable for outdoor environments. But there's been some really great progress in this space in the last few years. It's a huge advantage to be able to bootstrap what we are doing with all the software we developed for Atlas.
Russ Tedrake: We have big plans for the robot. In addition to resolving basic issue in control (we’d like to be able to walk off-belay without any fear of falling), I want to push the limits of the hardware in terms of multi-contact locomotion like climbing ladders, and dexterous manipulation. This is also a chance to team with Leslie Kaelbling and Tomas Lozano-Perez to put their higher-level autonomy research together with my group’s lower-level planning and control.
This robot has a lot more sensing available than we had on Atlas, including joint side position and torque sensing, and consequently considerably higher fidelity torque control. I believe that this will be very enabling for some of the control approaches we want to try. The downside is that it has less power available, and relatively more distal mass (e.g. heavy feet). That will likely limit its ability to do step recovery (you just can’t move your feet fast enough). It’s a trade-off that we’re excited to evaluate.
Valkyrie getting calibrated at UMass Lowell. Photo: Eric Marcoux
IEEE Spectrum: Do you plan to work closely with other labs that have Valkyrie robots?
Taskin Padir: For a change, this is not a competition for us, as we are focusing on research and hosting SRC teams. We are already in contact with MIT, IHMC, and University of Edinburgh for collaborating on this project. I believe each team brings a unique approach to the table.
Holly Yanco: We definitely plan to share software between the teams. NASA has set up the ability for the teams to communicate easily with each other and NASA, and has also set up code repositories to facilitate sharing between groups.
Russ Tedrake: Absolutely. We’ve opened up our DRC code in the open humanoids project on Github. Maurice Fallon was on the MIT DRC team, and is now working with Val at Edinburgh—his team has been driving open humanoids while we waited for our robot to arrive. We’ve been talking with Northeastern to figure out how we’ll collaborate with them, too, but definitely plan on it.
Sethu Vijayakumar: Since about a year ago, we have been building on a common software base: a user interface (MIT Drake based), lower body control software (IHMC SCS based), and planning software (Edinburgh Exotica Planner, Real Time execution modules) that are shared across all four partners, including NASA, MIT, and IHMC. All core development is pushed across these partners, so it is a joined up effort, since this is essential for a project of this magnitude. We will also collaborate with Northeastern as part of the SRC effort.
First steps at UMass Lowell! Photo: Eric Marcoux
IEEE Spectrum: Apart from what you’ve told NASA you’re going to do with Valkyrie, what do you think it might be fun to try to get the robot to do?
Sethu Vijayakumar: One area we want to focus on is dynamic, compute efficient re-planning to deal with dynamic objects. Perhaps manipulating and balancing while we throw some balls that the robot needs to avoid. I think we will stick to sponge balls for now till I find another couple of million dollars to spare.
Taskin Padir: The ideas are plenty. We are interested in practical and real scenarios where Valkyrie can be used. It will be fun to see robot to do some activities of daily living in the home, for example.
Holly Yanco: One of our first bits of code on the robot is to send Valkyrie’s stereo cameras to an Oculus Rift (we’re still waiting for the new one; if anyone from Oculus is reading, it would be great to get our new headset a bit sooner than you’re predicting).
Taskin Padir and I had the opportunity to visit the Savannah River Site in South Carolina last December for an NSF-funded workshop organized by Richard Voyles. We spent four days learning about the challenges that robots could help with during decontamination and decommissioning (D&D). One of the tasks was D&D of gloveboxes that were used to make Plutonium-238 for NASA missions. Both Taskin and I think this application would be a great place to deploy a robot like Valkyrie. Currently, people are doing this D&D work; we envision sending the robot in to do the work, controlled by humans outside of the radiation area.
Russ Tedrake: I really want to push the hardware to the limits and perform very dynamic tasks. You’ll have to stay tuned!
There’s still a fourth Valkyrie in residence at IHMC in Florida. We’ll have updates for you on all of these bots just as soon as they start doing cool stuff, so look forward to lots more awesome Valkyrie videos over the next few years. And of course, the Space Robotics Challenge itself will be starting up this year, and we’ll have more coverage on that, too.
[ MIT CSAIL ]