Play Live Radio
Next Up:
0:00
0:00
0:00 0:00
Available On Air Stations

Walking on Mars: OSU researcher talks training robots for outer space

Training four-legged robots at White Sands, New Mexico
Justin Durner
Training four-legged robots at White Sands, New Mexico

The following transcript was generated using automated transcription software for the accessibility and convenience of our audience. While we strive for accuracy, the automated process may introduce errors, omissions, or misinterpretations. This transcript is intended as a helpful companion to the original audio and should not be considered a verbatim record. For the most accurate representation, please refer to the audio recording.

MICHAEL DUNNE: I'm Michael Dunne. If you ever make the trek up the slopes of Mount Hood, you'll definitely see a lot majestic views of the Willamette Valley, spectacular glaciers and brightly colored climbers making their push for the summit not that long ago, if you happen to be in the exact right place at the exact right time, you might have seen something that almost defies belief, four legged robots desperately trying to walk up the steep ice-covered slopes. Today, on the show, you'll hear from an OSU professor and researcher who was part of that testing operation to see if specialty robots could traverse such a challenging landscape as a proving ground for much more inhospitable terrain, the surface of Mars. It's the stuff of science fiction, but it's real, and you'll hear all about it next on Oregon On The Record.  Christina Wilson, an assistant professor of research in the robotics Institute as part of the College of Engineering at Oregon State University. Thanks so much for coming on and talking with us.
CHRISTINA WILSON: Thank you for having me.

MICHAEL DUNNE: Tell us about the Lassie project. What is it?

CHRISTINA WILSON: Yes. So, the Lassie Project is an acronym. It stands for legged autonomous surface science in analog environments. So, there's a few parts of that. One is the legged autonomy. So, we're dealing with quadrupeds or dog style robots, also in reference to the Lassie acronym, but we're doing surface science with those robots. So, what that means is we're using the legs themselves of the quadruped to test properties of the surface that it's walking over. And we're doing that in analog environments. So, we're doing that in preparation for future missions on the Moon and Mars, where we anticipate quadrupeds could be deployed to kind of address a gap in the current robot forms that you see on other planets. So, we have examples of rovers, we have examples of a drone with ingenuity, but we haven't yet to see a quadruped deployed through NASA, and so that's what our group is working towards.

MICHAEL DUNNE: And how do you test those quadrupeds here on Earth?

CHRISTINA WILSON: Yes, so we take them to environments that are analogs. Most recently, we visited White Sands National Park, which we treated as a Mars analog. In the past, we've also gone to Mount Hood here in Oregon, which we treated as a lunar analog because of the very steep slopes of volcanic material and the presence of ice at White Sands, you know, there's a lot of sand. So that's our Mars analog. And we're particularly kind of thinking about the mobility challenges that rovers have had traversing across the sandy environments, and how, how can the quadruped kind of learn about the surface of the sandy environment to potentially assist a rover and more effective traversal or also assist a human teammate. Because the idea of the NASA moon to Mars program is that in the future, there will be humans deployed on the surface, working beside robots. And so, our project is also considering, how can a quadruped in the way, you know, maybe that the Lassie is a great example of a teammate that assisted a human in, you know, safety and other tasks. And so, we are kind of thinking about Lassie as, how can a quadruped assist a future astronaut as they're traversing these different surfaces.

MICHAEL DUNNE: For a non-expert such as myself is it a matter of something like this - a rover might be on treads, and so it might run into an obstacle that it can't get over, whereas a four-legged quadruped could step over a rock or something like that, is it as simple as that?

CHRISTINA WILSON:  To some degree. That's one component of it. I would also say, like, if you've ever driven on a beach in Oregon, you'll know that sometimes wheels and sand don't mix. Sure. So, there's examples of rovers, you know, on previous Mars missions getting stuck in sand. And so part of it is the locomotion of being able to step over things, but another part of it is that the legs themselves can actually provide information about the stiffness of what they're walking over, and so a quadruped would be able to sense that it had entered a very soft environment where there was just loose sand, and that would be information that it could use to change how it's walking in the same way that a person who encounters, you know, loose sand modifies their gait, or a quadruped is capable of doing the same thing.

MICHAEL DUNNE: So, game this out for us. Once these things are proven, it sounds like they are both an assistant to a potential human on a surface, but also, they're a diagnostic tool of the surface. Is that correct? Am I hearing that right?

CHRISTINA WILSON:  Yes, yeah. I think that's a great way of phrasing it. They can essentially act as a scout, so they can traverse the surface first, and as you said, like diagnose or provide some kind of map of what the strength of the surface is, and then that information can be used by humans, and it can also be used by other large assets, like rovers, to traverse safely. The other thing, though, is that that's actually interesting information to mission scientists. So, it's not purely just safety, its actual science value. So the quadruped through this type of walking we like to use the phrase like sensing through locomotion, or, you know, an experiment with every step to just reinforce the idea that it's, it's the leg surface interaction itself. That is the data. This isn't like just something that the flutter pit is carrying, okay, it's, it's that. It's the legs themselves. And so that type of data can tell us about the properties of the surface and the immediate subsurface, and that turns out actually to be really interesting to earth and planetary scientists, because they can use this as a complementary source of information to more traditional sensors that they're using. So we've looked at things like, how does this data relate to, like, the thermal inertia of an environment, or kind of the multi spectral composition of an environment? So our collaborators, and earth and planetary scientists are in science, are actually using this as data for their discipline.

MICHAEL DUNNE: Is it fair to say that these quadrupeds, and specifically their legs, are learning the environment? Is that the right word to use?

CHRISTINA WILSON:   It's a great question. I would actually say no, just because in the field of robotics learning has a very specific connotation. And you know, I guess in a kind of human language as well. But in robotics, learning is taking that data and actually applying algorithms to then predict with some amount of certainty what will happen next. And so, we have yet to take the step of actually trying to learn from the data that the robot legs are collecting, but that's definitely something that we're interested in, like starting to pursue at the moment. This is just data that is recorded for use by human teammates or potentially other robot teammates, and its data that just has value to the earth science discipline.
 
MICHAEL DUNNE: How did the quadrupeds do in these environments on Mount Hood and White Sands? Did they experience challenges? I mean, was it a success? In your view? Was it a limited success? You're a professor, give them a grade, if you will.

CHRISTINA WILSON:   Anytime you take robots out of the lab. It's really challenging. A lot of the videos, you know, the number one reaction I get, what I tell people I work in robotics, is basically. Basically, you know something about the robots taking over, or, you know, ascribing a lot of potential to robotic capabilities. And I always joke to people that, like, if you want to avoid robots, just go up like a steeper than 30-degree slope with obstacles, and you'll be fine, like they will never reach you. And it's, you know, it's super challenging. We take for granted our ability to easily locomote over certain surfaces, but it's difficult for current technologies to do that reliably. And so, although legs have a huge advantage, like in soft sand. So, we see that the, you know, the quadrupeds actually can go up, you know, decently steep us, like slip faces of dunes where there's a lot of just loose sand compiled. So just think, like, if you've ever been to the Oregon coast, and you're walking up like, you know, that steep side of the sand dune, and you're just, every step you take, you're slipping back several inches. So that's what the quadruped does, too. But it can get it, you know, it can eventually get up. We actually had a lot of locomotion challenges at Mount Hood, just because of the differences in like cobble to Boulder sized rocks and those are those are challenging, we just essentially had to avoid them entirely, and then also ice. It turns out ice and robot legs are a challenging combination. There are some epic robot falls, especially at Mount Hood. We saw some epic falls slipping from the robot, but I think it performed locomotion wise, as expected. And our team, the collaborators at USC, who are really focused on the locomotion component, we're able to learn a lot about how to improve the gait. And then also we're able to learn at the start of the project, our robot, essentially, the quadruped, walked one way. Now that we're at the end of the project, the quadruped is able to walk many ways, but the robot is crawling very slowly. And so actually these like sensing locomotion tradeoffs are something that we are able to develop because of the environments that we were working in and the challenges to locomotion that were to count.

MICHAEL DUNNE: Mount Hood can be an extreme environment, extremely cold. White Sands can be an extreme environment, extremely hot. And of course, the Moon and Mars have their just astronomically more challenging environments. What were things like temperature to these quadrupeds?

CHRISTINA WILSON:   Yes. So, I think the thing that comes to mind is because we were just at White Sands. We were there in August, kind of scheduling conflicts and so, yeah, the heat was extreme, and we were monitoring conditions for the safety of our science team. But it turned out that there was a point at which we were monitoring the conditions for the safety of our human team. But it turned out that the point at which it was like decided we would need to begin packing up and exiting the field. We called it the withdrawal period. That was the exact same moment that the quadruped batteries also started to have major issues. So, it turned out to be a good alignment between human safety and robot safety, both needing to exit the field at the same time. That being said, it really gave our team an appreciation for the type of challenges that these robots are going to have and human astronauts are going to have to face. But the challenge it's like on a minuscule scale, compared to actually deploying these on Mars, and so we work with a team at JSC Johnson Space Center, and we're very committed to the idea of putting a quadruped on the moon and on Mars, and it's something that we're actively pursuing, but there's a lot of work that needs to be done to ensure that that deployment can happen safely and successfully.

MICHAEL DUNNE: Can you contextualize how much improvement robots have made? Maybe in just the last decade, because I would imagine listening to you, you couldn't have done this 10 or 20 years ago, just the technology didn't exist. Give us an idea of, again, how robots and specifically quadrupeds, have improved over the years.

CHRISTINA WILSON:   Yes, so the advancements in robot technology over the last 10 years are just incredibly impressive from where we were to where we are now, I'd say, particularly for legged robotics. And now you see that most recently in the kind of proliferation of different humanoid robots which all have legs, and so legs and robots, and the study of that with quadrupeds really served as a foundation for some of the work that's being done in humanoids. So, it's incredibly impressive. I actually come from a background in cognitive psychology, so I am somewhat of an outsider, or I have an outside perspective on the field of robotics, because it's not something that I received classic training in. It's just a field that I have found myself in because of the work that I do with human scientists, and so it's fascinating to see such a young discipline compared to the discipline that I was trained and received my doctorate in, and just the speed at which these advancements have occurred, it's not something that ever it's not something I ever had experience with coming from like a different discipline.

MICHAEL DUNNE: Maybe talk a little bit about AI is obviously on everybody's minds these days about where it is and where it's going. Can you talk a little bit, perhaps in your classic research, but also hear how artificial intelligence is part of this entire equation?

CHRISTINA WILSON:   Yes, yeah. So, there's the aspect of the project where the legged robot is kind of a data collection machine. So, it's just walking around collecting data through its locomotion. But the other side of that is, where should the robot go, and how should it walk? Because it has these different gaits that have different tradeoffs, classically, those decisions of where to go and how to collect data are made by human scientists, but with advancements in robotics and in artificial intelligence, there's this ability for the robot to be more than just a Data Collection machine, but actually to have assistive capabilities. And so, the other part of the project that OSU is particularly involved in, and is like leading the collaboration, is the human robot teaming, which is using algorithms to support human decision making. So, it's a type of artificial intelligence.

MICHAEL DUNNE: Okay, well, to that point, and I know this is more of a philosophical question, but the idea of a human and a quadruped working together, and this has probably been something theorized and talked about for decades, ever since man ventured to space, is, could we see a situation where we wouldn't have humans risk their lives doing something like this and that robots could do it all?

CHRISTINA WILSON:   Yeah, I mean, it's a classic view of robots. It was basically a replacement for humans. In dull there's a phrase in the literature that's dull, dirty and dangerous domains, okay, so in these domains, those are cases where it would be potentially good for a robot to act as a replacement, and so that that is one view, and if we think about even in the same space of like robots for science, the idea of underwater remotely operated vehicles that are actually or, you know, these gliders are that are used in oceanography to collect data in places that like, yeah, otherwise, you would have to dive to potentially get this data. So that's an example of where robots are used as a replacement. And so, you could think about the same thing in space. And really, that's what the Mars Rover. Drivers are at this point, you know, if you talk to mission scientists who are part of the Mars operation, so these are scientists who are back on Earth, who are making the decisions of where the rovers should go and how they should and what data should be collected. There's a major limitation to this, which is where you can go is first constrained by where the engineering team has deemed is safe for the rover to operate. So there, there are places that the scientists would love to have data but that are just not accessible by the rover, which is why it's, you know, there's this idea of basically a team of robots that all have different capabilities. So, a rover is great for carrying a lot of sensors. It has a great payload. It's a reliable vehicle, but it has limitations on the types of trains it can access. A quadruped. Can access more terrains, but it can carry less equipment. So, there's kind of a tradeoff there, and then you think about a drone, you're accessing a whole different spatial scale, like you're able to move in this vertical space in a way that a ground robot can't.

MICHAEL DUNNE: Well, this is fascinating stuff. Christina Wilson, Assistant Professor of research in the robotics Institute within the College of Engineering at Oregon State University. Really appreciate you coming on and talking about this.

CHRISTINA WILSON:   Yes, thank you for having me.

MICHAEL DUNNE: That's the show for today. All episodes of Oregon On The Record are available as a podcast at KLCC.org. Tomorrow on the show, you'll hear from a reporter at The Oregon Capitol Chronicle about the latest with the legislative special session and the massive transportation funding package. I'm Michael Dunne, and this has been Oregon On the Record from KLCC. Thanks for listening.

Michael Dunne is the host and producer for KLCC’s public affairs show, Oregon On The Record. In this role, Michael interviews experts from around Western and Central Oregon to dive deep into the issues that matter most to the station’s audience. Michael also hosts and produces KLCC’s leadership podcast – Oregon Rainmakers, and writes a business column for The Chronicle which serves Springfield and South Lane County.