I was fortunate to get to see Robonaut in its home lab on two separate occasions 2006 - 2008. It is/was mainly a platform for research in teleoperation and hand-gripping. I was confused when I first saw a news item, years ago, about it going up to the ISS, because while it was forward looking to this purpose, it didn't seem like the prototype itself was intended for this. I think they'd have done some things differently if they knew the prototype itself was going to go up - probably make it smaller and lighter for one. But do remember, when this was started was before things like the Raspberry Pi, so a lot of components (computers, cameras, actuators) were bigger then than they would be now.
It was interesting to see the design evolve. For instance the first-generation Robonaut hand was based on human anatomy. One of the researchers spoke to hand surgeons and even observed real surgeries to learn about this anatomy. So the fingers of the first-generation hand were endoskeleton-supported. But on my second visit I got to see parts from a version of the hand they had just (or were just about to) put in. These were more of an exoskeleton design, like a loop/outline of each segment of a finger. (Unlike an insect exoskeleton, these segments were not fully enclosed; think of like a 2D loop not 3D tube.) Apparently the second design gave more room to put sensors in the fingers (and, I imagine, protected the sensors better). Once the glove is on the hand, you can't tell the difference.
I don't know, seems a little goofy, to make a humanoid robot. I know, it has to operate the same equipment that humans operate etc. But humans suck at zero gravity - there has to be a better solution for the ISS. Space spider? Zero-g-octopod? Slithering station snake? Something.
I got to see him in 2010 before he went up. Actually got to see the prototype with the acrylic chest. It was kind of awkward because my friend who worked at NASA took us into the room where the engineers were all working on him and just stared at us gawking at the robot.
Not to discount the work of roboticists, but that seems like a huge waste of space and weight. Were there significant gains from putting a prototype into space instead of perfecting it on the ground?
If you want to learn about how to do robots in microgravity, you have to really do it. It’s not just controls but also how lubricants behave, how dust and crud behaves in motors, and all the little things.
I think the human form factor is not the most useful, but it’s going to be important eventually to be able to do manipulation in orbit.
The human(ish, see the legs) form factor is likely preferable because they had ground controllers operating it in a telepresence design. If a human is trying to do things a human would do if they were there, a human shape is a good one. And many things they will manipulate were designed to be manipulated by humans.
Since we can't beam a specialist into space when we break something, the next best thing would be if that specialist could take over control of a humanoid robot in space they can directly control there to intervene and help fix a problem.
Perhaps someday when they have a problem with their humanoid robot in space, a roboticist can take control of another humanoid robot up there to perform the repair work.
Yes, those are the arguments they used. In particular it had to be able to use astronaut hand tools, which are designed for pressure suit gloved hands. On the downside the form factor is rather complex.
I recall a presentation when one of the very talented engineers that built it said (roughly) we made the robot now all you have to do (you being the robot AI) people in the room) is make it autonomous. A 30 DOF semi-android in microgravity? Piece of cake.... :|
Most power systems in space are far from cutting edge technology. If you're going to spend many dollars to put hardware into space, your priority shifts away from peak performance and towards steady, proven reliability. Lots of old tech is robust and well-understood.
Also, work on robonaut was initiated 20 years ago.
You can get plenty of modern stuff to go into a MicroTCA chassis. For what is essentially an industrial computer hooked up to motors and sensors, that kind of base hardware is perfect for the role.
The modern approach to robotics uses a high (-ish) speed serial bus - CAN 20 years ago, EtherCAT today - and a distributed network. Centralized chassis based systems are fragile and difficult to upgrade!
It was interesting to see the design evolve. For instance the first-generation Robonaut hand was based on human anatomy. One of the researchers spoke to hand surgeons and even observed real surgeries to learn about this anatomy. So the fingers of the first-generation hand were endoskeleton-supported. But on my second visit I got to see parts from a version of the hand they had just (or were just about to) put in. These were more of an exoskeleton design, like a loop/outline of each segment of a finger. (Unlike an insect exoskeleton, these segments were not fully enclosed; think of like a 2D loop not 3D tube.) Apparently the second design gave more room to put sensors in the fingers (and, I imagine, protected the sensors better). Once the glove is on the hand, you can't tell the difference.