Taylor Frey-Baker, a mechanical engineer, answers questions about why the current iteration of Atlas has a head, how the brain of the robot works, and her path to becoming a roboticist. Read the interview below and watch a conversation with Taylor about how Atlas understands the world around it.

Working at Boston Dynamics

What is your role at Boston Dynamics?
I’m a mechanical engineer on the compute and sensing team for Atlas. I work on the inside of the head, what we call the compute core. So I basically work on the brain of the robot.

What does one version of a typical day of work look like for you?
I don’t really have a typical day, but if I had to generalize I’d come in and grab some breakfast, check my messages, and then go to a morning meeting with the wider head team. Everybody gets together to talk about any issues that have come up, and how the builds are going.

And then maybe I have to work on a new part. So I would design something in our 3D modeling program and then pop it onto the 3D printer. Maybe I’d take some meetings about how a part is going to get made or about future testing, and then I’d pop up to lunch with some of my colleagues if it’s really nice to eat outside or on the grass. And then I come back and grab my 3D print test, put it into a head to make sure that it’s easy to assemble and I can get my tools in there. If all that looks good, I’ll go ahead and get the documentation out to make the change official.

A head above the rest

Why does Atlas have a head?
Hydraulic Atlas didn’t have a head, mainly because it was a research platform doing a lot of athletic and acrobatic work. It really only needed to see directly in front of it so that it could tell what terrain it was on. You can imagine, if you were going to do a forward roll you’d really have to tuck your head in.

So electric Atlas has a head because it’s going to be working around other people, and so it’s important for electric Atlas to have a better sense of the environment around it, to have a wider field of view. So we moved those cameras up into the head, and then the head is also a convenient place for us to indicate what Atlas is doing — we use lights and sounds, so that Atlas can show what it’s doing for other people around it.

What informed the design of the head?
A few things informed the design of the head. For the mechanical design, there are a lot of parts that we need to fit inside of the head. So we literally have to Tetris that into the head, make it all fit and mount to one another.

There are four perception modules, and their main feature is that they have stereo vision. So there’s two camera sensors. Similar to how humans have two eyes, and our brain kind of interprets the signals from both to understand the world around us — the sensor modules and Atlas are doing the same. So the stereo vision allows us to understand how far away different objects are.

What sets Atlas’ head apart from other humanoids? 
One of the things that makes Atlas’ head unique among humanoids is that it’s very modular. The head itself can swap onto and off of robots very easily. And then the compute core inside of the head comes in and out of the head like a video game cartridge. So what that modularity does for us is it lets us keep up with the pace and advancement in compute technology, so if there’s some new technology we don’t have to redesign the entire robot from the ground up. We can just redesign the head, put it onto an existing robot, and essentially we’ve upgraded the brain. We have a new robot with upgraded capabilities.

Why does the head rotate all the way around?
Continuous rotation is helpful on the head in a couple of ways. Atlas reuses a lot of actuators throughout the entire body, and we reused one actuator instead of artificially limiting ourselves with a hard stop or with a new actuator. That continuous rotation also makes it easy to tell where Atlas is looking when the head is rotating.

Getting into robotics

How can aspiring students prepare for a career in robotics?
Get interested in how things work and take things apart as they break. There’s a lot to learn from how something is put together and what choices the designer made. Don’t be afraid to start small, with really simple robots or at-home kits. Don’t feel like you have to wait for the perfect idea for this really complex robot or a really advanced mechanism.

What first sparked your interest in robotics?
I first got interested in robotics through my first job in high school, so I was 17 or 18. I worked at a children’s science museum and makerspace, so I was building very simple robots. But it kind of piqued my interest as I was going into college and figuring out what I wanted to do. When I got to college, I took a couple of mechanical engineering classes, did some simple robots there, and used that to transition into my first internship, which was working on robotic toys. And that was really a moment where I said, “Oh, this is something that you could do for work.”

My career path has grown in how complex the robots are that I’ve worked on. I started on robotic toys, which were very fun, somewhat simple mechanisms. I went on and worked in consumer robotics, robotic vacuums, and now working on an industrial robot. It’s the most complex system I’ve ever worked on. I feel like there’s so much room to grow and to learn, which is exciting.