IHMC develops full-size humanoid robot Nadia

The Florida Institute for Human & Machine Cognition (IHMC) at IHMC is one of the world's leading research institutes for the control of bipedal humanoid walking motion. Since 2015, IHMC's research has included the Boston Dynamics Atlas (DRC version) and A Valkyrie - making significant progress in pushing these platforms toward reliable mobility and manipulation. But fundamentally, when it comes to humanoid robots with human-like strength, speed, and dexterity, there haven't been many good alternatives.

A few years ago, IHMC decided it was time to build its own robot from scratch, and in 2019, we saw some very cool concepts for Nadia - a humanoid robot that can perform useful tasks at human speeds in human environments. After 16 beta versions of the experiment, Nadia is now a real robot, and it already looks pretty impressive.

For a long TIme, we've been pushing up agnst the limits of the hardware with DRC Atlas and Valkyrie. We haven't been able to fully explore the limits of our software and our controls. With Nadia, it's the other way around, which is really exciting from a research perspective.
—Robert Griffin, IHMC

Nadia’s design is essentially the next generation of the DRC Atlas and Valkyrie, which is faster, more flexible, and more stable, enough to be an excellent research platform. It is a mix of electric and hydraulic actuators: 7 degrees of freedom (DoF) electric arms and 3 DOF electric pelvis, as well as 2 DOF hydraulic torso and 5 DOF hydraulic legs. The hydraulic system is an integrated smart actuator. Nadia’s joints are arranged to maximize the range of motion, which means it has a dense operating space in front (where it really matters) as well as highly flexible legs. And, the carbon fiber shell covering most of the robot can be safely in contact with the environment.

IHMC

This large manipulation workspace is particularly important because it places a heavy emphasis on immersive VR teleoperation with Nadia. Not to say autonomy isn’t important, but the concept behind Nadia was shaped (at least in part) by IHMC’s experience in the DARPA Robotics Challenge finals: ideally, the robot would be able to manage everything on its own while giving humans more direct control over complex tasks. While full robot autonomy is not a fantasy, the fact is that in the near term, this approach is an immediately viable way to get robots to reliably do useful things in unstructured environments. Overall, Nadia aims to perform urban exploration and manipulation tasks at as close to human speed as possible. If that involves a human directly teleoperating the robot, or calling it the best way to do things, Nadia is designed to handle it.

IHMC

For more details about Nadia, we spoke to Robert Griffin, a research scientist at IHMC.

Q

IEEE Spectrum: How is Nadia unique among humanoid robots?

Robert Griffin: I think one of the great things about Nadia is that many of the joints have a very large range of motion. What sets Nadia apart from many other humanoid robots is not just the range of motion, but also the speed and power. Nadia has a much higher thrust-to-weight ratio than DRC Atlas and is significantly faster, which improves its overall speed and stability.

Q

Can you expand your and Nadia's goals into the field of humanoid robots?

Robert Griffin: We are trying to make Nadia a robot that can work in an unstructured environment, where you don't know what it needs to do ahead of time. So we are trying to build a robot with hardware, software, and interfaces that enable a human-machine team to go into an unknown environment, explore it, map it, and then perform meaningful tasks, rather than having a fixed sequence or number of tasks.

IHMC

Q

IHMC is known for helping complex humanoid robots like DRC Atlas and Valkyrie move. How did that experience feed into the design of Nadia?

Robert Griffin: You're right, we've been working with NASA JSC since DRC was built, and when you see Valkyrie walk, it's using IHMC software and controllers. When it came to applying that knowledge to Nadia, we paid special attention to the range of motion of the legs. Unlike many humanoid platforms, Nadia has a good hip roll (or adduction) setup, which means it can do very wide gaits, or very narrow gaits where the feet overlap. We think this will help with flexibility by adjusting stride length and reaching challenging footholds.

In addition to the hardware, we were able to use our tools and control stack to really help us specify the hardware design needed to accomplish these tasks. Currently, we are unable to optimize all of the joints of the robot and we are pretty much stuck with a hydraulic piston in the leg, but we hope that future designs will be better integrated and optimized. The difference with many humanoid robots is that we have full ankle actuation. Our approach from the beginning was to complete the necessary control rights with Nadia's feet to be able to do precise foot placement and use it to achieve relatively fast reactive movement.

Q

What compromises did you have to make in transitioning Nadia from concept to hardware?

Robert Griffin: Even though the robot has made a lot of improvements, it's still not as flexible as a normal person. I was surprised at how much range of motion we lost in the mechanism, especially around the hips, just because of the hydraulic hose routing. We wanted to avoid having big loops of hose hanging from the robot. We want to eventually bring Nadia into the real world, so we didn't want to be encumbered by that.

Nadia is not as fast as a human in some joints, and faster in others. We hope to make it a little more compact in the future - we are still a little off, right now it's about 90kg, not including hydraulic pumps or batteries.

Q

Why did you decide to focus on how humans could directly control Nadia in VR, rather than placing more emphasis on autonomy?

Robert Griffin: Autonomy works very, very well when you know what the operating conditions are and what you want to achieve. It's not that automation is easy to achieve, but it is doable. We have tried some autonomous planning and manipulation and found that to be a little limited because it doesn't leverage the full capabilities of our platform. So we are working on automating key parts of the system, like footprint planning and grasp planning, which can be automated very well. But we also need the ability to interact with new environments, which is why we are working on developing ways to collaborate with humans through interfaces such as virtual reality. Combining the cognitive capabilities of the human with the manipulation capabilities of the robot, we believe you get a lot of versatility.

Q

How much of a difference is it to control Nadia in VR, with the human-level flexibility of the hardware and the human-level workspace?

Robert Griffin: For manipulation, it's huge. We don't want users to feel limited in their abilities because of the limitations of the robot. A lot of the manipulation we can do on Nadia is not possible on other platforms.

There are some tradeoffs, of course - the arm is designed more for speed than for power, which is a little limiting, but we hope to improve on that in the long term. I think that designing the robot, especially the control system and software architecture, with a human operating it in mind, has developed a lot of capabilities that we hope will allow us to iterate quickly on many different paths.

Q

What unique abilities will Nadia have?

Robert Griffin: We hope that the robot will be able to uniquely perform multi-touch movement, and that's the direction we really want to go in. But I don't want to say it's better than other robots because I don't know what other robots can do in the future.

Q

Are you hoping that Nadia will be able to do some actual tasks at some point in the future?

Robert Griffin: Tasks like disaster response, nuclear repair, bomb disposal are mostly done by humans today because we don't have robots that are really capable of doing these tasks, even though roboticists have spent a lot of time on them. I think through a combination of great robotics engineering, incredible operator interfaces, and human cognitive capabilities, these tasks will be done by robots in the future.

Although Nadia is a brand new robot, IHMC has planned a number of upgrades for it, most notably placing batteries and custom hydraulic pumps in Nadia's torso, as well as higher torque arms and lighter, more streamlined legs. At the same time, there is still a lot of work to be done on the software side in order for Nadia to reach its full potential.