Digit from Agility Robotics: why humanoid robots are more efficient for many tasks
During the coronavirus pandemic, the Digit robot from Agility Robotics took a short break from work. However, it must be admitted that most of what the company showed the world over the past year and a half was a research project. For example, Digit competed with a human in squats and learned not to fall.
All of this is very important, but Agility's goal is to turn Digit into a useful and practical robot. It is not a research platform. The company claims the robot is designed to "boost business productivity and help people lead fuller lives." Digit is indeed meant to spend its time performing boring, repetitive tasks so that people don't have to do them.
In a new video, Digit showed how it can help with the tedious shuffling of warehouse boxes. The key is the combination of leg mobility and object manipulation. The video showed the robot fully squatting, grabbing a box, stepping back to pull it out, and then standing up again. Such details are the main points, since all the serious challenges are solved in this scenario.
Digit has no problems lifting objects, walking around, and not bumping into people. The fact that the robot can work effectively in such an environment is also reflected in other small things that are obvious to a human but often determine success for a robot.
An obvious question arises: why is Digit (or even any other biped robot) suitable for this kind of work? There are other robots already performing similar tasks in warehouses, but they typically have wheeled bases and control systems specifically designed for moving boxes. Such robots may cost less, be safer, and do the job more efficiently.
In essence, Digit cannot outperform a box-moving robot. However, it is important to note that Digit can do virtually anything thanks to its versatile humanoid design.
“Digit is very, very flexible automation. The value lies in what we do together and in having a robot that can work with boxes for three or four hours, then unload boxes from trailers for three or four hours, keeping pace with you. Many of these spaces are designed specifically for the human form factor, and a robot like Digit is capable of performing all these boring, repetitive tasks”
The value of a humanoid robot in human environments becomes apparent as soon as you start thinking about typical warehouse situations that would be trivial for a human to solve but impossible for wheeled robots. According to Hurst, Digit can even use a stool to reach items on high shelves.
Of course, you could design a wheeled robot with an extension system that lets it reach high shelves, but this all increases cost and complexity. The whole point of a general-purpose humanoid robot is that in a human environment you don't need to worry about surrounding problems. Digit has a mostly humanoid form factor, but it looks unusual.
“We did not set out to build humanoid robots. We set ourselves the task of solving the problem of mobility and took a methodical path toward understanding physical interaction with the world. Agility started with our robot Cassie, and one of the big problems with Cassie was that we lacked body inertia to counteract the leg swinging forward, which is why Digit has an upright torso. Our goal was to create a device that can move wherever people can and manipulate things in the world, and in the end we arrived at this form factor. For us this is a very different path than for the vast majority of humanoid robots, and our device has many subtleties that other devices don't have”
When Digit walks, its arms swing. This was not added to make the motion look right and "human-like." It looks right because it is literally right — the physics of mobility. The robot demonstrated impressive mobility skills. For now, the company is looking for the earliest, most suitable, and most valuable use cases. That is why it showed the demonstration in a limited-capability warehouse.
The robot has many impressive features, and the company does not intend to leave it as an ordinary packing robot. For example, Digit can not only walk perfectly on flat floors but also avoid tripping over curbs, steps, uneven surfaces, or other objects on the ground. It is being built for general purposes, but it is starting with the simplest use cases.
“We are trying to narrow the industry space to reach something where there is a value proposition with a partner, and deploy there. We can respect the complexity of a general-purpose use case and work toward early, profitable deployment instead of continuing to promote other applications”
This is a problem every robotics company faces sooner or later: the drive to move beyond demos toward real robot sales in a long-term, sustainable way. No matter how brilliant the robot's advanced capabilities are, if it cannot cost-effectively perform boring, dirty, or dangerous tasks, no one will pay for it. And in biped robotics, cost-effective usefulness is one of the biggest challenges.
