Building Robots for Human Spaces

One of our biggest goals here at Agility Robotics is to build low-cost, adaptable robots that can be used in human spaces. In the words of Jonathan Hurst, our co-founder and chief technology officer, “That’s the dream.”

 Achieving that goal has hinged upon the iterative development of some early robot versions, first ATRIAS, then Cassie, two robots that informed the creation of our current commercially available robot, Digit. These were all bipedal robots, meaning they could walk on two legs, a type of engineering that was no easy feat.

Of all the different kinds of animals that are bipeds, almost all of them have some additional way to keep from falling over when they move. Studying biomechanics and applying what we’ve learned through engineering and computer simulation helped us create a robot that can walk, run, jump, and even skip.

In nature, animals have different ways of ensuring stability when they move. It might be a big tail. Or in the case of land birds, such as ostriches, it’s their wings, which catch the wind when they’re running and enable them to turn. Similarly humans have arms that pump back and forth, providing a similar effect, the ability to balance and keep from falling over. Digit’s arms swing in a related manner. This motion looks somewhat natural and familiar because the movement is capturing the same physics that a person or a bipedal animal uses when moving.

Another engineering hurdle involved making sure Digit could carry out basic warehouse tasks, such as lifting heavy boxes, activities that are too repetitive or physically demanding for humans. When a human picks up a box, they are estimating how heavy that box may be and how much force is needed to lift it. Very quickly, a human can adjust how much strength is required to pick up that box. 

With Digit, instead of muscles, the robot has actuators, mechanical components that enable physical movement.These actuators are incredibly force sensitive and do not use position trajectories, which allows Digit to pick up, carry, and set down large objects. These dynamically advanced robots combined with advanced machine learning are the future that we’re working toward.