It works by dragging and adjusting Spot's pre-programmed moves on a timeline. The video describes some of the movements that Spot can perform, including body, step, dynamic transition, and knee movements. Within each of those categories, there are subcategories of movements that can be dragged to the timeline. But the most interesting video was an unprecedented behind-the-scenes account of how Boston Dynamics engineers developed and trained Atlas to run the parkour track.
The video shows some of Atlas's failures and is a break with the company's tradition of showing highly polished results from this work. The video and an accompanying blog post provide very important information about the challenges of creating humanoid robots. Stretch has a box-like base with a set of wheels that can move in all directions. At the top of the base is a large robotic arm and a perception mast.
The robotic arm has seven degrees of freedom and an array of suction cups that can grip and lift boxes. The perception mast uses cameras and sensors driven by computer vision to analyze its environment. It also contains elements of Pick, a computer vision-based depalletizing solution mentioned in the press release that declared Hyundai's acquisition of Boston Dynamics. So, the basic premise is that if you can make a robot do backflips, jump platforms, jump barriers, and run very narrow paths, you would have taught it all the other basic physical and movement skills that all humans possess.
Stretching may seem boring compared to other things Boston Dynamics has done in the past. Even Atlas, which is by far the most advanced bipedal robot, is a long way from achieving the smooth and versatile motor skills of humans. First, Boston Dynamics will leverage its decades of experience to boost the versatility of its robots without sacrificing robustness and safety. Software engineers create three-dimensional environments in which a virtual version of the robot can be trained at a very fast pace and without the costs of the physical world.
But in chaotic, unknown or ill-defined environments, reliance on rules makes robots notoriously bad at dealing with anything that can't be accurately predicted and planned beforehand. Aligning these two goals is very difficult, and proof of this is that Boston Dynamics has changed ownership several times in the last decade, moving from Google to SoftBank and Hyundai. Hmmmm, I doubt you think that robotizing people outside their jobs would be a great idea if you were one of the victims. Turning a robot into a good teammate can be difficult, because it can be difficult to find the right amount of autonomy.
Both have found interesting applications in different industries, and with Hyundai's manufacturing capability, Boston Dynamics could turn them into profitable companies. Atlas may not be a commercial product anytime soon, but it is providing Boston Dynamics and the robotics industry with a great platform to learn about the challenges nature has solved. An important thing when creating a new configuration is that you need to look at how it affects the weight of your robot, since it cannot exceed 114 kilograms. Boston Dynamics has accumulated a mini zoo of robotic beasts over the years, with names like BigDog, SandFlea and WildCat.