Boston Dynamics next generation electric, as opposed to hydraulic, Atlas Robot is less messy than its predecessor which left a thin film of hydraulic fluid in its wake. Thanks to AI, Atlas is quickly becoming more capable of everyday tasks as demonstrated in a recent video.

In this video we see Atlas performing a mundane, repetitive task. Just the sort of task that you might imagine would be handed off to a robot. In fact the task demands skills that are second nature to human beings, but which robots find hard. This concept is referred to as Moravec's Paradox and was first articulated by Hans Moravec, a prominent AI and robotics researcher in the 1980s, who summed up the situation as:

"It is comparatively easy to make computers exhibit adult-level performance on intelligence tests or playing checkers, and difficult or impossible to give them the skills of a one-year-old when it comes to perception and mobility."

 We have already seen, in Atlas In Action With Reinforcement Learning how AI has transformed Altas's mobility. Now it's the turn of perception and object manipulation.

The task that Atlas performs in the video is part sequencing, a common logistics task in automotive facilities and similar manufacturing environments. In an automotive plant, many different car models and trim levels are assembled on the same line; installers need the correct individual parts for each of those vehicles in the right order. Sequencing is a pre-processing step where those parts are selected from containers are arranged correctly in a dolly for the cars that are being assembled that day.

What to a human is a simple task, pick up a car part and put it in the correct slot, breaks down into multiple steps, each requiring extensive knowledge about the environment. First, Atlas must detect and identify the object. Many car parts are either shiny and metallic or low-contrast and dark, making them hard for the robot’s camera to distinguish clearly. Adding AI into the mix greatly enhances its perceptual abilities.

Then Atlas needs to interact with the object, starting with having to reason about the location of the object in order to grasp it. Is it out in the open inside a container with limited line of sight?  Having picked the object, Atlas has to decide where to place it and how to get it there. Finally, Atlas needs to place the object accurately - just a couple centimeters off in any direction and the object might get stuck or simply fall. Atlas therefore also needs to be able to take corrective actions when things go wrong. If an insertion fails, for example, the robot can search for and pick up the dropped part from the ground.

These challenges require new methods and this has led to a new object pose tracking system for Atlas called SuperTracker which fuses different streams of information: robot kinematics, vision, and when necessary forces. Kinematics information from Atlas’s joint encoders determines where Atlas’s grippers are in space. When Atlas recognizes it has grasped an object, this information provides a strong hint about where the object should be as Atlas moves its body. By fusing kinematic data, Atlas can handle situations where objects are visually occluded or out of view of its cameras, as well as informing Atlas of when an object slips from its grasp. 

All-in-all, the video reveals that Atlas is well on the way to having the fundamental skills necessary to operating in real-world environments and that Moravec's Paradox can be assigned to history.

More Information

Boston Dynamics Atlas

Getting Real with Humanoids

Related Articles

Atlas In Action With Reinforcement Learning

Hydraulic Atlas Bows Out, Welcome Electric Atlas

To be informed about new articles on I Programmer, sign up for our weekly newsletter, subscribe to the RSS feed and follow us on Twitter, Facebook or Linkedin.

Comments

Make a Comment or View Existing Comments Using Disqus

or email your comment to: comments@i-programmer.info