Almost a year ago, we announced our new Atlas® robot—faster, stronger, more compact, and less messy. We’re designing the world’s most dynamic humanoid robot to do anything and everything, maybe even change work as we know it.

How do we get there? One step at a time.

At Boston Dynamics, we’ve learned a few things about building robots that make an impact. At the top of that list is the art of breaking down big goals into practical problems. A great problem lets you tune out the noise and pull the threads that really matter. It should:

  • Target the hardest, riskiest parts of the technical problem
  • Provide real value over existing solutions

These factors help you find the fastest path to a complete, useful solution—capable enough to outperform people and other automation technology, and grounded enough for the task at hand.

We’re fortunate to be partnered with Hyundai, a world leader in manufacturing. When something hasn’t been automated in a Hyundai facility, there’s a good reason for it. Together, we found the ideal first problem for Atlas: starting with part sequencing, we are putting Atlas to work in manufacturing facilities and opening up the future.

What Is Part Sequencing?

Part sequencing is 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 arranged correctly for the cars that are being assembled that day.

Thousands of parts flow into these manufacturing facilities from suppliers, arriving in single-SKU containers each with one part variant; a person then takes those parts and re-sorts them onto an output dolly in the right order for what’s needed on the line.

For this particular test, we have a bunch of inlet containers that contain engine covers from the original suppliers that are being sequenced into a dolly. These dollies would then be presented next to the assembly line, so that assembly workers could efficiently retrieve those parts and put them into the cars as they’re appearing on the line.

Sequencing is Hard – That’s Good!

Part sequencing bundles up many of the core problems that are both challenging for conventional automation and important for making a humanoid that delivers on general purpose use. Next time you watch a robot video, ask yourself, is it grappling with these complexities autonomously? Atlas is.

Task Diversity

A complete part sequencing solution will require Atlas to manipulate thousands of different parts and complete different types of pick, carry, and place tasks. These parts are different sizes, irregular shapes, and a range of weights. In order to solve these problems, we’ve built on our solid foundation of robotics techniques and behavior creation pipelined around Reinforcement Learning  and Foundation Models. 

Behavior Complexity

Sequencing is a multi-part behavior that requires high reliability. Atlas needs to understand all of the different ways a pick, transport, or insert can go wrong, recognize if one of those things happen, and know how to handle them.

A robot that needs tons of painstaking work to program every edge case, requires frequent help from a person, or damages parts isn’t valuable! This aspect of sequencing drove us to rebuild our behavior authoring system for complex manipulation tasks. This system is the springboard for both effective human and AI behavior authoring.

Environmental Complexity

Next, we wanted an application that reflected the real world—messy, complicated, and designed for people. Part sequencing involves, both literally and figuratively, a lot of moving pieces.

The environment changes, with new inlet containers arriving and filled output dollies being moved out. The parts are also not stored in robot-friendly configurations. Atlas needs to pick them from containers with limited visibility—the parts are in dark cubbies occluded by walls—and use its full range of motion to reach parts that are low to the ground or at the top of the container. Keeping up with complexity in an initial application sets us up for success in future applications.

This complexity is also why part sequencing has real ROI behind it. It’s monotonous and ergonomically challenging—involving repetitive motions like lifting, squatting, and twisting with heavy parts like strut assemblies. Automating the process with Atlas creates opportunities to reskill people into other areas, reduce the risk of injury, and improve employee retention, all while improving throughput and efficiency.

What’s Next

The promise of humanoids is that they can easily slide into existing processes and adapt to the environment. They have the potential to be as generalizable and versatile as people themselves. The future is wide open, but a robot that can theoretically do everything is still only as useful as the tasks it can do well.

Atlas has strength, range of motion, dexterity, and now the spark of intelligence to take on the human world. Beyond the technological foundations, we are going full speed to bring Atlas to scale—starting with testing in Hyundai facilities this year and then working with additional pilot customers to iterate, ideate, and expand Atlas’ capabilities. 

Ultimately, we believe the scale of adoption of autonomous mobile robots will mirror the rise of other revolutionary technologies that shape how we live and work.

With industry-leading hardware, a new generation of AI breakthroughs, access to real-world data, a parent company that understands mass manufacturing, and Boston Dynamics’ experience taking behaviors from the lab to commercial environments, Atlas is ready for that future.