Boston Dynamics Unveils an Electric Atlas Humanoid Robot

SUMMARY

In 30 seconds

Matthew Berman reacts to Boston Dynamics’ April 2024 unveiling of an all-electric Atlas following the retirement of its hydraulic predecessor. He discusses its movement, the appeal of humanoid form, the company’s existing Spot platform and the infrastructure needed to deploy robots at work. The demonstration shows a prototype, not proof of commercial readiness or general autonomy. Boston Dynamics said it would test applications with a small group of customers, beginning with Hyundai, over the following years.

1Electric Atlas follows hydraulic AtlasBoston Dynamics retired its hydraulic research platform and introduced an all-electric successor designed for future real-world applications.
2Movement in the first demonstrationBerman is impressed by the robot’s unusual joint rotation and ability to rise from the floor, shown in a controlled demonstration.
3Why humanoid form?Berman points to human-oriented environments and the availability of human video data as possible advantages, while their effectiveness requires testing.
4A testing partnership with HyundaiThe company said Hyundai would be an early testing ground and that selected customer applications would be developed over the coming years.
5The deployment ecosystemBoston Dynamics describes the need for IT, safety, workflows and support alongside hardware, citing over 1,500 Spot deployments at the time.
6Software and facility modelsThe company discusses simulation, model-predictive control, reinforcement learning, computer vision and facility data; Berman speculates about competing approaches.

Takeaways

The new Atlas shown in 2024 uses electric actuation, succeeding Boston Dynamics’ retired hydraulic research robot.
The unveiling shows the prototype standing from the floor and moving its joints beyond human ranges; it does not demonstrate a complete industrial task.
Boston Dynamics planned to test and refine applications with Hyundai and other selected partners over several years.
A practical deployment needs workflows, safety, connectivity, operational support and facility data as well as the robot itself.
Berman’s comparison of control strategies among robot companies is an interpretation, not a verified description of their complete systems.

TopicsRobotics Autonomous Systems AI Adoption

FULL TEXT

Read the edited text by chapter

Show transcript

Transcript only

A new electric Atlas

Berman introduces the electric prototype and plays its unveiling. He contrasts it with the retired hydraulic Atlas and reflects on its design.

Boston Dynamics retired its hydraulic Atlas research robot and unveiled a new all-electric version in April 2024. The brief demonstration is striking: Atlas rises from the floor through unusual joint rotations and walks away. I am impressed by the movement and futuristic appearance, but the footage shows a controlled prototype demonstration, not a complete industrial deployment.

SLIDE 1 / 4Electric Atlas begins to riseThe launch demonstration shows the robot using a non-human range of joint motion to get up from the floor.

The speed, movement and head design make this one of the most interesting-looking robots I have seen. Appearance may matter if humanoid robots eventually enter homes, although that is a future possibility rather than a demonstrated use for this machine. Boston Dynamics describes the new platform as all-electric, in contrast to the hydraulic actuation of the retired Atlas.

I have followed Boston Dynamics’ videos for years, and the progression of Atlas is remarkable. Earlier research demonstrations included dynamic movement and parkour. The company says the electric platform builds on that history while aiming at real-world applications; it does not claim that the 2024 prototype is already generally available for commercial work.

Spot and the humanoid form

He compares the growing robotics field with the commercial Spot platform. He suggests reasons robots may adopt human-like forms.

The robotics landscape now includes many companies, including 1X and Figure. Boston Dynamics already has a commercial quadruped, Spot, used for tasks such as inspection in industrial or hazardous environments. Its existing deployments should not be confused with completed deployments of the new Atlas.

Why build humanoid robots? I see two possible reasons. Much of the built environment and many workflows are designed around human bodies, while the abundance of videos of people may provide useful training data. Neither factor alone guarantees that a humanoid can safely replace a person or that human video transfers directly to robot skills.

Hyundai testing and manipulation

Boston Dynamics plans to begin testing Atlas applications with Hyundai. Berman discusses the robot’s movement and planned gripper variations.

SLIDE 2 / 4Hyundai as an early testing groundBoston Dynamics’ announcement describes its planned partnership to test Atlas applications.

Boston Dynamics says the journey will begin with Hyundai, its investor and a potential automotive-manufacturing testing ground. Its [2024 announcement](https://bostondynamics.com/blog/electric-new-era-for-atlas/) says it plans to test and refine Atlas applications with a small group of customers over the following years. This is a commercialization plan, not evidence that the newly unveiled robot was already ready for broad sale.

I find Atlas’ motion exceptionally impressive, particularly the way it gets up from a flat position. Boston Dynamics says the electric version is intended to be stronger and have a broader range of motion than previous generations. Those are company claims about the platform; the brief launch video does not establish comparative performance across every humanoid robot.

SLIDE 3 / 4Planned strength and gripper developmentThe company describes its goals for the electric robot and its exploration of different grippers.

The company says its previous hydraulic Atlas could manipulate heavy, irregular objects and that it is exploring new gripper variations for customer needs. I compare that with Spot’s configurable attachments: different jobs may call for different tools. The specific commercial manipulation abilities of the new Atlas still have to be tested.

The deployment ecosystem and software

Hardware must be supported by infrastructure, safety and operational processes. Berman discusses facility models, simulation and machine learning.

A capable robot is only part of a commercial solution. Boston Dynamics points to IT infrastructure, connectivity, employee buy-in, workflows, safety standards and operational processes. At the time of its announcement, the company cited more than 1,500 Spot deployments as experience helping organizations work with mobile robots.

SLIDE 4 / 4Deployment infrastructure and Spot experienceThe announcement lists operational requirements and cites more than 1,500 Spot deployments.

Boston Dynamics argues that detailed facility models and operational data can help humanoids work effectively, and says Spot can help create digital twins. I wonder how that compares with the end-to-end neural approaches discussed by Tesla and Figure. That is my reading of limited public information, not proof that their robots lack maps or other supporting systems.

The company also describes experience with simulation and model-predictive control, plus reinforcement learning and computer vision. I wonder whether it will build everything itself or use tools from companies such as NVIDIA. Its Orbit software provides centralized fleet and site-data management; the 2024 announcement said it was available for Spot, with Stretch and Atlas integration planned.

Mobility beyond human limits

The company explains why it builds legged robots and why Atlas need not share human joint limits. Berman returns to the demonstration.

Boston Dynamics traces Atlas back through PETMAN and the hydraulic robot’s parkour research. It says legged machines can balance and move dynamically across difficult or unfamiliar terrain, and that a humanoid shape fits environments designed for people. The shape, however, does not require the robot to copy the limits of human joints.

That design choice is clear in the launch clip: Atlas rotates its legs and pushes itself up in a way a person cannot. I find the approach exciting and want to see more demonstrations. This is a mobility demonstration, not proof of autonomous performance across factory tasks; the video closes by replaying the clip.

The video description also promotes Pinecone, a vector database service, but does not specify the terms of that promotion. The rest of the presentation focuses on Boston Dynamics’ public announcement and Berman’s reaction.