Boston Dynamics' flagship Atlas humanoid robot picks up and places heavy automotive struts with ease in new footage.
In this latest demonstration of Atlas' capabilities, the robot uses only its on-board sensors to detect the objects before using its grippers to pick up the struts from storage and insert them into a nearby flow cart. The footage also gives us a glimpse of the action from Atlas' perspective.
Atlas — which Boston Dynamics describes as the "world's most dynamic humanoid robot" — made its public debut in July 2013 and has been iterated upon over the years. The footage of Atlas handling automotive struts is the latest in a series of demonstrations, with previous footage showing Atlas running a.nd performing parkour.
Related: Robot hand exceptionally 'human-like' thanks to new 3D printing technique
Last year, the company released the first footage of Atlas using grippers to handle objects and deliver them to a human companion. Atlas also navigated elevated platforms by running and jumping, and even performed a backflip.
While handling struts isn't as visually spectacular as these previous efforts, it demonstrates one of the most plausible real-world deployments of humanoid robots; working in factory and warehouse environments.
Since Boston Dynamics debuted Atlas more than 10 years ago, other companies including Tesla and Figure have launched their own humanoid robots designed to be deployed in industrial settings.
Tesla teased its Optimus Gen-2 robot in December 2023 by showing it walking, doing squats and handling delicate objects like eggs. In January, Figure also released promotional footage of its Figure 01 machine making coffee — having been trained using just 10 hours of footage. This latter footage was significant because it purported to show the robot using artificial intelligence (AI) to learn by watching humans perform certain tasks.
Boston Dynamics chief strategy officer Marc Theermann shared the Atlas video on LinkedIn and quipped that the team "considered making coffee or picking up an egg," but instead decided to focus on "real world manipulation challenges."
