What we’re watching next in humanoids

Boston Dynamics has stepped up its game with the latest version of Atlas, which boasts a staggering 42 degrees of freedom (DOF)—double what it achieved in 2022. This leap forward not only enhances Atlas's dexterity but also opens new avenues for applications in complex environments where precise manipulation is crucial.

The latest demonstration footage reveals Atlas executing a series of intricate tasks, from folding laundry to performing gymnastics. In one striking scene, the robot smoothly transitions from a handstand to a backflip, showcasing its impressive balance and control. The challenge remains, however, to translate this high-performance capability from controlled environments into real-world applications.

Engineering documentation shows that the new Atlas features a power source capable of sustaining a runtime of up to 90 minutes per charge, with a charging time of under two hours. This efficiency is a significant improvement over previous generations, which struggled with battery longevity during extended tasks. The payload capacity remains around 30 kg, suitable for a range of materials and tools, but raises questions about how well Atlas can handle heavier objects without compromising its agility.

Despite these advancements, one notable limitation persists: Atlas's operational speed in dynamic environments. While it excels in pre-programmed routines, real-time adaptability in unpredictable settings remains a hurdle. Lab testing confirms that Atlas still occasionally falters in obstacle avoidance strategies, particularly when navigating cluttered spaces or responding to sudden changes.

The technical specifications reveal that Atlas has improved its gait cycle efficiency, a critical factor for long-term operational viability. However, achieving seamless integration of these capabilities in field environments requires further tuning and testing. The current Technology Readiness Level (TRL) for Atlas stands at TRL 5—validated in relevant environments but not yet fully operational in the field.

In comparison to its predecessor, the enhancements in Atlas's agility, dexterity, and efficiency illustrate a significant evolutionary leap. The previous model, with only 21 DOF, could perform basic tasks but struggled with complex manipulations. The new version's capabilities suggest a shift towards more practical, real-world applications, though the pace of deployment will depend heavily on overcoming existing limitations.

### What we’re watching next in humanoids

Monitor Atlas’s integration of machine learning for real-time adaptability in dynamic environments.

Watch for updates on battery technology advancements that could extend operational runtimes.

Evaluate the impact of payload capacity on Atlas’s performance in practical applications, especially in logistics or manufacturing.

Assess the feedback from initial deployments to understand performance in unstructured environments.

Keep an eye on competition from other humanoid developers, particularly regarding speed and efficiency improvements.

Sources

IEEE Spectrum Robotics

The Robot Report

Boston Dynamics