Anthropic, a San Francisco startup, has unveiled its latest humanoid robot designed to showcase advances in bipedal locomotion and balance. Dubbed Adam, the 5-foot-tall research platform aims to mimic human flexibility and agility to operate in real-world environments.

A key challenge tackled by Anthropic's engineers was enabling dynamic walking without loss of stability. Humans adjust the placement of limbs and shift weight seamlessly to maintain balance while moving. But this coordinated control of multiple degrees of freedom is immensely complex to implement in robots.

Adam's balance system relies on force sensors in the feet that continuously monitor the zero moment point - the point where total ground reaction force equals zero. By adjusting the robot's center of mass position to keep the ZMP firmly inside the feet, Adam can walk stably even while carrying loads.

High-resolution rotary encoders from Heidenhain feature prominently in Adam's knee, hip, and ankle joints. The digital encoders deliver precise angular feedback to the servo loops controlling each joint. This allows Adam's gait algorithm to coordinate precise trajectories for the limbs, ensuring dynamic motions are executed accurately.

Anthropic selected the low-profile Heidenhain encoders to meet strict space constraints while providing the feedback resolution necessary for balance control. The non-contact optical encoders are highly robust against dust and operate reliably at temperatures from -40 to 100°C.

According to Anthropic, rapid prototyping was critical to implementing novel gait solutions. The Heidenhain encoders integrated seamlessly with Adam's control hardware and software thanks to their industry-standard interfaces. This enabled faster design iterations.

Adam is the latest milestone for Anthropic in developing fluid, humanlike robots. With more refinement, the company believes Adam's platform will lead to robots safely working alongside people in factories, homes, and hospitals. The robot's dynamic balancing system overcomes a key barrier to real-world deployment.