AthenaZero researcher explains why the baseball-playing robot has yielding joints

A robot arm can need a forceful throw and a gentle catch within the same task. AthenaZero was built to explore that range. In an October 9 Tech Xplore interview, researcher Andrew S. Morgan explained how the Robotics and AI Institute’s two-armed system uses its physical design to support the throwing, catching and batting demonstrated in Science Robotics.

Rigid arms, yielding joints

AthenaZero has a central torso, two arms with hands, and motors and gears. Its arm segments are rigid, but its joints can give way under outside forces. The robot is not made entirely soft, yet it can respond when an object or person pushes against it.

Morgan contrasts that approach with conventional industrial robots built for strength, stiffness and precise movement. Those properties help with lifting large objects, he said, but make softer contact with the surroundings difficult. His team wanted a machine capable of delicate interactions as well as powerful movements.

The other design priority is low inertia: reducing how strongly the arm resists changes in motion. Lower effective inertia can help an arm speed up and slow down more readily. The researchers are investigating those mechanical properties, rather than treating improved control software as the only route to better object handling.

“Our paper reopens the conversation about how robot manipulators (arms) should be built in order to contact the world more delicately.”

Baseball makes contact and timing visible

The team chose baseball because it brings several manipulation challenges into one test setting. Throwing requires coordinated motion to transfer force into a ball. Catching requires responding to an approaching object and managing contact forces. Batting adds two-handed coordination, with both arms positioning and swinging the bat at the right moment.

The original paper, submitted to arXiv on September 15, supplies quantitative context. The authors report throws above 30 meters per second, and catching and batting at speeds above 14 meters per second over a short 7.3-meter distance. They describe these as interactions on human-comparable timescales, where milliseconds matter.

The demonstrations also included batting practice and games of catch between robots and between a human and a robot. The authors present those variations as evidence of adaptability in tasks requiring high acceleration.

A mechanical experiment, not a job-ready robot

The evaluations demonstrate the potential of the mechanical design, not broad real-world job capability. The larger aim is dexterous manipulation: handling objects skillfully and reliably. Morgan described the baseball tasks as proxies for harder problems, rather than the final purpose of the project.

Morgan said custom actuators—the mechanisms producing movement—define AthenaZero’s mechanical properties and distinguish it from commercially available robots. He called it the team’s first low-inertia design attempt, citing advances in electric motors and computation as catalysts.

The researchers plan further studies to improve AthenaZero’s movements and manipulation skills, then assess a wider range of tasks. Their stated goal is reliable object handling in dynamic, real-world settings.