First reported Sep 19 — we wrote this up later than the original.
Video Friday: Water-Pumping Robots and a Humanoid Factory Milestone
This week's robotics roundup spans an undulating water robot, an underwater quadruped, and UBTECH's rapid humanoid production ramp.
Every week, IEEE Spectrum rounds up notable robotics videos from labs and companies around the world, and this week's batch ranges from playful demos to serious engineering milestones.
One of the more inventive builds comes from NYU Tandon School of Engineering, where researchers created an undulatory robot called WorMa that pumps water between its head and tail to change how it moves. Shifting water toward the head increases the force pressing the robot's front contact points against a surface, boosting traction on inclines. That head-heavy configuration was the only one that let WorMa climb a 19.5-degree slope — every other setup slid back down — and it cut the robot's cost of transport by at least 33% compared with alternative water distributions. Clearing steps required a more complex sequence: WorMa approaches with water in its head for grip, shifts water to the tail and raises its head and neck so the tail can push it forward, then anchors its head on the step's edge before shifting the water back. Using this method, the robot cleared steps up to 15 centimeters high.
Underwater mobility got attention too. Researchers at the Norwegian University of Science and Technology's Autonomous Robots Lab built a reproducible underwater quadruped using custom waterproof motor housings machined from polyoxymethylene plastic, sealed with off-the-shelf O-rings and dynamic shaft seals. The team derived a simplified dynamics model capturing how drag on the robot's spherical end effectors transmits torque to its floating base, then built a closed-loop attitude controller using an error formulation on the special orthogonal group SO(3). Tested in a water tank, the robot successfully tracked target roll, pitch, and yaw orientations.
On the manufacturing side, UBTECH shared a look at what it calls its benchmark humanoid robot smart factory, where one robot now rolls off the line every 10 minutes. The company says it scaled production from 1,000 to 10,000 units in less than nine months — though as IEEE Spectrum's editors noted, the more interesting question may be what all those robots will actually do.
Elsewhere, MIT showed off a robotic optics lab in which an arm autonomously builds, runs, and dismantles laser-based optics experiments. The system arranges mirrors, lenses, and other components housed in custom QR-encoded mounts to align a red laser beam, and MIT suggests the approach could accelerate testing for solar cells, sensors, displays, and quantum technologies.
Agility Robotics posted footage focused on the safety systems built into its latest Digit 5 humanoid, while Figure shared a home-robot demo that drew a skeptical reaction from Spectrum's editors, who wrote they wanted proof it's "safe to be in my home and a realistic alternative to a human." CLIMB Lab showed off a quadrotor built from two paired birotor units, and ETH Zurich's Robotic Systems Lab posted a clip whose final 30 seconds stood out as a highlight.
Smaller items rounded out the week: a Kickstarter campaign for Kubi 2.0, a desktop telepresence robot aimed at connecting homebound or hospitalized students with classrooms; footage from DEEP Robotics showing an unusual use case for a wheeled quadruped; Japan's Ugo Nova semi-humanoid robot aimed at industrial deployment; a two-finger gripper demo from Daimon Robotics; highlights from NASA's Lunabotics student rover competition ahead of its 2027 edition; and a CMU Robotics Institute seminar on the EU-funded RoboRoyale project, which uses robotics and AI to monitor honeybee colonies and support hive health.
As always, the full lineup — with all the videos — is available on IEEE Spectrum's site.
Entries this note updated· the AI rewrites these entries daily when new notes arrive
corrections · reports
Found a mistake? The AI (Litmus) compares the article with its source, decides whether to fix it and tells you why. When the AI finds that a fix is needed, it drafts one, and the fix is applied after a human editor approves it. Every fix is listed here and in the changelog.
Sources
This story was written by Robopedia based on the sources below.
Learn more