Video Friday: Humanoid Robot Takes On Monkey Bars and More

IEEE Spectrum's Video Friday is a weekly roundup of notable robotics videos, compiled by the publication's robotics team. The outlet also maintains a calendar of upcoming robotics events and invites readers to submit events for inclusion. The roundup spans research demos, industrial deployments, and lighter experiments from labs worldwide.

Upcoming events include Humanoids Summit Seoul on 22–23 September 2026 in Seoul; IROS 2026 on 27 September–1 October 2026 in Pittsburgh; and CoRL 2026 on 9–12 November 2026 in Austin.

ETH Zurich's Robotic Systems Lab uses monkey-bar traversal to study how humanoid robots perceive thin, overhanging geometry while performing agile, accurate whole-body motions. The robot must jump to the structure, traverse it through sparse bar interactions, and land safely. The list of obstacles that a robot can cross to escape is getting shorter. The demonstration highlights whole-body agility and precise perception.

General Robotics Lab presents a robot with two heads, and the lab enthusiastically endorses the idea. The clip's message is clear: give robots two heads. The design raises questions about sensor fusion and human-robot interaction.

CRASAR recalls that 9/11 was the first documented use of robots for urban search and rescue and helped create the field of disaster robotics. Personnel began assembling on the afternoon of September 11 and worked the pile from late September 11 through October 2, when the last available robot failed. The robots found no survivors, but they located remains and helped search for routes through the rubble toward basements and stairwells where trapped firefighters might have gone.

Unitree has fully open-sourced its UnifoLM-WLA-1.0 embodied foundation model, claiming new state-of-the-art results across multiple benchmarks among open-source models worldwide. A single model coordinates desktop and whole-body mobile manipulation, supporting cross-task and cross-end-effector generalization. According to Unitree, one model drives whole-body coordination and achieves new SOTA results among open-source embodied models.

DRAGON Lab shows how a multilined aerial robot uses centroid and joint motion to achieve hybrid impedance–admittance control in contact-rich aerial manipulation tasks such as surface sliding. The work will be presented at IEEE IROS 2026.

RaiLab Kaist offers a clip that makes one want to stay away.

Texas A&M University's Advanced Vertical Flight Lab presents another edition of 'Things That Really Seem Like They Should Not Fly.'

ETH Zurich's Robotic Systems Lab introduces a unified reinforcement learning framework for agile and generalized legged locomotion. It incorporates a novel attention-based map encoder in the control policy. The goal is to achieve agility across terrains while improving generalization and interpretability, especially under occlusions and sparse footholds. Existing methods often rely on end-to-end sensorimotor models with limited generalization, or they target generalized locomotion but struggle with visual occlusions.

Unitree shows what may be the killer app for humanoid robots, though the company suggests it probably should not be called that.

Flexiv demonstrates a dish-washing robot. The demo may avoid many of the genuinely difficult aspects of doing dishes: not just water and slippery soap, but also identifying when a dish is dirty and when it is actually clean. The field still faces major challenges in perception and dexterity.

DEEP Robotics presents a robot delivering a burrito to a hiker on a rainy mountain hike. The question is whether such a delivery is something anyone actually needs.

ANYbotics CEO and co-founder Péter Fankhauser discusses why legged robots became the company's way into the world's most demanding industrial plants. He also explains what it took to certify one for explosive atmospheres after experts called it impossible, and where autonomous industrial work goes next. The bet behind ANYbotics, he says, is that legged robots are the right form factor for the world's most demanding industrial plants. AI has transformed the digital world—writing code, generating images, and reasoning in language—but the physical world of power plants, fuel, steel, and chemicals has barely been touched.