Videos: Synchronized Dancing Robots, Dorm Movers, More

Video Friday is your weekly choice of superior robotics movies, collected by your mates at IEEE Spectrum robotics. We additionally publish a weekly calendar of upcoming robotics occasions for the following few months. Please send us your events for inclusion.

CLAWAR 2025: 5–7 September 2025, SHENZHEN, CHINA

ACTUATE 2025: 23–24 September 2025, SAN FRANCISCO

CoRL 2025: 27–30 September 2025, SEOUL

IEEE Humanoids: 30 September–2 October 2025, SEOUL

World Robot Summit: 10–12 October 2025, OSAKA, JAPAN

IROS 2025: 19–25 October 2025, HANGZHOU, CHINA

Get pleasure from at this time’s movies!

Boston Dynamics is again and their dancing robot dogs are greater, higher, and bolder than ever! Watch as they bring about a “lifeless” robotic to life and unleash a by no means earlier than seen synchronized dance routine to “Good Vibrations.”

And rather more curiously, right here’s a dialogue of how they made it work:

[ Boston Dynamics ]

I don’t particularly care whether or not a robot falls over. I care whether or not it will get itself again up once more.

[ LimX Dynamics ]

The robotic autonomously connects a number of wires to the setting utilizing small flying anchors—drones outfitted with anchoring mechanisms on the wire ideas. Guided by an onboard RGB-D digicam for management and environmental recognition, the system permits wire attachment in unprepared environments and helps simultaneous multi-wire connections, increasing the operational vary of wire-driven robots.

[ JSK Robotics Laboratory ] at [ University of Tokyo ]

Thanks, Shintaro!

For a robotic that hardly has a face, that is some fairly good emoting.

[ Pollen ]

Studying abilities from human motions gives a promising path towards generalizable insurance policies for whole-body humanoid management, but two key cornerstones are lacking: (1) a scalable, high-quality movement monitoring framework that faithfully transforms kinematic references into sturdy, extraordinarily dynamic motions on actual {hardware}, and (2) a distillation method that may successfully be taught these movement primitives and compose them to unravel downstream duties. We tackle these gaps with BeyondMimic, a real-world framework to be taught from human motions for versatile and naturalistic humanoid management through guided diffusion.

[ Hybrid Robotics ]

Introducing our open-source metal-made bipedal robotic MEVITA. All parts could be procured by e-commerce, and the robotic is constructed with a minimal variety of components. All {hardware}, software program, and studying environments are launched as open source.

[ MEVITA ]

Thanks, Kento!

I’ve at all times thought that having the ability to lease robots (or exoskeletons) that will help you transfer furnishings or in any other case carry stuff can be very helpful.

[ DEEP Robotics ]

A brand new research explains how tiny water bugs use fan-like propellers to zip throughout streams at speeds as much as 120 physique lengths per second. The researchers then created an identical fan construction and used it to propel and maneuver an insect-sized robotic. The invention gives new prospects for designing small machines that might function throughout floods or different difficult conditions.

[ Georgia Tech ]

Dynamic locomotion of legged robots is a essential but difficult matter in increasing the operational vary of mobile robots. To attain generalized legged locomotion on numerous terrains whereas preserving the robustness of learning-based controllers, this paper proposes to be taught an attention-based map encoding conditioned on robotic proprioception, which is educated as a part of the end-to-end controller utilizing reinforcement learning. We present that the community learns to concentrate on steppable areas for future footholds when the robotic dynamically navigates numerous and difficult terrains.

[ Paper ] from [ ETH Zurich ]

Within the fifth installment of our Moonshot Podcast Deep Dive video interview sequence, X’s Captain of Moonshots Astro Teller sits down with Google DeepMind’s Chief Scientist Jeff Dean for a dialog in regards to the origin of Jeff’s pioneering work scaling neural networks. They talk about the primary time AI captured Jeff’s creativeness, the earliest Google Mind framework, the group’s stratospheric developments in image recognition and speech-to-text, how AI is evolving, and extra.

[ Moonshot Podcast ]