Video Friday: Spot Plays to Show Its Whole Body Manipulation
IEEE Spectrum Robotics’ Video Friday highlights demos and research on whole-body manipulation — including a Spot demo and work from RAI — and describes an ETH Zurich contact-centric learning framework that combines motion imitation with reinforcement learning to enable humanoid mobility and autonomous adaptation to new terrains. The excerpt also notes perception failure modes (e.g., a tire being misidentified in depth data).
Why this matters for robot rental
Advances in whole-body manipulation and contact-centric learning directly affect how robots handle heavy or awkward objects and move in unstructured environments — capabilities that increase the usefulness of rental robots for tasks like material handling, site work, and field service. These are research demos and frameworks (not commercial performance claims); the ETH description explicitly mentions autonomous adaptation from learned priors, while the Spot footage is a demonstration of physical manipulation and perception limits. Understanding both capability and failure modes helps operators decide when to rent autonomous-capable robots versus relying on teleoperation or human crews.
Source context
Video Friday is your weekly selection of awesome robotics videos, collected by your friends at IEEE Spectrum robotics. We also post a weekly calendar of upcoming robotics events for the next few months. Please send us your events for inclusion. CoRL 2026 : 9–12 November 2026, AUSTIN Humanoids 2026 : 6–9 December 2026, SANTA CLARA, CA Enjoy today’s videos! Some top-notch whole-body manipulation here from RAI, although it’s a little hard to tell whether there’s direct torso involvement, which is something that humans do with heavy or awkward stuff all the time. “Anything that shows up as a blob in a depth image can be taken for the tire, whether a person walking past or a stray object on the floor. In a few runs the robot left the tire and went after something else.” That “something else” seems suspiciously vague, doesn’t it? [ Robotics and AI Institute ] We present a contact-centric learning framework for versatile humanoid mobility in complex and unstructured environments. Our approach bridges motion imitation and reinforcement learning by leveraging human motion data as a training prior while enabling autonomous adaptation to previously unseen terrains. [ ETH Zurich Robotic Systems Lab ] You’ve literally heard me say it literally a thousand times: the only things that could possibly make Reachy Mini better would be some wheels and a whole bunch of fancy hats. But of course the only way I could possibly recommend the wheels is if Reachy Mini can totally wreck Microduck in a race. [ Pollen Robotics ] Luxonis is launching OAK 4 FFC, a 70 x 53 mm vision computer that lets humanoid and drone builders put up to five cameras wherever the robot needs them. The board runs sync, stereo depth and AI itself, and calibrates from the scene instead of a checkerboard. [ Luxonis ] Than
Transparent curation
Rent2Bots summarizes the linked reporting for market intelligence. The original publisher remains the source of record. We do not republish full articles, and local tags are used only when a location is supported by the source.
Read our editorial standards