<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>RoboticHub.ai - Research &amp; Signals</title>
    <link>https://robotichub.ai</link>
    <description>Research &amp; Signals coverage from RoboticHub.ai. Curated coverage of Physical AI: robotics, autonomous systems, defense technology and smart infrastructure.</description>
    <language>en</language>
    <lastBuildDate>Wed, 11 Nov 2026 19:30:45 GMT</lastBuildDate>
    <ttl>60</ttl>
    <atom:link href="https://robotichub.ai/feed/research-signals.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>Berkeley and NVIDIA release AI agents that code industrial robots in simulation</title>
      <link>https://robotichub.ai/article/article_1794425444579</link>
      <guid isPermaLink="true">https://robotichub.ai/article/article_1794425444579</guid>
      <pubDate>Wed, 11 Nov 2026 19:30:45 GMT</pubDate>
      <category>Research &amp; Signals</category>
      <category>Robotics</category>
      <category>Industrial Automation</category>
      <category>Computer Vision</category>
      <category>Breakthrough</category>
      <category>PoC</category>
      <description><![CDATA[<p><img src="https://www.therobotreport.com/wp-content/uploads/2026/11/robosimstudio.jpg" alt="" /></p><p>UC Berkeley and NVIDIA have released Graph-as-Policy (GaP) and VABAR, an open-source framework that enables AI coding agents to program industrial robots by iterating in simulation. The approach eliminates the need for large training datasets and produces human-readable code that engineers can audit before production deployment, potentially compressing robot deployment timelines from months to hours.</p><p><strong>Key signal:</strong> Berkeley and NVIDIA release open-source AI agent framework that codes and simulates industrial robots without large training datasets.</p><p>Source: The Robot Report — Can AI agents program industrial robots without training data? Berkeley says yes (https://www.therobotreport.com/can-ai-agents-program-industrial-robots-without-training-data-berkeley-says-yes/)</p><p><a href="https://robotichub.ai/article/article_1794425444579">Read the full article on RoboticHub.ai</a></p>]]></description>
    </item>
    <item>
      <title>Boston Dynamics tests wheeled Atlas to merge classical control with whole-body RL</title>
      <link>https://robotichub.ai/article/article_1794262453861</link>
      <guid isPermaLink="true">https://robotichub.ai/article/article_1794262453861</guid>
      <pubDate>Mon, 09 Nov 2026 22:14:14 GMT</pubDate>
      <category>Research &amp; Signals</category>
      <category>Robotics</category>
      <category>Humanoids</category>
      <category>Autonomous Systems</category>
      <category>PoC</category>
      <category>Breakthrough</category>
      <description><![CDATA[<p><img src="https://www.therobotreport.com/wp-content/uploads/2026/11/wheeled-atlas-humanoid.jpg" alt="" /></p><p>Boston Dynamics released video of its Atlas humanoid using scooter wheels attached to ankle actuators, controlled via a hybrid system that merges classical model-based control with whole-body reinforcement learning. The experiment demonstrates that Atlas can dynamically adjust pose and force in real time to handle unmodeled disturbances and irregular surfaces, shifting between fast rolling and articulated legged motion. It&apos;s a research testbed, not a product, but signals growing interest in hybrid locomotion approaches as humanoid developers weigh commercial trade-offs between human-like form factors and operational versatility.</p><p><strong>Key signal:</strong> Boston Dynamics tests Atlas with scooter wheels, merging classical control and reinforcement learning for real-time adaptation on unstructured terrain.</p><p>Source: The Robot Report — Boston Dynamics adds wheels to Atlas humanoid to test whole-body control (https://www.therobotreport.com/boston-dynamics-adds-wheels-to-atlas-humanoid-to-test-whole-body-control/)</p><p><a href="https://robotichub.ai/article/article_1794262453861">Read the full article on RoboticHub.ai</a></p>]]></description>
    </item>
    <item>
      <title>AGIBOT&apos;s GO-3 Preview tops three embodied AI benchmarks with full-lifecycle learning</title>
      <link>https://robotichub.ai/article/article_1793945811546</link>
      <guid isPermaLink="true">https://robotichub.ai/article/article_1793945811546</guid>
      <pubDate>Fri, 06 Nov 2026 06:16:51 GMT</pubDate>
      <category>Research &amp; Signals</category>
      <category>Robotics</category>
      <category>Computer Vision</category>
      <category>Edge AI</category>
      <category>Breakthrough</category>
      <category>PoC</category>
      <description><![CDATA[<p><img src="https://roboticsandautomationnews.com/wp-content/uploads/2026/11/GO-3_Training_Data.gif" alt="" /></p><p>AGIBOT has released GO-3 Preview, an embodied foundation model that achieved first place on the RoboDojo, RoboTwin, and RoboColiseum benchmarks. The system implements &quot;full-lifecycle scaling&quot; — connecting large-scale pre-training, hybrid post-training across 100 physical robots and simulation, and post-deployment learning — allowing robots to develop general capabilities and refine them through real-world work.</p><p><strong>Key signal:</strong> AGIBOT&apos;s GO-3 Preview tops RoboDojo, RoboTwin, RoboColiseum benchmarks with full-lifecycle learning across pre-training, post-training, and deployment.</p><p>Source: Robotics and Automation News — AGIBOT unveils GO-3 Preview and takes top spot on three embodied AI benchmarks (https://roboticsandautomationnews.com/2026/11/05/agibot-unveils-go-3-preview-and-takes-top-spot-on-three-embodied-ai-benchmarks/106711/)</p><p><a href="https://robotichub.ai/article/article_1793945811546">Read the full article on RoboticHub.ai</a></p>]]></description>
    </item>
    <item>
      <title>Robot moth grows UV sensors using circulating artificial blood</title>
      <link>https://robotichub.ai/article/article_1793908308831</link>
      <guid isPermaLink="true">https://robotichub.ai/article/article_1793908308831</guid>
      <pubDate>Thu, 05 Nov 2026 19:51:49 GMT</pubDate>
      <category>Research &amp; Signals</category>
      <category>Robotics</category>
      <category>Sensors</category>
      <category>Edge AI</category>
      <category>PoC</category>
      <category>Breakthrough</category>
      <description><![CDATA[<p><img src="https://assets.newatlas.com/dims4/default/ab2597d/2147483647/strip/true/crop/2417x1269+231+0/resize/1200x630!/quality/90/?url=https%3A%2F%2Fnewatlas-brightspot.s3.ap-southeast-2.amazonaws.com%2Fcc%2F4a%2F4acd235442af8a74baa0b5d94f47%2Frobot-moth-gained-an-e-2.jpg&amp;na.image_optimisation=0" alt="" /></p><p>Researchers at the University of Tartu and the Italian Institute of Technology have built a moth-shaped robot that grows new sensors on demand by circulating photoreactive chemicals through artificial veins in its 3D-printed body. When UV light hits the wing surface, pyrrole in the fluid polymerizes into a conductive patch that functions as a photodetector, demonstrating environmentally-triggered hardware reconfiguration without pre-installed components.</p><p><strong>Key signal:</strong> Tartu and IIT robot grows UV sensors in its wings using photoreactive artificial blood triggered by environmental light exposure.</p><p>Source: New Atlas — Robot moth grows new sensor using its own blood (https://newatlas.com/robotics/tartu-iit-robot-moth-artificial-blood-uv-sensor/)</p><p><a href="https://robotichub.ai/article/article_1793908308831">Read the full article on RoboticHub.ai</a></p>]]></description>
    </item>
    <item>
      <title>LiberAI launches Liber-0 world action model trained on 2.2M hours of human experience</title>
      <link>https://robotichub.ai/article/article_1793771852253</link>
      <guid isPermaLink="true">https://robotichub.ai/article/article_1793771852253</guid>
      <pubDate>Wed, 04 Nov 2026 05:57:33 GMT</pubDate>
      <category>Research &amp; Signals</category>
      <category>Robotics</category>
      <category>Humanoids</category>
      <category>Computer Vision</category>
      <category>Launch</category>
      <category>Breakthrough</category>
      <description><![CDATA[<p><img src="https://roboticsandautomationnews.com/wp-content/uploads/2026/11/Liber-0-achieves-36.0-overall-success.png" alt="" /></p><p>LiberAI has released Liber-0, a World Action Model pretrained on over 2.2 million hours of human activity to enable robots to learn manipulation tasks from human experience rather than robot-only demonstrations. The system combines large-scale human video, instrumented glove recordings capturing hand motions, and embodiment-aware learning to bridge the gap between watching humans perform tasks and executing those tasks with different robotic bodies.</p><p><strong>Key signal:</strong> LiberAI launches Liber-0, a foundation model trained on 2.2M hours of human video and glove-recorded motions to teach robots manipulation without extensive robot-only demos.</p><p>Source: Robotics and Automation News — From human experience to robot action: LiberAI launches Liber-0 world-action model (https://roboticsandautomationnews.com/2026/11/04/from-human-experience-to-robot-action-liberai-launches-liber-0-world-action-model/106378/)</p><p><a href="https://robotichub.ai/article/article_1793771852253">Read the full article on RoboticHub.ai</a></p>]]></description>
    </item>
    <item>
      <title>Leiden study confirms autonomous UV-C robot meets hospital disinfection standards</title>
      <link>https://robotichub.ai/article/article_1793211535855</link>
      <guid isPermaLink="true">https://robotichub.ai/article/article_1793211535855</guid>
      <pubDate>Wed, 28 Oct 2026 18:18:56 GMT</pubDate>
      <category>Research &amp; Signals</category>
      <category>Robotics</category>
      <category>Autonomous Systems</category>
      <category>Healthcare</category>
      <category>Breakthrough</category>
      <category>PoC</category>
      <description><![CDATA[<p><img src="https://roboticsandautomationnews.com/wp-content/uploads/2026/10/Ryberg-UV-C-disinfection-robot-1.jpg" alt="" /></p><p>A peer-reviewed study from Leiden University&apos;s Institute of Biology found that Ryberg Healthcare&apos;s autonomous UV-C disinfection robot met international disinfection standards, achieving bacterial reductions exceeding 7-log for E. coli and other pathogens. The research, published in Infection Prevention in Practice, demonstrated that autonomous operation outperformed static UV-C devices at the same duration and intensity, particularly in partially shadowed areas—a critical finding for real-world hospital environments where line-of-sight disinfection fails.</p><p><strong>Key signal:</strong> Leiden University study confirms Ryberg&apos;s autonomous UV-C robot achieves 7-log bacterial reduction, outperforming static systems in hospital disinfection tests.</p><p>Source: Robotics and Automation News — Leiden University study finds autonomous UV-C robot meets disinfection standards (https://roboticsandautomationnews.com/2026/10/28/leiden-university-study-finds-autonomous-uv-c-robot-meets-disinfection-standards/106410/)</p><p><a href="https://robotichub.ai/article/article_1793211535855">Read the full article on RoboticHub.ai</a></p>]]></description>
    </item>
    <item>
      <title>ETH Zurich trains robotic hand to crawl using its own fingers</title>
      <link>https://robotichub.ai/article/article_1792168313355</link>
      <guid isPermaLink="true">https://robotichub.ai/article/article_1792168313355</guid>
      <pubDate>Fri, 16 Oct 2026 16:31:53 GMT</pubDate>
      <category>Research &amp; Signals</category>
      <category>Robotics</category>
      <category>Edge AI</category>
      <category>Sensors</category>
      <category>PoC</category>
      <category>Breakthrough</category>
      <description><![CDATA[<p><img src="https://roboticsandautomationnews.com/wp-content/uploads/2026/10/wuji-hand-1.png" alt="" /></p><p>ETH Zurich researchers have demonstrated a robotic hand that uses its fingers for both locomotion and manipulation, eliminating the need for a separate arm or mobile base. The system, built using an off-the-shelf WUJI hand, can crawl untethered, steer, recover from falls, and perform manipulation tasks like typing on a keyboard—all while supporting its own weight. The work represents an unconventional approach to mobile manipulation that challenges the traditional robot arm paradigm.</p><p><strong>Key signal:</strong> ETH Zurich trains WUJI robotic hand to walk on its fingers, performing locomotion and manipulation without a separate mobility system.</p><p>Source: Robotics and Automation News — ETH Zurich scientists turn WUJI robotic hand into self-propelled mobile manipulator (https://roboticsandautomationnews.com/2026/10/16/eth-zurich-scientists-turn-wuji-robotic-hand-into-self-propelled-mobile-manipulator/105991/)</p><p><a href="https://robotichub.ai/article/article_1792168313355">Read the full article on RoboticHub.ai</a></p>]]></description>
    </item>
    <item>
      <title>Skild AI uses self-play to train robots beyond human capability</title>
      <link>https://robotichub.ai/article/article_1792168263532</link>
      <guid isPermaLink="true">https://robotichub.ai/article/article_1792168263532</guid>
      <pubDate>Fri, 16 Oct 2026 16:31:03 GMT</pubDate>
      <category>Research &amp; Signals</category>
      <category>Robotics</category>
      <category>Humanoids</category>
      <category>Edge AI</category>
      <category>PoC</category>
      <category>Breakthrough</category>
      <description><![CDATA[<p><img src="https://roboticsandautomationnews.com/wp-content/uploads/2026/10/skild-ai-model-4.png" alt="" /></p><p>Skild AI has demonstrated a self-play training method that allows robots to develop physical skills autonomously, using its S1 foundation model to learn soccer through over 140 years of simulated competition in Nvidia Isaac Sim. The approach could eliminate the bottleneck of human-generated training data and enable robots to exceed human capability in complex physical tasks including factory work, construction, and household manipulation.</p><p><strong>Key signal:</strong> Skild AI trains humanoid robot to play soccer using 140 years of simulated self-play, demonstrating autonomous skill acquisition beyond human demonstrations.</p><p>Source: Robotics and Automation News — Skild AI trains robots to learn physical skills through self-play (https://roboticsandautomationnews.com/2026/10/16/skild-ai-trains-robots-to-learn-physical-skills-through-self-play/105994/)</p><p><a href="https://robotichub.ai/article/article_1792168263532">Read the full article on RoboticHub.ai</a></p>]]></description>
    </item>
    <item>
      <title>Tesla patents scalable touch-sensitive skin for robot hands</title>
      <link>https://robotichub.ai/article/article_1792168228181</link>
      <guid isPermaLink="true">https://robotichub.ai/article/article_1792168228181</guid>
      <pubDate>Fri, 16 Oct 2026 16:30:28 GMT</pubDate>
      <category>Research &amp; Signals</category>
      <category>Robotics</category>
      <category>Humanoids</category>
      <category>Sensors</category>
      <category>PoC</category>
      <category>Strategic</category>
      <description><![CDATA[<p><img src="https://roboticsandautomationnews.com/wp-content/uploads/2026/10/tesla-robot-hand-1.png" alt="" /></p><p>Tesla has filed a US patent application for a touch-sensitive &quot;skin&quot; that can be molded onto robot fingers and palms, using flexible substrates with embedded conductive regions to detect contact, force, and pressure. The filing emphasizes manufacturability through established processes like screen printing and thermoforming, potentially enabling production at scale for humanoid manipulation tasks.</p><p><strong>Key signal:</strong> Tesla files patent for scalable touch-sensitive skin using printable sensors and thermoforming for robot hands.</p><p>Source: Robotics &amp; Automation News — Tesla patent reveals touch-sensitive skin for robot hands (https://roboticsandautomationnews.com/2026/10/16/tesla-patent-reveals-touch-sensitive-skin-for-robot-hands/105999/)</p><p><a href="https://robotichub.ai/article/article_1792168228181">Read the full article on RoboticHub.ai</a></p>]]></description>
    </item>
    <item>
      <title>MIT&apos;s SANDO drone planner guarantees collision avoidance in unknown terrain</title>
      <link>https://robotichub.ai/article/article_1791740370389</link>
      <guid isPermaLink="true">https://robotichub.ai/article/article_1791740370389</guid>
      <pubDate>Sun, 11 Oct 2026 17:39:30 GMT</pubDate>
      <category>Research &amp; Signals</category>
      <category>Drones</category>
      <category>Autonomous Systems</category>
      <category>Computer Vision</category>
      <category>PoC</category>
      <category>Breakthrough</category>
      <description><![CDATA[<p><img src="https://assets.newatlas.com/dims4/default/dfa4d3f/2147483647/strip/true/crop/1280x672+0+91/resize/1200x630!/quality/90/?url=https%3A%2F%2Fnewatlas-brightspot.s3.ap-southeast-2.amazonaws.com%2F0f%2F4e%2F2488e4d74a3ba0331f21a618a8bf%2Fplanning-system-ensure.jpg&amp;na.image_optimisation=0" alt="" /></p><p>MIT researchers developed SANDO, a drone navigation system that mathematically guarantees collision avoidance in disaster zones and unmapped environments. Rather than predicting obstacle paths, SANDO creates expanding virtual spheres around hazards based on maximum possible speed, then plans flight corridors through safe space—successfully avoiding all obstacles across simulations and 12 real-world test flights.</p><p><strong>Key signal:</strong> MIT develops SANDO drone planner with mathematical collision avoidance guarantee for disaster zones and unmapped terrain.</p><p>Source: New Atlas — MIT&apos;s drone brain neutralizes hazards in disaster zones (https://newatlas.com/drones/mit-sando-planning-obstacle-prediction-collisions/)</p><p><a href="https://robotichub.ai/article/article_1791740370389">Read the full article on RoboticHub.ai</a></p>]]></description>
    </item>
    <item>
      <title>Swiss researchers develop flexible fiber motor for soft robotics</title>
      <link>https://robotichub.ai/article/article_1791740343175</link>
      <guid isPermaLink="true">https://robotichub.ai/article/article_1791740343175</guid>
      <pubDate>Sun, 11 Oct 2026 17:39:03 GMT</pubDate>
      <category>Research &amp; Signals</category>
      <category>Robotics</category>
      <category>Sensors</category>
      <category>Humanoids</category>
      <category>Breakthrough</category>
      <category>PoC</category>
      <description><![CDATA[<p><img src="https://assets.newatlas.com/dims4/default/c933064/2147483647/strip/true/crop/6240x3276+0+114/resize/1200x630!/quality/90/?url=https%3A%2F%2Fnewatlas-brightspot.s3.ap-southeast-2.amazonaws.com%2F96%2F7d%2F0a1503e3423abeb286202efb8fd1%2F6240x3504-1.jpg&amp;na.image_optimisation=0" alt="" /></p><p>Researchers at Switzerland&apos;s EPFL have developed the FiberMotor, a thin, flexible electrostatic motor designed to actuate soft robots without the rigidity of conventional motors. The fiber-within-a-fiber design uses electrostatic forces to generate linear motion while remaining back-driveable, silent, and conformable to complex shapes. The breakthrough addresses a fundamental problem in soft robotics: how to add actuation without destroying compliance.</p><p><strong>Key signal:</strong> EPFL develops flexible fiber motor that actuates soft robots without adding rigidity — targets wearables, prosthetics, and haptic systems.</p><p>Source: New Atlas — Minuscule Swiss fiber motor keeps robots squishy and pliable (https://newatlas.com/robotics/fibermotor-flexible-motor/)</p><p><a href="https://robotichub.ai/article/article_1791740343175">Read the full article on RoboticHub.ai</a></p>]]></description>
    </item>
    <item>
      <title>MIT robot swims using living muscle cells controlled by light</title>
      <link>https://robotichub.ai/article/article_1791003881101</link>
      <guid isPermaLink="true">https://robotichub.ai/article/article_1791003881101</guid>
      <pubDate>Sat, 03 Oct 2026 05:04:41 GMT</pubDate>
      <category>Research &amp; Signals</category>
      <category>Robotics</category>
      <category>Sensors</category>
      <category>PoC</category>
      <category>Breakthrough</category>
      <category>Emerging</category>
      <description><![CDATA[<p><img src="https://assets.newatlas.com/dims4/default/ae18159/2147483647/strip/true/crop/1800x945+0+127/resize/1200x630!/quality/90/?url=https%3A%2F%2Fnewatlas-brightspot.s3.ap-southeast-2.amazonaws.com%2F8d%2F0a%2F2c4e524944869787ad1acf136f17%2Fmit-robomuscle-01-press-0.jpg&amp;na.image_optimisation=0" alt="" /></p><p>MIT researchers built a chewing gum-sized robot powered by a single layer of living muscle cells that contracts when exposed to light. The thin construction uses fewer cells than competing biohybrid designs and demonstrates basic navigation through a water maze, though it remains lab-bound and requires external steering.</p><p><strong>Key signal:</strong> MIT demonstrates light-steered swimming robot powered by single-layer living muscle cells — thinner and cheaper than previous biohybrid designs.</p><p>Source: New Atlas — MIT robot swims through water via living muscle cells (https://newatlas.com/robotics/mit-robot-swims-light-living-muscle-cells/)</p><p><a href="https://robotichub.ai/article/article_1791003881101">Read the full article on RoboticHub.ai</a></p>]]></description>
    </item>
    <item>
      <title>Cyborg cockroaches deliver drug injections in search-and-rescue tests</title>
      <link>https://robotichub.ai/article/article_1788619183573</link>
      <guid isPermaLink="true">https://robotichub.ai/article/article_1788619183573</guid>
      <pubDate>Sat, 05 Sep 2026 14:39:43 GMT</pubDate>
      <category>Research &amp; Signals</category>
      <category>Robotics</category>
      <category>Sensors</category>
      <category>PoC</category>
      <category>Emerging</category>
      <category>Global</category>
      <description><![CDATA[<p><img src="https://spectrum.ieee.org/media-library/large-brown-cockroach-fitted-with-an-electronic-device-and-wires-on-its-back.jpg?id=67724302&amp;width=1200&amp;height=600&amp;coordinates=0%2C125%2C0%2C125" alt="" /></p><p>Australian researchers have developed &quot;Paraborgs&quot; — cyborg cockroaches fitted with wireless controls and spring-loaded drug injectors — designed to navigate disaster sites and deliver emergency medication to trapped victims. In proof-of-concept tests, the system achieved 72% injection accuracy over a 2.5-meter obstacle course, demonstrating feasibility but raising questions about reliability, sterility, and victim acceptance in real-world rescue scenarios.</p><p><strong>Key signal:</strong> Australian researchers test cyborg cockroaches with remote-controlled drug injectors for search-and-rescue, achieving 72% injection accuracy in lab trials.</p><p>Source: IEEE Spectrum — Cyborg Roaches Can Stab You With Needles (https://spectrum.ieee.org/cyborg-rescue-roach)</p><p><a href="https://robotichub.ai/article/article_1788619183573">Read the full article on RoboticHub.ai</a></p>]]></description>
    </item>
    <item>
      <title>Autonomous vehicles learn to take passenger requests via LLM interface</title>
      <link>https://robotichub.ai/article/article_1788120187291</link>
      <guid isPermaLink="true">https://robotichub.ai/article/article_1788120187291</guid>
      <pubDate>Sun, 30 Aug 2026 20:03:24 GMT</pubDate>
      <category>Research &amp; Signals</category>
      <category>Autonomous Systems</category>
      <category>Computer Vision</category>
      <category>Edge AI</category>
      <category>PoC</category>
      <category>Breakthrough</category>
      <description><![CDATA[<p><img src="https://spectrum.ieee.org/media-library/jw-your-self-driving-car-now-takes-requests.jpg?id=67659079&amp;width=1200&amp;height=600&amp;coordinates=0%2C125%2C0%2C125" alt="" /></p><p>Researchers at TU Delft have developed a system that uses an LLM to translate passenger requests like &quot;go faster, I&apos;m late&quot; into parameter adjustments for a self-driving car&apos;s motion planner. The approach keeps the AI away from direct vehicle control, instead tuning criteria like speed and smoothness while requiring human confirmation before making changes — offering personalization without compromising the safety guarantees of deterministic planning algorithms.</p><p><strong>Key signal:</strong> TU Delft uses GPT-4o-mini to let passengers adjust autonomous vehicle driving style via natural language while maintaining deterministic safety bounds.</p><p>Source: IEEE Spectrum — Self-Driving Cars Could Someday Take Requests (https://spectrum.ieee.org/autonomous-vehicles-motion-planner-llm)</p><p><a href="https://robotichub.ai/article/article_1788120187291">Read the full article on RoboticHub.ai</a></p>]]></description>
    </item>
    <item>
      <title>VicOne releases free NVIDIA Isaac Sim extension for robot cybersecurity testing</title>
      <link>https://robotichub.ai/article/article_1786444726361</link>
      <guid isPermaLink="true">https://robotichub.ai/article/article_1786444726361</guid>
      <pubDate>Tue, 11 Aug 2026 10:38:56 GMT</pubDate>
      <category>Research &amp; Signals</category>
      <category>Robotics</category>
      <category>Computer Vision</category>
      <category>Edge AI</category>
      <category>PoC</category>
      <category>Vulnerability</category>
      <description><![CDATA[<p><img src="https://www.therobotreport.com/wp-content/uploads/2026/08/VicOne_DefCon.jpg" alt="" /></p><p>VicOne has released a free NVIDIA Isaac Sim extension that allows developers to test how cyberattacks could alter robot behavior before deployment. The tool is based on findings from the company&apos;s DEF CON 34 capture-the-flag exercise, where researchers tested attack paths across AI models, perception systems, and embedded controls. It addresses a gap in robotics testing: functional safety is validated extensively, but cyber-induced failure modes often fall outside standard test plans.</p><p><strong>Key signal:</strong> VicOne releases free NVIDIA Isaac Sim extension for testing robot cyberattack scenarios based on DEF CON 34 research.</p><p>Source: The Robot Report — VicOne releases free NVIDIA Isaac Sim cybersecurity extension based on DEF CON 34 research (https://www.therobotreport.com/vicone-releases-free-nviida-isaac-sim-cybersecurity-extension-based-def-con-34-research/)</p><p><a href="https://robotichub.ai/article/article_1786444726361">Read the full article on RoboticHub.ai</a></p>]]></description>
    </item>
    <item>
      <title>Shark robot vacuums exposed to remote takeover via AWS cloud flaw</title>
      <link>https://robotichub.ai/article/article_1786439523457</link>
      <guid isPermaLink="true">https://robotichub.ai/article/article_1786439523457</guid>
      <pubDate>Tue, 11 Aug 2026 09:12:32 GMT</pubDate>
      <category>Research &amp; Signals</category>
      <category>Robotics</category>
      <category>Sensors</category>
      <category>Computer Vision</category>
      <category>Vulnerability</category>
      <category>High Risk</category>
      <description><![CDATA[<p><img src="https://vacuumwars.com/wp-content/uploads/2026/07/Shark-vulnerability-title.png" alt="" /></p><p>Security researcher tokay0 disclosed a cloud vulnerability affecting Shark robot vacuums that could have allowed remote command execution, camera access, and Wi-Fi credential theft via an overly permissive AWS IoT policy. SharkNinja confirmed post-publication that it has &quot;completely addressed the identified vulnerability,&quot; though no CVE has been assigned and there is no evidence the flaw was exploited in the wild.</p><p><strong>Key signal:</strong> Shark robot vacuums vulnerable to remote takeover via AWS cloud flaw; SharkNinja says issue resolved, no evidence of exploitation.</p><p>Source: Vacuum Wars — Shark Robot Vacuum Security Vulnerability Report from Researcher (https://vacuumwars.com/shark-robot-vacuum-security-vulnerability/)</p><p><a href="https://robotichub.ai/article/article_1786439523457">Read the full article on RoboticHub.ai</a></p>]]></description>
    </item>
    <item>
      <title>Three critical flaws expose Ghost Robotics Vision 60 to remote takeover</title>
      <link>https://robotichub.ai/article/article_1786439041643</link>
      <guid isPermaLink="true">https://robotichub.ai/article/article_1786439041643</guid>
      <pubDate>Tue, 11 Aug 2026 09:06:10 GMT</pubDate>
      <category>Research &amp; Signals</category>
      <category>Robotics</category>
      <category>Defense</category>
      <category>Vulnerability</category>
      <category>High Risk</category>
      <category>Strategic</category>
      <description><![CDATA[<p><img src="https://www.incibe.es/sites/default/files/contenidos/blog/estar-al-dia-las-amenazas-ciberseguridad-nuestro-servicio-avisos/incibecertavisos-800.jpg" alt="" /></p><p>INCIBE has disclosed three high-severity vulnerabilities in Ghost Robotics&apos; Vision 60 quadruped robot, with CVSS scores up to 8.7. The flaws — lack of authentication, weak session control, and missing cryptographic integrity — enable attackers on the robot&apos;s Wi-Fi network to seize control, intercept camera feeds, and execute movement commands while blocking legitimate operators. No fix is available.</p><p><strong>Key signal:</strong> Three unpatched CVEs allow attackers on Ghost Robotics Vision 60&apos;s Wi-Fi to hijack control, intercept feeds, and block operators.</p><p>Source: INCIBE-CERT — Multiple vulnerabilities in Ghost Robotics&apos; Vision 60 (https://www.incibe.es/en/incibe-cert/notices/aviso-sci/multiple-vulnerabilities-ghost-robotics-vision-60)</p><p><a href="https://robotichub.ai/article/article_1786439041643">Read the full article on RoboticHub.ai</a></p>]]></description>
    </item>
    <item>
      <title>Ubiquiti patches critical UniFi OS command injection flaw</title>
      <link>https://robotichub.ai/article/article_1783586826153</link>
      <guid isPermaLink="true">https://robotichub.ai/article/article_1783586826153</guid>
      <pubDate>Thu, 09 Jul 2026 08:47:22 GMT</pubDate>
      <category>Research &amp; Signals</category>
      <category>Smart Cities</category>
      <category>Industrial Automation</category>
      <category>Edge AI</category>
      <category>Vulnerability</category>
      <category>High Risk</category>
      <description><![CDATA[<p><img src="https://securityaffairs.com/wp-content/uploads/2021/01/Ubiquiti-Logo.png" alt="" /></p><p>Ubiquiti patched seven UniFi OS vulnerabilities, including CVE-2026-50746 (CVSS 10.0), a critical command injection flaw in the UniFi Connect Application used to manage commercial building systems. Network-positioned attackers could exploit the vulnerability to execute arbitrary commands on affected devices running versions 3.4.16 and earlier.</p><p><strong>Key signal:</strong> Ubiquiti patches critical command injection flaw in UniFi OS smart building platform — CVSS 10.0, affects EV chargers and lighting controls.</p><p>Source: Security Affairs — Ubiquiti Patches Critical UniFi OS Flaws Allowing Command Injection and Privilege Escalation (https://securityaffairs.com/194978/security/ubiquiti-patches-critical-unifi-os-flaws-allowing-command-injection-and-privilege-escalation.html)</p><p><a href="https://robotichub.ai/article/article_1783586826153">Read the full article on RoboticHub.ai</a></p>]]></description>
    </item>
    <item>
      <title>500,000 hours of Ukraine drone footage released for AI training</title>
      <link>https://robotichub.ai/article/article_1782061347019</link>
      <guid isPermaLink="true">https://robotichub.ai/article/article_1782061347019</guid>
      <pubDate>Sun, 21 Jun 2026 17:02:43 GMT</pubDate>
      <category>Research &amp; Signals</category>
      <category>Drones</category>
      <category>Computer Vision</category>
      <category>Defense</category>
      <category>PoC</category>
      <category>Strategic</category>
      <description><![CDATA[<p><img src="https://defensescoop.com/wp-content/uploads/sites/8/2026/06/EI-Ukrainian-FMV-Thermal-Example-1.png" alt="" /></p><p>Enabled Intelligence has released a collection of over 500,000 hours of labeled Ukraine conflict drone footage for AI model training, covering electro-optical, infrared, and SAR sensors. The dataset provides real-world combat conditions — weather variability, terrain diversity, dynamic battlefield scenarios — that simulated training data cannot replicate, accelerating development of autonomous targeting and ISR systems.</p><p><strong>Key signal:</strong> 500,000 hours of labeled Ukraine drone footage released for AI training by Enabled Intelligence — available to US, NATO, and Ukrainian developers.</p><p>Source: Defense Scoop — Data from &apos;half a million hours of Ukraine conflict drone footage&apos; now available to train AI (https://defensescoop.com/2026/06/16/data-from-half-a-million-hours-of-ukraine-conflict-drone-footage-now-available-to-train-ai/)</p><p><a href="https://robotichub.ai/article/article_1782061347019">Read the full article on RoboticHub.ai</a></p>]]></description>
    </item>
    <item>
      <title>Vision language models teach robots to read emotions — but trust still hinges on performance</title>
      <link>https://robotichub.ai/article/article_1781367586047</link>
      <guid isPermaLink="true">https://robotichub.ai/article/article_1781367586047</guid>
      <pubDate>Sat, 13 Jun 2026 16:19:56 GMT</pubDate>
      <category>Research &amp; Signals</category>
      <category>Robotics</category>
      <category>Computer Vision</category>
      <category>PoC</category>
      <category>Research &amp; Signals</category>
      <category>Emerging</category>
      <description><![CDATA[<p><img src="https://spectrum.ieee.org/media-library/image.jpg?id=66888132&amp;width=1200&amp;height=600&amp;coordinates=0%2C224%2C0%2C276" alt="" /></p><p>Researchers at the University of Melbourne trained collaborative robots to interpret human emotions using vision language models that analyze full-scene context, not just facial expressions. The VLM outperformed conventional emotion recognition systems, and volunteers preferred emotionally adaptive robot apologies over scripted ones — but trust collapsed when robots failed their core tasks, regardless of how they apologized.</p><p><strong>Key signal:</strong> Vision language models improve robot emotion recognition, but functional competence — not apology quality — determines human trust in collaborative robots.</p><p>Source: IEEE Spectrum — Are Emotion Reading Robots Still Missing What Matters Most? (https://spectrum.ieee.org/robot-emotions-visual-language-models)</p><p><a href="https://robotichub.ai/article/article_1781367586047">Read the full article on RoboticHub.ai</a></p>]]></description>
    </item>
    <item>
      <title>Software architecture now biggest bottleneck for robotics teams, QNX survey finds</title>
      <link>https://robotichub.ai/article/article_1780232216489</link>
      <guid isPermaLink="true">https://robotichub.ai/article/article_1780232216489</guid>
      <pubDate>Sun, 31 May 2026 12:57:13 GMT</pubDate>
      <category>Research &amp; Signals</category>
      <category>Robotics</category>
      <category>Industrial Automation</category>
      <category>PoC</category>
      <category>Strategic</category>
      <category>Emerging</category>
      <description><![CDATA[<p><img src="https://www.therobotreport.com/wp-content/uploads/2026/05/QNX-Robotics-Architecture-Report-copy.jpg" alt="" /></p><p>A QNX-commissioned survey of 1,000 robotics developers found that software architecture and integration (27%) now outpaces hardware (16%) as the primary development bottleneck. The shift reflects industry-wide recognition that future progress depends less on new sensors or actuators and more on building predictable, safety-certifiable software foundations as robots move from controlled factories into dynamic human environments.</p><p><strong>Key signal:</strong> QNX survey finds 27% of robotics developers now cite software architecture as biggest bottleneck — surpassing hardware constraints for first time.</p><p>Source: The Robot Report — Software becoming the biggest bottleneck to physical AI innovation, finds QNX research (https://www.therobotreport.com/software-becoming-biggest-bottleneck-physical-ai-innovation-finds-qnx/)</p><p><a href="https://robotichub.ai/article/article_1780232216489">Read the full article on RoboticHub.ai</a></p>]]></description>
    </item>
    <item>
      <title>NIST launches humanoid robot baseline benchmark for industry</title>
      <link>https://robotichub.ai/article/article_1780231995464</link>
      <guid isPermaLink="true">https://robotichub.ai/article/article_1780231995464</guid>
      <pubDate>Sun, 31 May 2026 12:53:36 GMT</pubDate>
      <category>Research &amp; Signals</category>
      <category>Humanoids</category>
      <category>Robotics</category>
      <category>PoC</category>
      <category>Strategic</category>
      <category>US</category>
      <description><![CDATA[<p><img src="https://www.nist.gov/themes/custom/nist_www/img/homepage/nist_mark.png" alt="" /></p><p>The National Institute of Standards and Technology (NIST) has announced a Humanoid Robot Baseline Performance Benchmark — a standardized suite of locomotion and manipulation tasks designed to assess minimum expected capabilities across commercial humanoid platforms. The initiative addresses a critical gap: no rigorous cross-platform performance measurement has existed since the 2014 DARPA Robotics Challenge. NIST will distribute test apparatus to U.S. manufacturers and regional facilities under data-sharing agreements that protect IP while establishing industry-wide capability baselines.</p><p><strong>Key signal:</strong> NIST launches standardized humanoid robot benchmark to establish comparable performance metrics across commercial platforms — first rigorous cross-industry evaluation since 2014.</p><p>Source: NIST — Humanoid Robot Baseline Performance Benchmark (https://www.nist.gov/el/intelligent-systems-division-73500/humanoid-robot-baseline-performance-benchmark)</p><p><a href="https://robotichub.ai/article/article_1780231995464">Read the full article on RoboticHub.ai</a></p>]]></description>
    </item>
    <item>
      <title>Yarbo robotic lawnmowers contain hardcoded backdoor and bypass emergency stop</title>
      <link>https://robotichub.ai/article/article_1779732454006</link>
      <guid isPermaLink="true">https://robotichub.ai/article/article_1779732454006</guid>
      <pubDate>Mon, 25 May 2026 18:09:38 GMT</pubDate>
      <category>Research &amp; Signals</category>
      <category>Robotics</category>
      <category>Vulnerability</category>
      <category>High Risk</category>
      <category>PoC</category>
      <category>Smart Cities</category>
      <description><![CDATA[<p>Security researcher Andreas Makris discovered critical vulnerabilities in Yarbo&apos;s modular robotic lawnmowers and snow blowers that allow remote takeover, emergency stop bypass, and extraction of owner data including Wi-Fi passwords and GPS coordinates. All 11,000+ deployed devices share a hardcoded root password and a built-in remote access backdoor that cannot be disabled, enabling attackers to control movement, activate blades, or pivot into home networks.</p><p><strong>Key signal:</strong> Researcher discloses hardcoded backdoor in 11,000+ Yarbo robotic lawnmowers enabling remote takeover and emergency stop bypass.</p><p>Source: HackMag — Researcher finds backdoor and multiple vulnerabilities in robotic lawnmowers (https://hackmag.com/news/yarbo-flaws)</p><p><a href="https://robotichub.ai/article/article_1779732454006">Read the full article on RoboticHub.ai</a></p>]]></description>
    </item>
    <item>
      <title>Universal Robots patches critical command injection flaw in cobot OS</title>
      <link>https://robotichub.ai/article/article_1779201234111</link>
      <guid isPermaLink="true">https://robotichub.ai/article/article_1779201234111</guid>
      <pubDate>Tue, 19 May 2026 14:34:15 GMT</pubDate>
      <category>Research &amp; Signals</category>
      <category>Robotics</category>
      <category>Industrial Automation</category>
      <category>Vulnerability</category>
      <category>High Risk</category>
      <category>Manufacturing</category>
      <description><![CDATA[<p><img src="https://www.securityweek.com/wp-content/uploads/2026/05/cobot-industrial-robot-arm.jpeg" alt="" /></p><p>Universal Robots has patched CVE-2026-8153, a critical command injection vulnerability in its PolyScope 5 operating system that powers collaborative industrial robots. The flaw allows unauthenticated attackers with network access to execute arbitrary commands on the underlying Linux OS, potentially compromising entire robot fleets in flat industrial networks.</p><p><strong>Key signal:</strong> Universal Robots patches CVE-2026-8153, a critical command injection flaw in PolyScope 5 that threatens industrial cobot fleets.</p><p>Source: SecurityWeek — Critical Vulnerability Exposes Industrial Robot Fleets to Hacking (https://www.securityweek.com/critical-vulnerability-exposes-industrial-robot-fleets-to-hacking/)</p><p><a href="https://robotichub.ai/article/article_1779201234111">Read the full article on RoboticHub.ai</a></p>]]></description>
    </item>
    <item>
      <title>Physical Intelligence trains robots to master precision tasks in 15 minutes</title>
      <link>https://robotichub.ai/article/article_1778956971663</link>
      <guid isPermaLink="true">https://robotichub.ai/article/article_1778956971663</guid>
      <pubDate>Sat, 16 May 2026 18:44:21 GMT</pubDate>
      <category>Research &amp; Signals</category>
      <category>Robotics</category>
      <category>Computer Vision</category>
      <category>Edge AI</category>
      <category>Breakthrough</category>
      <category>PoC</category>
      <description><![CDATA[<p><img src="https://physicalintelligence.company/images/rlt/og.png" alt="" /></p><p>Physical Intelligence has developed RL tokens (RLT), a reinforcement learning method that allows vision-language-action models to refine precise manipulation skills using just 15 minutes of real-world robot experience. The technique adds a compressed &quot;RL token&quot; output to the base model, enabling a lightweight policy to train in real time without retraining the full VLA. Across four tasks—driving screws, fastening zip ties, and inserting cables—RLT improved execution speed up to 3× and outpaced human teleoperation.</p><p><strong>Key signal:</strong> Physical Intelligence&apos;s RL tokens method trains robots to master precision tasks in 15 minutes, achieving 3× speed gains over base models and surpassing human teleoperation.</p><p>Source: Physical Intelligence — Precise Manipulation with Efficient Online RL (https://www.pi.website/research/rlt)</p><p><a href="https://robotichub.ai/article/article_1778956971663">Read the full article on RoboticHub.ai</a></p>]]></description>
    </item>
    <item>
      <title>Unitree robots vulnerable to firmware poisoning via hard-coded encryption keys</title>
      <link>https://robotichub.ai/article/article_1778956533145</link>
      <guid isPermaLink="true">https://robotichub.ai/article/article_1778956533145</guid>
      <pubDate>Sat, 16 May 2026 18:37:44 GMT</pubDate>
      <category>Research &amp; Signals</category>
      <category>Robotics</category>
      <category>Defense</category>
      <category>Vulnerability</category>
      <category>High Risk</category>
      <category>PoC</category>
      <description><![CDATA[<p><img src="http://takeonme.org/og-image.png" alt="" /></p><p>Security researchers have disclosed CVE-2026-1442, a critical vulnerability in Unitree&apos;s firmware update system that uses hard-coded encryption keys. The flaw allows anyone with physical or remote access to create and deploy malicious firmware packages that Unitree robots — including models deployed by police and military units in the US and Europe — will trust and execute. Public exploit tooling (UniTEABag) is now available, and Unitree has not responded to disclosure attempts.</p><p><strong>Key signal:</strong> Unitree robots vulnerable to firmware poisoning via hard-coded keys; public exploit available, no patch from vendor.</p><p>Source: TakeOnMe.org — CVE-2026-1442 (https://takeonme.org/cves/cve-2026-1442/)</p><p><a href="https://robotichub.ai/article/article_1778956533145">Read the full article on RoboticHub.ai</a></p>]]></description>
    </item>
    <item>
      <title>Researcher breaks Tegra X2 secure boot, exposing Tesla Autopilot hardware</title>
      <link>https://robotichub.ai/article/article_1778956256445</link>
      <guid isPermaLink="true">https://robotichub.ai/article/article_1778956256445</guid>
      <pubDate>Sat, 16 May 2026 18:31:09 GMT</pubDate>
      <category>Research &amp; Signals</category>
      <category>Autonomous Systems</category>
      <category>Computer Vision</category>
      <category>Vulnerability</category>
      <category>High Risk</category>
      <category>PoC</category>
      <description><![CDATA[<p><img src="https://static.media.ccc.de/media/congress/2025/1454-b98918cb-489e-5f5e-aa06-26753cb48418_preview.jpg" alt="" /></p><p>Security researcher Elise Amber Katze disclosed multiple vulnerabilities in NVIDIA&apos;s Tegra X2 SoC, including an unpatchable BootROM exploit affecting Tesla Autopilot 2 and 2.5 hardware. The research, presented at 39C3, demonstrates complete secure boot chain bypass via USB recovery mode. Since the vulnerability exists in read-only BootROM, no firmware update can fix it in deployed systems.</p><p><strong>Key signal:</strong> Unpatchable BootROM vulnerability disclosed in NVIDIA Tegra X2, affecting Tesla Autopilot 2/2.5 hardware with no firmware fix possible.</p><p>Source: Chaos Communication Congress (39C3) — Making the Magic Leap past NVIDIA&apos;s secure bootchain and breaking some Tesla Autopilots along the way (https://media.ccc.de/v/39c3-making-the-magic-leap-past-nvidia-s-secure-bootchain-and-breaking-some-tesla-autopilots-along-the-way)</p><p><a href="https://robotichub.ai/article/article_1778956256445">Read the full article on RoboticHub.ai</a></p>]]></description>
    </item>
    <item>
      <title>Engineer accesses 7,000 robot vacuums via DJI backend flaw</title>
      <link>https://robotichub.ai/article/article_1778956176242</link>
      <guid isPermaLink="true">https://robotichub.ai/article/article_1778956176242</guid>
      <pubDate>Sat, 16 May 2026 18:29:43 GMT</pubDate>
      <category>Research &amp; Signals</category>
      <category>Robotics</category>
      <category>Sensors</category>
      <category>Computer Vision</category>
      <category>Vulnerability</category>
      <category>High Risk</category>
      <description><![CDATA[<p><img src="https://i.guim.co.uk/img/media/e1930fc7978d19ba529fa4a85675d89415632585/144_22_4775_3820/master/4775.jpg?width=1200&amp;height=630&amp;quality=85&amp;auto=format&amp;fit=crop&amp;precrop=40:21,offset-x50,offset-y0&amp;overlay-align=bottom%2Cleft&amp;overlay-width=100p&amp;overlay-base64=L2ltZy9zdGF0aWMvb3ZlcmxheXMvdGctZGVmYXVsdC5wbmc&amp;enable=upscale&amp;s=d57f439b405ea129598f9d891f9fd482" alt="" /></p><p>Software engineer Sammy Azdoufal reverse-engineered his DJI Romo vacuum cleaner and discovered a backend vulnerability granting access to live camera feeds, microphone audio, and floor plans from nearly 7,000 devices globally. The disclosure highlights systemic security weaknesses in consumer IoT robotics, where backend infrastructure remains a critical attack surface even when devices themselves appear secure.</p><p><strong>Key signal:</strong> Engineer discovers DJI backend flaw granting access to 7,000 robot vacuums&apos; live cameras, audio, and floor plans across 24 countries.</p><p>Source: The Guardian — The accidental hacker: how one man gained control of 7,000 robots (https://www.theguardian.com/lifeandstyle/2026/feb/24/accidental-hacker-how-one-man-gained-control-robots-)</p><p><a href="https://robotichub.ai/article/article_1778956176242">Read the full article on RoboticHub.ai</a></p>]]></description>
    </item>
    <item>
      <title>Military drone lessons expose self-driving car safety gaps</title>
      <link>https://robotichub.ai/article/article_1778955813973</link>
      <guid isPermaLink="true">https://robotichub.ai/article/article_1778955813973</guid>
      <pubDate>Sat, 16 May 2026 18:24:04 GMT</pubDate>
      <category>Research &amp; Signals</category>
      <category>Autonomous Systems</category>
      <category>Drones</category>
      <category>High Risk</category>
      <category>Strategic</category>
      <category>Research</category>
      <description><![CDATA[<p><img src="https://spectrum.ieee.org/media-library/image.jpg?id=65098234&amp;width=1200&amp;height=600&amp;coordinates=0%2C312%2C0%2C313" alt="" /></p><p>Self-driving car companies are repeating mistakes the U.S. military solved decades ago in unmanned aerial vehicle operations, according to autonomous systems researcher Missy Cummings. Remote supervision of autonomous vehicles requires low latency, careful interface design, operator workload management, rigorous training, and contingency planning — yet commercial deployments show gaps in all five areas, with some operators now based in the Philippines where network delays amplify control risks.</p><p><strong>Key signal:</strong> Military drone research from the 1990s shows self-driving car companies are repeating critical remote supervision mistakes that caused UAV accidents.</p><p>Source: IEEE Spectrum — What Military Drones Can Teach Self-Driving Cars (https://spectrum.ieee.org/military-drones-self-driving-cars)</p><p><a href="https://robotichub.ai/article/article_1778955813973">Read the full article on RoboticHub.ai</a></p>]]></description>
    </item>
    <item>
      <title>Octopus-inspired gripper shifts from soft to rigid in under 2 seconds</title>
      <link>https://robotichub.ai/article/article_1778942420345</link>
      <guid isPermaLink="true">https://robotichub.ai/article/article_1778942420345</guid>
      <pubDate>Sat, 16 May 2026 14:40:28 GMT</pubDate>
      <category>Research &amp; Signals</category>
      <category>Robotics</category>
      <category>Sensors</category>
      <category>PoC</category>
      <category>Breakthrough</category>
      <category>Research &amp; Signals</category>
      <description><![CDATA[<p><img src="https://assets.newatlas.com/dims4/default/3d60b35/2147483647/strip/true/crop/1662x873+0+118/resize/1200x630!/quality/85/?url=https%3A%2F%2Fnewatlas-brightspot.s3.amazonaws.com%2F9f%2F13%2F959fc23c4713bebfec0c4a3c6f19%2F%E5%9B%BE%E7%89%871-1.jpeg&amp;na.image_optimisation=0" alt="" /></p><p>Researchers at Peking University have developed an octopus-inspired underwater gripper that transitions between soft and rigid states in 1.3 seconds — 10x faster than existing shape memory polymer systems. The OUT-Robot uses electrically heated PLA arms that grip fragile or irregular objects when soft, then lock rigid for zero-energy vertical lift via buoyancy control, carrying payloads over 500 grams.</p><p><strong>Key signal:</strong> Octopus-inspired gripper shifts soft-to-rigid in under 2 seconds using heated PLA, enabling zero-energy underwater lift for payloads over 500 grams.</p><p>Source: New Atlas — Heated PLA robo-octopus gripper goes limp to grapple, hard to lift (https://newatlas.com/robotics/octopus-inspired-gripper/)</p><p><a href="https://robotichub.ai/article/article_1778942420345">Read the full article on RoboticHub.ai</a></p>]]></description>
    </item>
  </channel>
</rss>
