50 articles found
CMU researchers developed human-inspired robotic fingertips that combine force and vibrotactile sensing to enable robots to handle delicate objects like pills and potato chips. The low-cost fingertips (under $100 each) attach to the DeltaHands robotic hand and use a strain gauge and contact microphone to detect pressure, texture, and slipping, enabling fine manipulation tasks previously difficult for robots.
Carnegie Mellon University researchers have developed LAMP (Long-Horizon Adaptive Manipulation Planning), a system that enables multiple robots to collaborate and coordinate their movements to transport objects through cluttered, crowded spaces like warehouses. The system combines learned models of object manipulation with search-based planning algorithms, and includes a lazy variant that updates plans dynamically based on real-world feedback, successfully completing complex multi-step tasks that previous approaches could not handle.
IROS 2026, the IEEE/RSJ International Conference on Intelligent Robots and Systems, will be held September 27 - October 1 in Pittsburgh, featuring keynote talks from leading researchers covering topics including autonomous surgery, soft robotics, embodied AI, manipulation systems, and disaster robotics. The conference program includes 86 workshops and tutorials, plenary panels on startups, space robotics, and generalist vs. specialist robots, and community forums on government funding, robotics education, and foundational research.
The 2026 IEEE International Conference on Robotics and Automation (ICRA) held in Vienna from June 1-5 has made keynote and plenary talk recordings publicly available. The event featured talks on topics including aerial robots, humanoid robotics, robot perception, manipulation, learning, human-robot interaction, and autonomous vehicles, along with panel discussions on humanoid robotics startups, sustainable robotics, and trustworthy AI for automation.
Two major security breaches in as many weeks have exposed a troubling truth: the safety guardrails protecting enterprise AI systems are alarmingly fragile. When both Grok and Microsoft Copilot can be tricked into revealing their own vulnerabilities through simple prompt manipulation, we need to ask whether AI security is being treated seriously or as a checkbox exercise.
Google's new sign language translation AI isn't just another accessibility feature — it's a reminder that the robotics and AI community has spent years solving the wrong problems. While we've obsessed over humanoid form factors and general-purpose manipulation, some of the most impactful applications require us to actually listen to disabled communities first.
Tacta Systems launched TactaBot, a dexterous robotic hand system featuring proprietary fluidic tendon actuation, high-resolution tactile sensors, and an AI model designed for high-skilled manufacturing tasks. The company also introduced a data-capture glove that enables robots to learn complex manipulation skills by recording human demonstrations, addressing the critical data shortage in physical AI training.
MIT researchers have developed SceneSmith, a system that uses three AI agents (a designer, critic, and orchestrator) to automatically generate realistic 3D virtual environments for robot training and simulation. The system leverages vision-language models to create detailed indoor scenes with up to six times more objects than previous methods, enabling robots to practice manipulation tasks in diverse, lifelike settings before real-world deployment.
This roundup highlights the top 10 robotics stories from July 2026, including major funding rounds for Chinese humanoid robot companies, advances in reinforcement learning for manipulation, new humanoid robot launches (Apollo 2, Isaac 1), AMD's robotics platform, brain-computer interface demonstrations, and new U.S. import policies affecting humanoid and mobile robots.
Holiday Robotics, a Seoul-based startup, raised $105 million in Series A funding to develop FRIDAY, a wheeled humanoid robot designed for industrial manufacturing tasks. FRIDAY features 64 degrees of freedom with 40 dedicated to dexterous hands, safety-focused hardware design, and a vision-language-skill (VLS) approach to AI that combines perception, learning, and real-world deployment for reliable operation alongside human workers.
While the industry races toward humanoid deployment, a quieter revolution is happening in how robots understand physical contact. From full-body tactile sensing to friction-aware world models, researchers are finally addressing the gap that's kept manipulation tasks frustratingly brittle.
Researchers propose VμA, a new world model class for robotics that incorporates friction (coefficient of static friction μ) as a first-class input to improve robot learning and manipulation. Current robot world models rely on visual and joint position data but lack direct tactile information about contact physics, limiting their ability to generalize to novel surfaces and objects; the new approach addresses this by making friction a primary conditioning signal for more reliable physical predictions.
This newsletter covers multiple AI and technology stories, with key focus on research findings that LLMs develop hiring biases beyond their training data and may stereotype job applicants more than humans, and a separate piece on the growing risk of weather data sabotage as prediction markets and AI weather forecasting create incentives for data manipulation.
Researchers have developed an Interactive World Simulator, a learned video prediction model that can simulate robot manipulation tasks in pixel space without a physics engine, enabling cheap data generation and scalable policy evaluation. The system is trained on an autoencoder and action-conditioned dynamics model, and has been demonstrated on four manipulation tasks including rope routing, mug grasping, and pile sweeping, allowing extended interactive simulation at 15 FPS.
Amazon and University of Michigan researchers developed HydroShear, a physics-based simulator that teaches robots to sense and understand tactile forces for complex manipulation tasks. The simulator enables robots to learn dexterous manipulation policies entirely in simulation and transfer them to real robots with 93% success rate, addressing a major challenge in contact-rich robotic tasks.
Boston Dynamics' Atlas humanoid robot demonstrated advanced soccer skills at the FIFA World Cup, including the complex Ghost Rabona kick, as part of Hyundai's expanded partnership with FIFA. The company discussed how reinforcement learning and motion capture techniques enabled Atlas to learn complex object manipulation and athletic behaviors more efficiently than traditional model-based control methods.
Humanoid robotics companies are racing to demonstrate "human-level" performance, with startups like Humanoid touting 99.9% manipulation reliability. But in high-volume industrial settings, that remaining 0.1% represents thousands of failures per day—and reveals why the industry's obsession with matching human capability misses the point entirely.
Humanoid, a London-based robotics startup, introduced KinetIQ Ascend, a reinforcement learning framework designed to enable humanoid robots to achieve 99.9% manipulation reliability at human speed and beyond. The company demonstrated the system across multiple industrial tasks including bin picking and bimanual handling, achieving throughput improvements of 42-85% and success rates exceeding 98% after only days of training, marking a shift from manual tuning to scalable, real-world deployment.
This roundup covers June 2026's top 10 robotics developments, including NVIDIA's physical AI tools, AGIBOT's 15,000th robot production milestone, humanoid deployments by Figure at BMW, major funding rounds for Standard Bots ($200M), NEURA Robotics ($1.4B), and General Intuition ($320M), Agility Robotics' SPAC merger announcement, RealSense's new AI-native depth camera, and Collaborative Robotics' Proxie Gen 2 mobile manipulation robot.
Bear Robotics' acquisition of Kinisi isn't just another M&A headline — it's part of a broader pattern where robotics companies are consolidating around AI capabilities rather than hardware. As the industry matures, intellectual property in perception and manipulation is becoming more valuable than the robots themselves.
Alphabet's Intrinsic Intelligence unveiled a modular, AI-powered industrial robotic workcell called IntrinsicOS that replaces traditional robot coding with drag-and-drop automation, aiming to democratize industrial AI for manufacturing applications. The company announced a global AI for Industry Challenge with an $180,000 prize pool focused on solving cable and connector manipulation in electronics assembly, attracting over 5,000 registrations from 1,600 teams across 115 countries.
Bear Robotics has acquired Kinisi Robotics to integrate the KR1 humanoid robot and its physical AI manipulation capabilities into Bear's deployed service robot fleet. The acquisition combines Bear's commercial navigation platform and 16,000+ deployed robots with Kinisi's vision-language-action models and robotic manipulation technology, enabling Bear's robots to move beyond delivery and cleaning to handling physical tasks like picking and sorting.
Collaborative Robotics unveiled Proxie Gen 2, an advanced mobile manipulator robot featuring bimanual manipulation, autonomous task identification via AI-powered 'autotasking', improved payload capacity, and self-swapping batteries. The second-generation platform, informed by nearly 13,000 operating hours of field deployments across healthcare and logistics, reduces parts by 40% while adding autonomous navigation and environment perception capabilities that can identify materials ready to move without human intervention.
CMU researchers have developed VideoManip, a system that trains robots to perform dexterous manipulation tasks by converting videos of people interacting with objects into training data. The system reconstructs 3D hand movements and object interactions from ordinary internet videos, eliminating the need for specialized equipment or teleoperated demonstrations, and addresses a major bottleneck in robot learning by enabling robots to learn from internet-scale datasets.
Kawasaki Robotics will debut its RL030N eight degree-of-freedom robot platform designed for physical AI applications at Automate 2026 in Chicago, along with demonstrations of its Pulseboard inspection technology and other industrial robots. The RL030N combines high-speed motion, enhanced dexterity, and real-time control capabilities to bridge the gap between traditional industrial robotics and AI-driven applications requiring adaptive motion and confined-space manipulation.
PSYONIC has partnered with ABB Robotics to advance dexterous robotic manipulation by combining ABB's GoFa collaborative robot with PSYONIC's Ability Hand, a dexterous gripper originally developed for prosthetics that incorporates touch sensing and compliant design. The collaboration leverages real-world data from prosthetic users to train robots to perform complex manipulation tasks, addressing a key challenge in developing physically intelligent robots capable of handling deformable objects and high-mix, low-volume operations.
Researchers at the Italian Institute of Technology have developed an octopus-inspired soft robotic arm equipped with tactile sensors integrated into artificial suction cups, enabling autonomous underwater grasping and object manipulation. The study, published in Nature Machine Intelligence, represents a significant advance in soft robotics by combining bioinspired design with distributed tactile sensing capabilities for complex autonomous tasks.
Hello Robot Inc. has been recognized by the World Economic Forum as a 2026 Technology Pioneer for its Stretch mobile manipulator robot, which assists older adults and people with disabilities in performing daily tasks. The company, founded by MIT and Google robotics veterans, emphasizes practical, human-centered robot design over spectacle, with Stretch already deployed at hundreds of research institutions and being piloted with individuals with severe mobility impairments.
X Square Robot has open-sourced XRZero-G0, a hardware-software framework that combines VR interfaces and multi-view cameras to enable efficient robot-free data collection for embodied AI training. The company released the G0-Dataset, a 2,000-hour multimodal dataset, along with code and methodology to help reduce real-robot training data requirements by up to 20× and accelerate research in generalizable robotic manipulation and cross-embodiment policy transfer.
Robot Talk Episode 159 features an interview with Maria Koskinopoulou, an Assistant Professor in Robotics and Computer Vision at Heriot-Watt University, discussing her research on autonomous robotic manipulators for surgery, industry, and related applications. Her work spans robotic manipulation, perception, robot vision, medical robotics, and machine learning, supported by major UKRI and EU-funded projects.
Daimon Robotics and Galbot have launched RobOmni, the first omni-modal evaluation benchmark for tactile perception and dexterous manipulation in robots, debuting at ICRA 2026. Built on NVIDIA Isaac Sim, RobOmni provides a standardized framework to measure how tactile sensing improves robotic manipulation tasks like grasping, insertion, and assembly, addressing the industry's lack of unified benchmarks for evaluating physical interaction capabilities beyond vision alone.
AGIBOT hosted the World Challenge 2026 competition with 526 teams from 27 countries competing in embodied AI tasks using real robots and standardized benchmarks. The competition evaluated two tracks—Reasoning to Action (task understanding and execution) and World Model (physical world prediction)—with finalists completing challenges using the AGIBOT G2 humanoid robot, marking a shift toward real-robot validation over simulation-only evaluation. The competition also introduced a supermarket benchmark track for end-to-end mobile manipulation in realistic retail environments.
Researchers from NYU Tandon and the Robotics and AI Institute demonstrated that robots can learn dexterous manipulation tasks more effectively from consistent, synthetic training data generated by planning algorithms than from traditional human teleoperation demonstrations. The study, published in IEEE Robotics and Automation Letters and awarded the IEEE RA-L Best Paper Award, shows that data quality and consistency matter more than quantity when teaching robots complex skills like in-hand object rotation and dual-arm coordination.
MIT researchers have developed Masked Inverse Reinforcement Learning (Masked IRL), a system that uses large language models to help robots understand vague instructions and focus on relevant details from demonstrations. The approach reduces the need for extensive human demonstrations by 80% while enabling robots to safely complete manipulation tasks in warehouses, offices, and homes by automatically clarifying ambiguous prompts and identifying crucial environmental constraints.
This roundup highlights the top 10 robotics stories from May 2026, including humanoid robot production scaling by 1X and partnerships with Bosch, advances in lidar sensors from Ouster, dexterous manipulation AI from Genesis AI, autonomous tire-changing systems from Automated Tire, and new benchmarks for humanoid testing from Fraunhofer IPA. The stories span humanoid development, industrial automation, AI breakthroughs, and emerging robotics startups.
A webinar panel discussion featuring industry leaders from RightHand Robotics, Locus Robotics, and Nomagic will explore advances in robotic case and item picking for order fulfillment, covering machine vision, AI-guided manipulation, and human-robot collaboration. Topics include addressing challenges in manipulating diverse shapes and materials, the role of AI training through data collection, and the future of human supervision in automated warehouse operations.
Researchers at Leiden University have developed microscopic flexible robots that move autonomously without sensors, software, or external control, instead relying on shape-environment interactions inspired by biological microswimmers. These 3D-printed chain-like microrobots demonstrate emergent intelligent behaviors including obstacle avoidance and object manipulation, with potential applications in targeted drug delivery and minimally invasive medical procedures.
NIST has proposed the first standardized performance benchmark for humanoid robots since 2015, establishing a baseline set of locomotion and manipulation tasks to evaluate minimum capabilities across commercial humanoid platforms. The benchmark includes testing for mobility, dexterity, whole-body awareness, and decision-making, with NIST planning to distribute testing apparatus to manufacturers and research facilities while publishing designs for virtual testing.
XELA Robotics will showcase its latest tactile sensing innovations at the 2026 Robotics Summit & Expo, including a novel six-axis force-sensitive robotic fingertip with a nail, magnetic interference compensation, and enhanced delicate grasping capabilities. The company is also demonstrating integrations with open-source platforms like the Universal Manipulation Interface (UMI) for human-robot skill transfer, along with improvements to sensor hardware, software, and communication protocols for industrial manipulation tasks.
Genesis AI, a Khosla-backed startup that raised $105M for foundational AI in robotics, unveiled its GENE-26.5 model along with a human-hand-shaped robotic hand it designed in-house. The startup developed a sensor-loaded data collection glove and uses simulation to accelerate model training, aiming to bridge the embodiment gap in robotics by collecting human manipulation data to train robots for complex tasks like cooking, lab work, and other real-world applications.
Hello Robot announced Stretch 4, an upgraded version of its mobile manipulation platform now priced at $29,950, featuring twice the speed of Stretch 3, extended reach, and a comprehensive sensor array with dual 3D lidars and multiple cameras for improved safety and autonomy. The robot is designed for general-purpose physical AI applications in dynamic environments like homes, with enhanced sensing capabilities to detect obstacles and people while performing manipulation tasks.
RLWRLD unveiled RLDX-1, a dexterity-first foundation model designed specifically for high-degree-of-freedom robotic hands in industrial applications. The model integrates vision, force sensing, temporal reasoning, and memory to handle complex manipulation tasks like pouring, object tracking, and precise contact control across single-arm, dual-arm, and humanoid embodiments.
Two major announcements this week signal a fundamental shift in logistics automation: Locus Robotics unveiled fully autonomous mobile manipulation, while Pickle Robot prepares to share hard-won lessons from real-world deployment. The message is clear — warehouses that don't automate soon won't be competitive.
Genesis AI unveiled GENE-26.5, an AI foundation model designed for dexterous robotic manipulation that enables robots to perform complex, human-level tasks like cooking, lab work, and piano playing. The company also introduced a proprietary robotic hand that mirrors human anatomy and a data-collection system with tactile sensing to enable direct skill transfer from humans to robots, addressing a fundamental bottleneck in training robotics foundation models.
Ghost Robotics, which has shipped over 1,000 legged robots since 2015, will present at the Robotics Summit & Expo in May 2026, with CEO Gavin Kenneally discussing 10 years of deploying quadruped robots in defense, security, and industrial applications. The company recently released a manipulator arm for its Vision 60 robots to expand their manipulation capabilities, and its systems are widely used throughout the Department of Defense.
Ouster released the Stereolabs ZED X Nano, a compact wrist-mounted stereo camera designed for robotic manipulation and imitation learning with high-resolution 1920×1200 RGB and depth capture at up to 120fps. The camera features sub-millimeter depth accuracy, ultra-low-latency processing with direct GPU integration, and ruggedized industrial connectivity, addressing key bottlenecks in robotic vision for physical AI applications. Pre-orders are available now with shipping beginning May 2026.
Chef Robotics announced it has completed 100 million meal servings across customer facilities, claiming this represents more production volume than all other food robotics companies combined. The milestone reflects deployments across a dozen facilities in the U.S., Canada, and Europe, and demonstrates the company's physical AI approach using real-world production data to train models for deformable material manipulation in food manufacturing.
Locus Robotics launched Locus Array, a mobile manipulation system combining an omnidirectional mobile robot, integrated picking arm, and AI-powered perception for autonomous warehouse fulfillment. The system represents the company's evolution toward fully autonomous warehouse operations, capable of handling multiple workflows including picking, putaway, and replenishment while working alongside existing Locus AMRs across 350+ facilities globally.
NVIDIA highlights Physical AI breakthroughs during National Robotics Week, announcing new technologies including Isaac GR00T language models for robot control, Cosmos world models for synthetic data generation, and Newton 1.0 physics engine for manipulation tasks. The company showcases expanded simulation capabilities and real-world applications like surgical robotics with PeritasAI and natural language control via NemoClaw integration with Isaac Sim.
Sanctuary AI demonstrated zero-shot in-hand manipulation using its hydraulic robotic hand, which successfully reoriented a lettered cube 10 consecutive times without dropping it. The system, trained entirely in simulation, transferred seamlessly to real-world performance, showcasing the company's advances in dexterous manipulation policies and embodied AI for its Phoenix humanoid robot.
