AI Robot Association (AIRoA)

We are Hiring! Service Robot System Engineer – Real-World Deployment | Tokyo (Heiwajima) ■Position: Service Robot System Engineer (R&D-025) ■What you'll do & need: ・Design service robot systems for real-world environments ・Plan and execute PoC studies and field experiments ・Select, improve, and iterate on robot hardware ■Experience: M.S. in Robotics or related field; hands-on with physical robots (ROS/ROS2, C++/Python) ■Location: Tokyo Join us building a global open robot data ecosystem! 🔗Apply here: jobs.workable.com/view/7AhghLd6L… #hiring #robotics #servicerobots #systemengineering

We are Hiring! Mechanical Engineer – Humanoid Robotics | Tokyo (Ota-ku) ■Position: Mechanical Engineer (R&D-024) ■What you'll do & need: ・Design robot structures, joints, links, and mechanisms (3D CAD mastery required) ・Select and integrate actuators, gearboxes, sensors for optimal performance ・Create manufacturing drawings and iterate fast from prototype to testing ・Collaborate hands-on with software/electronics engineers on real hardware ■Experience: Mechanical design background (structures/mechanisms); robotics or precision machinery a plus ■Location: Tokyo Join us building a global open robot data ecosystem! 🔗Apply here: jobs.workable.com/view/8uaTAknDW… #hiring #robotics #humanoid #mechanicalengineering

🔹Robotics NEXT Tokyo 2025 アーカイブ配信のお知らせ🔹 近年、Physical AI / Embodied AI に代表される大規模言語モデルのロボットへの適用、ヒューマノイド・スタートアップの台頭など、技術とビジネスの両面で新しい潮流が生まれています。 10月16日に開催された Robotics NEXT Tokyo 2025 では、こうした変化の最前線に立つ国内外の専門家が登壇し、ロボット技術の将来を多角的に展望しました。 Agility Robotics の Jonathan Hurst 氏、Preferred Networks の 岡野原 大輔 氏 による基調講演に加え、AIRoA CTO 松嶋 も登壇し、オープンなロボット基盤モデルプラットフォーム開発の取り組みを紹介しました。 経営層、企画部門、AI技術者、ロボット開発者など、ロボティクスの未来を捉えたいすべての方におすすめです。ぜひご視聴ください。 📆配信期間：2025年12月8日～2026年1月31日 💡参加料：無料 ▶ 視聴URL：events.nikkeibp.co.jp/event/2025/rnt… 🔹Now available: Robotics NEXT Tokyo 2025 On-Demand🔹 The robotics world is entering a transformative phase — from Physical AI and Embodied AI powered by large language models to the rapid growth of humanoid robotics startups around the globe. Robotics NEXT Tokyo 2025, held on October 16, featured global leaders and innovators exploring the future of robotics technology and its real-world adoption. Keynotes by Jonathan Hurst (Agility Robotics) and Daisuke Okhonohara (Preferred Networks) were joined by AIRoA CTO Matsushima, who introduced efforts to develop an open robotic foundation model platform. This webinar is ideal for executives, business strategists, engineers, and robotics professionals looking to understand both the technology frontier and the latest industry adoption trends. 📆 Available: Dec 8, 2025 – Jan 31, 2026 💡 Free access ▶ Watch here: events.nikkeibp.co.jp/event/2025/rnt… #Robotics #RoboticsNextTokyo #PhysicalAI

At the IREX (Nov 27), Prof. Tetsuya Ogata delivered a keynote on Physical AI—AI gaining intelligence through physical interaction and embodiment, now entering societal deployment. ◼️ Key Highlights: ・Reinforcement Learning Advances: Enables walking and whole-body control on affordable humanoids (e.g., Unitree G1 Kungfu Kid V6.0) using cheap QDD motors and simulation training for high-fidelity results.​ ・Imitation Learning Surge (2024+): Platforms like Figure Helix, Tesla Optimus, and 1X NEO use thousands of task demos + foundation models for manipulation skills.​ ・VLA & End-to-End Shift: Models (RT-2, OpenVLA, GR00T) predict future sensor-actions directly, turning human skills into robot "subjective world models." Trend: AI firms driving robotics hardware.​ ・AIRoA's Position in Industry: Launched Dec 2024 as a public organization, AIRoA builds a robot data ecosystem by collecting cross-industry operation data, developing and openly sharing foundation models to accelerate AI robot development and societal implementation.​ ◼️ Co-Evolution Path These three elements—robotics as multimodal AI testbeds, AI handling touch/sound for real-world actions, and social acceptance via pilots—interact mutually to drive AI-robot co-evolution based on societal needs. #PhysicalAI #AIRoA #Robotics #Humanoids #IREX

SPS 2025 is upon us at Nuremberg, Germany, Nov 25–27—a premier global hub for industrial automation and AI robotics innovation! sps.mesago.com/nuernberg/en.h… This year, over 1,200 exhibitors showcase cutting-edge AI and robotics solutions driving the future of smart manufacturing. Notable AI robotics companies expected include Delta Electronics, ABB Robotics, FANUC, Yasukawa, Boston Dynamics, Standard Bots, Tesla Optimus, and more.​ At SPS 2025, industry leaders demonstrate advancements from cobots with AI edge computing and digital twins to autonomous mobile robots and embedded vision—empowering factories to scale productivity, safety, and sustainability. Will your enterprise leverage SPS innovations to accelerate automation transformation? The three-day event offers keynotes, demos, and networking with robotics experts shaping intelligent industrial production worldwide. #SPS2025 #AIrobotics #SmartFactory #IndustrialAutomation #Innovation

[Physical AI News Nov.25] 📦 The Future of Logistics Automation: Introducing ROBOTERA's Star-Act L7 The Star-Act L7 humanoid robot by ROBOTERA(星动纪元) brings unprecedented agility, dexterity, and AI-driven flexibility to warehouse logistics, bridging the critical gap in automated item picking with its advanced Visual-Language-Action (VLA) embodied AI model and humanlike physical capabilities. ◼️ Background & Challenges Conventional logistics automation struggles with the final step of flexible item picking in warehouses. Automated Guided Vehicles handle transport well, but picking individual SKUs, especially during peak e-commerce times, remains labor-intensive and error-prone. Rigid robotic systems require constant reprogramming due to the sheer variety of SKUs. ◼️ Methodology Robotera's solution deploys the full-size Star-Act L7 humanoid robot with a proprietary Visual-Language-Action (VLA) AI brain. The L7 uses end-to-end embodied AI from raw visual and language input to motor action output, enabling real-time corrections and dynamic handling of new items without retraining. ◼️ Key Innovations The L7 robot features a highly dexterous five-finger hand with 12 degrees of freedom, a 3-DOF waist for vertical reach, and a cross-shaped wrist eliminating blind spots for stable grasping. The system's ERA-42 software integrates tightly with Warehouse Management Systems to autonomously manage picking workflows and data feedback loops optimizing performance. ◼️ Results Robotera claims this is the world's first deployment of an end-to-end VLA humanoid in real-world logistics, closing the flexible picking gap. The humanoid fits existing warehouse layouts, handles diverse SKUs without constant reprogramming, and supports scalable industrial automation. As logistics demands grow ever more complex and varied, the Star-Act L7 sets a new standard, inviting us to reconsider how humanoid robotics can deliver scalable, adaptable automation for the future of supply chains. How will this shape the next wave of smart warehouses? Here is an overview of The Star-Act L7: lnkd.in/du8C3tFM #AI #PhysicalAI #Robotics #Logistics #Automation

[Physical AI News Nov.25] 🚗 30,000+ BMW X3s Produced with Contribution from Figure 02 Bringing advanced humanoid robots from the lab to automotive production lines is a major technical and operational challenge. Over an 11-month deployment at BMW Group’s Spartanburg plant, Figure’s second-generation humanoid robot, Figure 02, ran daily 10-hour shifts loading more than 90,000 sheet-metal parts and supported the assembly of over 30,000 BMW X3 vehicles, covering approximately 1.2 million steps or over 200 miles on the factory floor. Key performance indicators included cycle time (84 seconds total, 37 seconds loading), placement accuracy (targeting over 99% success per shift), and zero human interventions per shift. Figure 02 achieved a demanding 5-millimeter placement tolerance within just 2 seconds using advanced locomotion and hand-eye coordination algorithms. Operational insights, especially relating to forearm hardware reliability, directly informed the design improvements in the next-generation Figure 03. This deployment marks a significant milestone, proving data-driven humanoid reliability and precision at scale. As Figure 03 rolls out, it will be fascinating to see how far these robots can evolve to redefine factory work. Here is an overview of Figure 02: figure.ai/news/productio… #AI #robotics #Figure02 #BMW #humanoid #manufacturing

Sunday has introduced an innovative household robot, Memo, capable of executing complex daily chores by learning from extensive real-world data collected via a unique Skill Capture Glove worn by 500 remote data collectors. This approach allowed the training of the ACT-1 model with over 10 million episodes from 500 homes, enabling Memo to perform tasks like laundry folding and dishwasher loading autonomously. Memo’s semi-humanoid wheeled design enhances safety and stability, especially by preventing falls during battery depletion, while achieving up to four hours of operation per charge. It also uses 3D home mapping with object location prediction to navigate unfamiliar environments effectively. Beta testing is scheduled for 2026, with commercial release around 2027-2028, priced near $10,000. The focus on dishwasher loading addresses an everyday household challenge. This development marks a significant step in making domestic AI robotics practical and reliable. How will such advanced data-driven learning transform the future of home automation and human-robot collaboration?

[Physical AI News Nov.21] Physical Intelligence Introduces π*0.6, Advancing Robot Learning from Experience ◼️Background & Challenges Conventional imitation learning typically caps at about 50% task success. The main stumbling block is cascading errors—for instance, a slight misalignment of 0.5mm during object grasping causes a ripple effect, leading the robot into unfamiliar states and compounded failures. ◼️Methodology ・Inspired by human skill acquisition, the process involves: Demonstrations by teleoperation to gather initial data. ・Real-time corrections by human operators during autonomous runs to fix mistakes. ・Repeated autonomous practice where the robot refines its actions based on feedback. ◼️Key Innovations A value function predicts remaining steps in a task by combining demonstration and correction data. Using this, actions are ranked as positive or negative, enabling the model to learn from experience by emphasizing better behaviors. ◼️Results Tasks like coffee making and laundry folding saw productivity double. The system achieved robust performance, successfully operating a coffee machine continuously for over 13 hours. For manufacturing, logistics, or any sector deploying collaborative robots, this evolution signals a new era of scalable autonomy. As robots advance from instruction to mastery, how could your organization leverage experiential learning to push operational boundaries? Full article is here: lnkd.in/dbug5X2y 🔎Join the Future of AI Robotics : airoa.org #AI #robotics #autonomy #factoryautomation #innovation

UBTECH’s Walker S2: Advancing Continuous Humanoid Robotics in Industry 🤖🔋 UBTECH Robotics has achieved a milestone by delivering its Walker S2 humanoid robots on a mass scale, enabling new levels of factory automation. These robots combine human-like mobility and dexterity with innovative autonomous battery swapping that supports nearly uninterrupted operation. With a payload capacity of 15 kg and precise spatial awareness from stereo vision, Walker S2 is engineered for complex industrial tasks. Its AI system allows independent task planning and problem-solving, exemplifying practical humanoid robot deployment beyond prototypes. 🔎Join the Future of AI Robotics : airoa.org #HumanoidRobotics #FactoryAutomation #AI #Robotics #UBTECH

[Physical AI News Nov.12] VLA-R1 is a breakthrough AI model that makes robots think before they act. By combining visual, language, and action understanding with step-by-step reasoning, it can better handle complex tasks in the real world. leaderobot.com/news/6633

[Physical AI News Nov.11] QianJue Technology’s research team published “When Do Neural Networks Learn World Models?”—a study that captured how neural networks spontaneously learn world models. arxiv.org/abs/2502.09297

[Physical AI News Nov.10] Tesla’s Master Plan Part 4: The Age of Physical AI⚙️ Tesla has unveiled "Master Plan Part 4," marking a strategic pivot from electric vehicles to AI and robotics. tesla.com/master-plan-pa…

[Physical AI News Nov.9] 🍔DoorDash x Serve Robotics: Redefining Urban Delivery DoorDash and Serve Robotics have joined forces to expand autonomous sidewalk delivery robots across major U.S. cities, beginning in Los Angeles. about.doordash.com/en-us/news/doo…

[Physical AI News Nov.8] Alibaba Group and NVIDIA have teamed up to embed Nvidia’s full “Physical AI” toolkit into Alibaba Cloud’s machine learning platform. aibusiness.com/robotics/aliba…

[Physical AI News Nov.7] 🌐Massachusetts Institute of Technology and Toyota Research Institute have unveiled “Steerable Scene Generation,” a generative AI platform that builds ultra-realistic 3D environments for training robots. news.mit.edu/2025/using-gen…

[Physical AI News Nov.6] Robots that Think, Plan, and Adapt 🧠 Google DeepMind’s new Gemini Robotics 1.5 is changing the game for embodied AI. deepmind.google/blog/gemini-ro…

[Physical AI News Nov.5] Unitree unveils its human-sized humanoid robot “Destiny”  A humanoid robot designed for real-world use. Standing 180 cm tall and weighing 70 kg, Destiny can move, balance, and work safely alongside people. unitree.com/H2