🚨 Only 1 week left to apply! Join the 2nd Robotic Collaborative Assembly (RoCo) Challenge @ #IROS2026 🤖 🏭 Industrial Board Assembly 🧱 Brick Assembly ⏰ Proposal deadline: Jul 12, 2026 Take a look at the accompanying demonstrations to see robotic manipulation in action 👀 1️⃣ DexMate bimanual precision assembly in simulation 2️⃣ Sharpa North teleoperation with tactile sensing 3️⃣ Collaborative LEGO assembly in simulation 4️⃣ Real-world bimanual LEGO assembly 🚀 Challenge: rocochallenge.github.io/RoCo… 📝 Register: forms.gle/d2NKNAE7dqSfYZB87 ❓ FAQ: rocochallenge.github.io/RoCo… 💬 Discord: discord.gg/BvxEN5vAh3 #RoCoChallenge #Robotics #PhysicalAI #EmbodiedAI #RobotLearning #RobotAssembly #IndustrialRobotics #BimanualManipulation #Simulation #Research #IROS2026

Proud to be a sponsor of the RoCo Challenge at IROS 2026! 🦾 The RoCo Challenge puts robotic collaborative assembly to the test in real-world manufacturing scenarios: precision gearbox assembly, error recovery, multi-step planning. These are the hard problems that matter most when you're deploying robots on actual factory floors. Challenges like this push the entire community forward, and we're thrilled to support it. See you in Pittsburgh! Let's build. 🔧 #RoCoChallenge #IROS2026 #Dexmate #PhysicalAI #DexterousManipulation

Our paper E2HiL has been accepted to RA-L 2026! Human-in-the-loop RL helps robots learn in the real world, but not every human correction is equally useful. E2HiL selects intervention samples by their effect on policy entropy, pruning both shortcut samples and noisy samples. Across 10 real-world manipulation tasks, E2HiL achieves 24.9% success rate with 9.3% fewer human interventions. Project: e2hil.github.io/ #RobotLearning #ReinforcementLearning #Robotics

🚨 Join Our LIVE INFO SESSION — May 21 at 9:30 PM EST 🚨 Curious about the 2026 #RoCoChallenge at #IROS? Join our live session to learn more about this year’s competition, tasks, evaluation, and how to participate! 🎥 Live session link available on our website: rocochallenge.github.io/RoCo… This year, the RoCo Challenge at #IROS26 is expanding to include: 🤖 More robot embodiments — including bimanual grippers and dexterous hands 🛠️ More assembly tasks — including industrial board assembly and brick assembly 🙌 Huge thanks to our co-organizers: Haichao Liu, @philipYHuang , @yushijinhun , Yuheng Zhou, Liming Chen, @Ken_Goldberg , Yejun Gu, @JiaoyangLi9 , Jun Liu, @soujanyaporia , Masayoshi Tomizuka, @Jianfei_AI, Steven Yeung, Hao Zhao, @ZiweiWangNTU , @ChangliuL. Special thanks to our sponsors: @DexmateAI, @SharpaRobotics, @TARSRobotics, @roboforce_ai for making this competition possible. 🔗 Ready to join us? 🔹 🌐 Learn more: rocochallenge.github.io/RoCo… 🔹 📝 Register here: docs.google.com/forms/d/e/1F… We’re excited to push the frontier of robot manipulation and embodied intelligence together with the community! 🔥 #PhysicalAI #EmbodiedAI #Robotics #AI #Assembly #BrickAssembly #IndustrialAssembly #Pittsburgh

🚀 RoCo Challenge @ AAAI 2026 — On-site Competition Live! RoCo Challenge @ AAAI 2026 (Robotic Collaborative Assembling for Human-Centered Manufacturing) is a research challenge jointly organized by Nanyang Technological University (NTU) and Agency for Science, Technology and Research (A*STAR). The challenge focuses on robotic collaborative assembly in industrial manufacturing, where robots must: 1. Perform high-quality assembly, 2. Understand human progress, 3. Detect and recover from realistic human errors. 🤖 Task: Gearbox assembly under evolving workspace conditions 📌 Key scenarios: 1. Assembly from scratch 2. Resume from partial state 3. Error detection & recovery After intense online competition, 6 teams advanced to the on-site physical track, taking place on Jan 24–25, 2026, with bilingual live streaming begins at Jan 25 13:30: 🌲 Chinese platforms: Rednote & Bilibili @ PINE_Lab_NTU 🌲 English platform: YouTube @ Pine-wn4gh 🏆 The challenge concludes with an award ceremony and technical presentations at AAAI Conference on Jan 26, 2026. 📡 Stay tuned for live demos of robust, error-aware robotic assembly in the real world!

🚀 Postdoc Position in Embodied AI & Robotics @ NTU (Singapore) PINE Lab (Perception and embodied INtElligence Lab) at Nanyang Technological University (NTU) is actively recruiting Postdoctoral Research Fellows to work on cutting-edge research in Embodied AI and Robot Learning. 👤 Supervisor Dr. Ziwei Wang is an Assistant Professor in the School of Electrical and Electronic Engineering, NTU. He received his B.E. (2018) and Ph.D. (2023) from Tsinghua University, and was a Postdoctoral Researcher at the Robotics Institute, Carnegie Mellon University. His research focuses on embodied foundation models and generalizable visuomotor policies across environments, tasks, and embodiments (single-arm, bimanual, mobile manipulation, dexterous hands). He has published 50 papers in top venues including CVPR, NeurIPS, TPAMI, ICRA, CoRL and IROS. 🔗 Homepage: lnkd.in/gthuiACn 🔬 Research Topics (including but not limited to): 1. VLA with Real-world Reinforcement Learning 2. Skill-based VLA 3. World Models for VLA Training and Planning 4. Mobile Bimanual Manipulation 5. Dexterous Manipulation & Tactile Sensing 🎯 Who We Are Looking For 1. PhD in Robotics, CS, AI, EE, or related fields 2. Strong background in robot learning / RL / vision / multimodal learning 3. Solid research and engineering skills; real-robot experience is a plus 4. Highly self-motivated and collaborative 📩 How to Apply Please email your application to ziwei.wang@ntu.edu.sg with subject: Postdoc Application – Your Name – Institution Include: CV, cover letter, representative papers, and referee contacts. 🌐 Lab website: lnkd.in/g8yRCfUH Feel free to share or reach out if interested

The first comprehensive survey of RL-VLA： A Survey on Reinforcement Learning of Vision-Language-Action Models for Robotic Manipulation！ 🔗：github.com/Denghaoyuan123/Aw… 🚩 Explain how RL enhances VLA systems by introducing reward driven exploration and self corrective behavior, which helps overcome the limitations of imitation based training and improves performance in real world and out of distribution scenarios. 🚩 Organize recent advance into a unified RL VLA framework that covers model architecture, training methodologies, deployment in real environments and evaluation practices. 🚩 Summarizes key challenges in action representation, reward design, dynamics modeling, stable optimization and safe exploration, while outlining important directions for future research. The survey provides a clear and structured reference for researchers working on robot learning, multimodal foundation models, VLA systems and autonomous manipulation.

🤖 Join the RoCo Challenge @ AAAI 2026! 🏁 The RoCo Challenge 2026 invites researchers and innovators worldwide to explore the future of Human-Robot Collaboration (HRC). Hosted during AAAI 2026, this competition—jointly organized by Nanyang Technological University (NTU), A*STAR, and Carnegie Mellon University (CMU)—focuses on embodied intelligence in complex, real-world manufacturing and service environments. 🏆 Participants will tackle multi-stage collaborative tasks in both simulation and physical settings, showcasing perception, reasoning, and coordination skills. Attractive cash prizes, AAAI exposure, and publication opportunities await top teams! rocochallenge.github.io/RoCo…

@ICCV2025 paper! 🆕 Meet AnyBimanual: a plug-and-play way to transfer any pretrained unimanual policy into a general bimanual policy with just a few demos. 🔧 Skill Manager: dynamically schedules reusable unimanual skill primitives for each arm 👀 Visual Aligner: soft-masking to align single-arm visual representations in bimanual scenes 📈 17.33% avg success on 12 RLBench2 tasks vs SOTA 🤖 84.62% average success on 9 real-world bimanual tasks 📄 Paper Link: arxiv.org/abs/2412.06779 🔗 Project Page: anybimanual.github.io/ #EmbodiedAI #Robotics #Bimanual #EmbodiedAI #ICCV2025

@ICCV2025 paper! 🆕 Meet GWM: Towards Scalable Gaussian World Models for Robotic Manipulation 🌍 3D Gaussian splats → scalable, geometry-aware world modeling 🎥 Action-conditioned future imagination with diffusion transformers 🤖 Serves as both a strong encoder for imitation & a neural simulator for RL ⚡️ 30% higher real-world success vs. SOTA 📄 Paper Link: arxiv.org/abs/2508.17600v1 🔗 Project Page: gaussian-world-model.github.… #EmbodiedAI #Robotics #WorldModels #GaussianSplating #ICCV2025

🚀 New @IROS2025 paper! 《 Embodied Instruction Following in Unknown Environments 》 🌍🤖 Tired of LLMs hallucinating actions in unseen spaces? This framework dynamically plans & executes tasks while exploring unknown scenes. 🔗 Read more: pine-lab-ntu.github.io/stati… #EmbodiedAI #Robotics #VLA #IROS2025

🚀 New @IROS2025 paper! 🆕 Meet AnyView: the 3D object detector that adapts to your frames — 1 or 50, it just works. 📦 No more overfitting to fixed views 🤖 Built for real-world real-time mobile robots ⚡️ Efficient, scalable, deployable 📄 Paper Link: pine-lab-ntu.github.io/stati… #Robotics #AI #ComputerVision #IROS2025