RoboSignal Daily · 2026-10-20 · Tuesday
ROBOSIGNAL
Tencent Robot X Team Creates 'Thinking and Imagining' Robot Brain - RxBrain
The Tencent Robot X Team, Fujita Lab, and Tencent Hunyuan Team jointly released a study proposing a new embodied cognition foundational model called Hy-Embodied-RxBrain (shortened as RxBrain). This model aims to enable robots to not only describe steps in language when planning actions but also generate mental images of the world state after each step in their mind.
Research
Tencent Robot X Team Creates 'Thinking and Imagining' Robot Brain - RxBrain
The Tencent Robot X Team, Fujita Lab, and Tencent Hunyuan Team jointly released a study proposing a new embodied cognition foundational model called Hy-Embodied-RxBrain (shortened as RxBrain). This model aims to enable robots to not only describe steps in language when planning actions but also generate mental images of the world state after each step in their mind.
arXiv Robotics — research abstractsScaling Online Exploration for Robot Agents
This research presents RAX, an open-source framework for robot agents that enables continuous self-improvement through online exploration. The framework allows agents to acquire experience, synthesize executable programs, and perform complex tasks without human demonstrations. RAX demonstrates significant performance improvements in both simulation and real-world settings, with results showing a +37.4% average improvement over top baseline models.
Zero-shot Deployment of a 5-Motor Bicycle Robot with Reinforcement Learning
Researchers from RAI Institute have developed a bicycle-shaped robot, UMV, capable of high-speed movement, jumps, and aerial flips using only five motors and reinforcement learning. The robot achieves 8 m/s speed, 1-meter jumps, and half-second aerial flips without specialized hardware for balance. Training occurs entirely in simulation, with zero-shot deployment to physical hardware. The study highlights the potential of reducing mechanical complexity and relying on learning algorithms for dynamic robotic tasks.
Industrial Scenario Needs for Embodied Intelligence: Doko Robotics Offers Answers with Over 5000+ Deployments
Doko Robotics unveiled its full product line of embodied intelligence solutions at the Industrial Expo, including humanoid robots, force-controlled mechanical arms, and joint modules, with applications in semiconductor manufacturing, welding, and automotive industries. The company highlights its integration of perception, motion, and intelligence, supported by over 5000+ deployments and industry awards.
Dexterous Hands in Robotics: A Shift from High Degrees of Freedom to Practicality
The article explores the evolving landscape of dexterous hands in robotics, noting a trend away from high degrees of freedom (DoF) toward more practical metrics like durability, cost, and operational efficiency. It highlights how industrial standards are influencing the market, with grippers and simpler hands becoming more prevalent. The text also discusses the limitations of current VLA models in controlling complex robotic hands and the role of data collection in training these systems.
Doku Robotics Unveils Full Line of Embodied Intelligence Products for Industrial Applications
Doku Robotics showcased its full range of embodied intelligence products, including joint modules, force-controlled mechanical arms, and humanoid robots, at the 2026 Global Leaders Conference. The company emphasized its focus on industrial deployment, with applications in semiconductor manufacturing, welding, and automotive sectors. The products are designed to meet the specific needs of industrial environments through advanced perception, motion control, and task execution capabilities.
arXiv Robotics — research abstractsNo Eyes, No Problem: What Does Touch Contribute to Blind Baoding Ball Manipulation?
The study combines sensor-aligned embodiment, tactile calibration, and novel domain randomizations to bridge the sim-to-real gap, setting a new benchmark for tactile-driven dexterous manipulation.
GPT-6-Astra's Embodied Capabilities Face Major Test
This research paper evaluates GPT-6-Astra's embodied intelligence capabilities across 12 simulated benchmarks and 8 real-world robot platforms. While it achieves high scores on structured tasks like LIBERO-Long++ and SIMPLE, real-world execution reveals significant limitations, including safety issues, precision challenges, and difficulty adapting to dynamic environments. The study emphasizes the gap between simulated performance and real-world deployment, highlighting the need for improved execution systems and task understanding.
Manipulation Solved by Feeling: Vision Isn't as Important as You Think
This paper presents a breakthrough in robotic manipulation by showing that contact-based joint position signals can enable precise, zero-shot insertion tasks such as pegs, gears, and nuts without cameras or tactile sensors. The approach leverages joint drift to perform blind, fine-grained assembly tasks, marking a significant step toward general-purpose robotics.
Nanjing's Embodied Intelligence Ecosystem Takes Shape at Physical AI and World Model Conference
The 2026 Physical AI and World Model Conference in Nanjing marks a significant step in the development of embodied intelligence, with local companies like Yuanfang Intelligent and Jiajia Vision showcasing advancements in world models and physical AI. The event underscores the region's commitment to fostering an ecosystem for real-world AI applications.
Tech Xplore — RoboticsRobot AI predicts future scenes to guide its next moves
A research team from Seoul National University College of Engineering has developed Dynin-Robotics, a robotic foundation model that can understand human instructions, predict future scenes, and generate actions to complete tasks. The model outperforms existing models in real-world robot manipulation tasks, achieving an average success rate of 78.4%. The paper has been accepted to NeurIPS 2026 and demonstrates the potential of integrating instruction understanding and scene prediction for robotic control.
Data & Training
China's Robotics Data Infrastructure Expansion
China has established over 70 training fields for embodied AI, with additional projects in planning, driven by government policies and industry investment. These centers aim to generate high-quality data for robot training, but face challenges such as data quality, sustainability, and market dynamics. The article explores both successful initiatives and failed projects, highlighting the evolving landscape of data collection and its role in advancing robotics.
Briefs
- 15 Humanoid Robots Start Work: Why Factories Are Willing to Wait for Them to Speed UpChinese robotics — Dataset and collection discovery
- LAW: Unifying Language, Action, and World Modeling for Agentic Physical IntelligencearXiv Robotics — research abstracts
- 538 Hours of 'Extremely Messy' Human Data, Real Robot Takes Less Than an Hour: Hong Kong University Teaches Robots to Unpack Packages, Use Hot Glue Guns, and Make Cold DrinksChinese robotics — Dataset and collection discovery
- The Robot I Didn’t Expect to BuildIEEE Spectrum — Robotics
- SenseSwarm: Swarm Exploration via Neural Structure Estimation with Shared World-Anticipation for Rapid MappingarXiv Robotics — research abstracts
- Swift-ROS2 3.0: Native Swift ROS 2 Client Library with Generated Common Interfaces and Own OrganizationROS Discourse — Community updates
- 15 Humanoid Robots Start Work: Why Factories Are Willing to Wait for Them to Speed UpChinese robotics — Dataset and collection discovery
- 15 Humanoid Robots Deployed in Factory for Automotive Part HandlingChinese robotics — Hardware and sensing
- RoboFollow: A Benchmark for Evaluating Language Understanding in Robot ControlChinese robotics — Research news discovery
- Asimov Open SDK, Humanoid Robot Competition Shifts to Developer EcosystemChinese robotics — Research news discovery