Embodied perception Latest coverage

TacClaw-Infinity: A Large-Scale Vision-Tactile-Language-Action Dataset for Dexterous Manipulation

This paper introduces TacClaw-Infinity, a large-scale structured dataset for contact-rich manipulation. Collected using the TacClaw tactile-enabled handheld interface, it includes over 10,000 hours of data, 1 million episodes, and 1 billion frames spanning more than 1,000 task types from homes, labs, industrial sites, retail stores, and outdoor environments. The dataset features synchronized visual observations, fingertip tactile signals, and end-effector trajectories without requiring a target robot during collection. A scalable pipeline for trajectory reconstruction, quality control, filtering, and structured annotation is developed. Validation at the data level assesses the collection interface and action-space compatibility, while model-level validation demonstrates cross-embodiment transfer across three heterogeneous robots without robot-specific demonstrations.

$S^3$-Nav: Structured Spatial-Semantic Decision Making for Language-Guided Robot Navigation

This research introduces $S^3$-Nav, a framework combining online mapping, persistent object-relation memory, and goal-conditioned navigation for robots operating under partial observability. Using RGB-D inputs and SLAM, the system maintains a structured state of object records and confirmed relations without needing a pre-existing map. A goal-conditioned evidence interface selects relevant information to guide high-level decisions on exploration, approach, and stopping. The framework validates provisional policy decisions against persistent memory and geometric evidence before execution. Evaluations across multiple environments demonstrate improved closed-loop navigation reliability and reduced language-model token consumption when using capacity-controlled evidence exposure.

Embodied RSI Paradigm Breaks Through Zero-Shot Generalization Bottleneck for Robots: Recursia Outperforms NVIDIA ASPIRE by 10 Points in Long-Term Tasks

This article introduces the Recursia framework, which achieves embodied recursive self-improvement (RSI) by organizing the system as an internal and external dual-loop structure, enabling generalization in robot zero-shot long-sequence tasks. Recursia achieved success rates of 28.4% and 53.4% in LIBERO-Pro Long tasks, respectively 5.8 and 15.1 percentage points higher than NVIDIA ASPIRE. The study emphasizes the importance of integrating execution feedback into high-level decision-making and explores the challenges of maintaining task success across different execution contexts.

CylindTrack: Depth-Aware Cylindrical Motion Modeling for Panoramic Multi-Object Tracking in Mobile Robots

CylindTrack is a depth-aware cylindrical tracking-by-detection framework for panoramic multi-object tracking in mobile robots. It addresses coordinate discontinuities and depth estimation fluctuations by jointly modeling panoramic topology and temporal depth consistency. The framework improves identity consistency and trajectory continuity, achieving HOTA scores of 33.67 and 31.12, and IDF1 scores of 40.45 and 34.33 on QuadTrack and JRDB datasets, respectively. It runs at 28.56 and 21.34 frames per second on an NVIDIA GeForce RTX 3090 GPU.

CoSMO: Conformal Semantic Mapping for Open-World Safe Robot Navigation

This research presents CoSMO, a framework that transforms vision-language model embeddings into online-calibrated coverage sets and safety-aware obstacle dilations to ensure safe navigation in open-world environments. By calibrating directly in the embedding space, CoSMO avoids score rescaling issues as vocabulary expands. Experiments on the Matterport3D dataset show a 61.6% reduction in semantic collisions compared to geometry-only baselines while maintaining formal coverage guarantees. Real-world tests on a quadruped robot achieved an empirical coverage of 90.73% against a 90% target, demonstrating generalization under distribution shift and delayed expert feedback.

Soft Robotic Fish Combines High-Speed Swimming with 3D Maneuverability

Researchers from the Shenyang Institute of Automation have developed a soft robotic fish that can swim at high speeds and perform agile three-dimensional maneuvers. The design integrates a rigid–flexible coupled propulsion system with dynamic buoyancy regulation, inspired by biological mechanisms such as muscle-skeleton coordination and swim bladder function. The robotic fish achieves a maximum speed of 2.13 body lengths per second and can perform complex 3D movements like turning, ascending, and spiral ascent, offering a potential approach for high-performance underwater robots.

MagicAtom's EchoPolicy-0 and MagicBrain Frameworks Achieve Top Performance in MolmoSpaces Benchmark

MagicAtom's EchoPolicy-0 framework achieved a 75.4% average success rate in MolmoSpaces Combined benchmark, ranking first globally, while MagicBrain achieved 72.4% with a 94.4% grasp success rate. Both frameworks address challenges in VLA by integrating visual language models with geometric tools and action policies, improving task execution and generalization in complex environments.

NovaSplat: Model-Free 6D Object Pose Estimation via 3D Gaussian Splatting Onboarding and Refinement

NovaSplat is a model-free pipeline for 6D object pose estimation that trains a 3D Gaussian Splatting model from a short smartphone video. It serves as both a template for detection and a differentiable backbone for pose refinement, outperforming model-based methods on a household-object benchmark and model-free competitors on an industrial benchmark with textureless and specular surfaces. A faster inference variant, NovaSplat-Lite, is also introduced.

Two-armed robot throws and catches balls with human-like movements

AthenaZero, a new dual-arm robotic system developed by researchers at the Robotics and AI Institute, demonstrates agile, human-like movements in tasks such as throwing, catching, and batting. The robot's design incorporates low inertia and backdrivability, enabling more natural and controlled interactions with objects. The study explores the challenges in robotic manipulation and presents a novel approach to improving dexterous interaction with the environment.

MagCilia: A Compact Magnetociliary Tactile Sensor with 3D Force Sensing for Robotic Contact Perception and Grasping Feedback

This paper introduces MagCilia, a compact tactile sensor that integrates a flexible magnetic-cilia structure with a Hall sensor to enable three-dimensional force sensing. Using quasi-static finite element analysis and causal history fusion regression (CHFR), the system reconstructs forces from coupled magnetic channels with root-mean-square errors of 0.40 N, 0.57 N, and 0.69 N for Fx, Fy, and Fz respectively, achieving coefficients of determination between 0.90 and 0.93. Robotic experiments demonstrate the sensor's ability to guide gripper adjustments based on tangential forces and monitor multi-axis loads under external perturbations. Additionally, frequency-domain features allow the system to distinguish six surface categories with 99.39% accuracy in cross-validation grouped by acquisition session, and an online demonstration successfully identifies all six tested surfaces.

Feather star-inspired underwater robot gets 3D maneuverability from only two actuators

Researchers have developed an aquatic soft robot inspired by feather stars that can move in all three dimensions using only two actuators. The design utilizes mechanical intelligence to achieve efficient 3D maneuverability through three modes: jellyfish, fish, and rotor. The robot can explore underwater spaces, lift objects, and work collaboratively with other robots, with future plans to develop a fully wireless version.

Chinese Robotics — Hardware and Sensing

The article covers a wide range of robotics and embodied AI developments, including product launches, industry events, and technological advancements. It highlights key players such as Agility Robotics, Figure AI, and others, showcasing their latest innovations in humanoid robots, industrial applications, and AI integration. The content also includes major funding rounds, new product releases, and real-world deployment examples across various sectors.

Threadlike motor uses sliding fibers to drive flexible robotic devices without gears

Researchers from EPFL's Soft Transducers Laboratory have developed FiberMotor, a threadlike motor that converts electrical energy into motion through the sliding of concentric fibers. This gearless design enables flexible, bidirectional movement, making it suitable for soft robotic applications. The motor's small size and lightweight construction allow for integration into smart textiles and wearable devices, with potential uses in assistive robotics and haptic systems.

After a $1.4 Billion Valuation, Skild AI Doesn't Want to Just Be the Brain

This platform aims to serve as a reference interface for physical AI, allowing developers to deploy Skild Brain models across various robotic bodies. The BRIDGE platform is designed to reduce the cost of adapting AI models to different robot bodies and is intended to become a standard in the industry. The release of BRIDGE is part of Skild AI's broader strategy to define the future of embodied AI and physical robot interfaces.

Factory Renovation Robots, Robots Redefine Factory Standards

The article explores how industrial robots are continuously optimized and adjusted in real production environments to meet factories' demands for continuous operation, high precision, and task completeness. Through the failure analysis of Figure 02 in the BMW factory, improvements to the UBTech Walker S1 and S2, and the deployment of Agility Digit in warehouses, the article demonstrates how robots adapt their design, endurance, and task execution capabilities to specific workstation requirements. The article emphasizes that factories not only define the functional boundaries of robots but also determine their technical priorities and product iteration directions.

Want to Transform Factory Robots, First Be Transformed by the Factory

The article explores how industrial robots are being shaped by factory environments.

Video: Dexterous robot hand learns the delicate art of silk embroidery

Agilink's OmniHand 3 Ultra, trained by a master of Suzhou embroidery, has learned to split silk strands, thread a needle, and stitch with precision. The hand uses vision-based sensors and direct-drive mechanics for fine control. The article contrasts this with other robotic hands, such as LinkerBot and Boston Dynamics' Atlas hand, to explore different approaches to dexterous manipulation in robotics.

Embodied Intelligence Education in Practice: Four-Stage Transition from Classroom to Production Line

This article presents a comprehensive guide to implementing embodied intelligence education in robotics, focusing on the integration of physical interaction, sensor data, and real-world problem-solving. It outlines a four-stage model for transitioning from classroom to industrial application, emphasizing sensor calibration, algorithm visualization, and real-world testing. The tutorial includes practical insights, common pitfalls, and real-world examples to support educators and students in developing hands-on, industry-relevant skills.

Rho: A Foundation for Efficiently Adaptable VLA Models

Rho is a new foundation model for vision-language-action (VLA) models in robotics, designed to improve adaptation to specific robots and tasks. The paper introduces a training recipe that separates embodiment adaptation from task adaptation, reducing the amount of task-specific data needed for fine-tuning. Rho variants are pre-trained on different robots and show strong performance on physical tasks, outperforming existing models in several benchmarks. The model also supports online adaptation through a latent policy that learns from human corrections.

LeRobot@LeRobotHF· 16:04SignalEditorial score6868

The author explores the LeRobot community dataset, using embeddings and FiftyOne to analyze 497 episodes across 50 robot types. They identify near-duplicates, filter sessions, and reveal cross-embodiment patterns like cloth folding. A curated subset is available for further exploration.