Robotics coverage relating to imitation learning.
Before humanoid robots mature, Physical AI is attempting to enter industrial scenarios still reliant on manual operations by deploying on mature robotic arms. Companies like Sanctuary AI and Skild are integrating AI models into robotic arms from different manufacturers to perform complex tasks such as wire insertion.
This article focuses on the content of Huaqianqian's 2027 Embodied Intelligence New Product Launch Event, combined with the author's hands-on experience in embodied intelligence projects, to break down the technical architecture, learning path, application scenarios, and common issues in development. Starting from the three-layer architecture of perception-decision-execution, the article explores the role of large models in embodied intelligence, analyzes key challenges such as the transition from simulation to real-world deployment, model inference latency optimization, and sensor calibration, and provides a complete learning path from Python and ROS 2入门 to the development of embodied intelligence agents.
This paper presents a comprehensive review of diffusion models in robotics, focusing on their application in robot learning, data scaling, reinforcement learning, and imitation learning. The authors conducted a systematic literature review, selecting peer-reviewed articles and high-impact preprints to analyze the current state of diffusion model research in robotics. The paper discusses various applications of diffusion models, including semantic-level data augmentation, cross-view and morphology synthesis, automated simulation asset and task generation, trajectory optimization, policy representation, hierarchical planning, and real-time deployment. It also addresses the challenges and limitations of current approaches, such as computational cost, inference latency, and data efficiency, while highlighting promising future directions for research and development in the field.
The source reports AMD’s acquisition of World Labs, co-founded by Fei-Fei Li, for approximately ¥55 billion, and Megmeet’s proposed ¥664 million acquisition of a welding subsidiary. It also covers funding and listing activity among Chinese robotics companies. These are the publisher’s reports, not independently verified transactions.
This paper presents LeaP, a learnable source prior for generative robot policies, which replaces fixed standard Gaussian distributions with state-dependent random initialization. LeaP jointly learns the mean and variance of an initial Gaussian distribution, improving action generation in robotic tasks. The method achieves an average success rate of 81.6% in simulation and 80.0% on real robotic arms, outperforming existing baselines like A2A and VITA. The paper is accepted at CoRL 2026, and the code is open-sourced.
Skild AI introduces their omni-bodied robotics foundation model, the Skild Brain, which is trained across diverse morphologies and data sources to enable generalization across tasks and hardware. The model uses a hierarchical architecture with low- and high-frequency policies, and is pre-trained using large-scale simulation and internet video data, with post-training on real-world data. This model represents a significant step toward creating a general-purpose robotic brain capable of operating on various robot types.
Professor Shi Qing's team from Beijing Institute of Technology was inspired by the high-speed running mechanism of the elephant shrew. They designed a micro quadruped robot named FLEXOR equipped with a dual-joint coupled spine. Through dynamic spine-leg coordination, the robot achieved a 31.6% increase in speed and a 32.2% reduction in energy consumption without upgrading its motors. The study reveals the critical role of the timing coordination between the spine and legs in locomotion performance and verifies the universality of this mechanism across different robot morphologies, providing new insights for highly mobile embodied intelligent robots.
The Stanford team's HomeBody project demonstrates how GPT-6 Astra controls the Unitree G1 robot to complete kitchen cleaning tasks through a three-step process: spatial exploration, digital twin simulation, and task execution. The system uses pre-trained models and a skill library to perform complex operations in new environments without additional training.
The article analyzes China's leading position in the field of humanoid robot patents, noting that as of the first half of 2026, China holds 60.8% of relevant patents, significantly surpassing the United States and South Korea. The article examines the evolution of patent strategies, the trend of technological shift from hardware to software, and the progress made by Chinese companies in patent quality, application scenarios, and global layout. It also points out that patent quantity does not equate to comprehensive technological leadership, emphasizing the challenges of manufacturing costs and quality, as well as the dual role of patents in industrial competition.
China is pursuing a strategic roadmap through its Five-Year Plans to dominate the global robotics industry, focusing on embodied AI and humanoid robotics. The 13th and 14th FYPs laid the foundation for industrial automation, while the 15th FYP emphasizes embodied intelligence, with state support for R&D and mass deployment of humanoids. Despite achievements in market share and production scale, challenges remain in precision hardware and semiconductor self-sufficiency, with risks of overcapacity and market volatility.
Robotic Systems Lab@leggedroboticsSignalEditorial score6363
This article explores how tactile data is helping robots improve their dexterous manipulation skills. Researchers are creating large tactile datasets and models that can use tactile feedback to enhance robot performance in tasks like folding laundry or turning keys. Challenges include the difficulty of integrating tactile data with vision-based models and the need for more diverse and scalable datasets to achieve significant improvements in robot dexterity.
Researchers at the University of Queensland have developed 'Paraborgs'—cyborg cockroaches equipped with needles and cameras to assist in search-and-rescue missions. These insects, grafted with electronic interfaces, can navigate rubble, deliver drugs, and operate in teams. The study addresses challenges in autonomy, communication, and human interaction with the technology.
Google DeepMind has released Gemini Robotics 2, a new model that enables robots to perform complex tasks with whole-body intelligence, advanced dexterity, and multi-robot collaboration. The model supports real-time reasoning, on-device adaptation, and safety features for human-robot interaction.
NVIDIA has released an open-source, GPU-accelerated Medical Physics Simulation framework as part of its Isaac for Healthcare platform. This tool enables developers to model anatomy-device interactions, generate complex scenarios, and train robot policies in simulation, significantly reducing development time and improving regulatory readiness.
Engineers at MIT and EPFL have developed a flapping-wing robot, or FAAV, that can swim underwater and then flap into the air, mimicking the movement of diving birds. The robot, weighing less than 300 grams, can transition between water and air using optimized wing size, flapping frequency, and tail angle. The study, published in Science, could enable new aerial-aquatic drones for oceanographic research and environmental monitoring.
NVIDIA has released a cross-embodiment foundation model for humanoid robots, leveraging a Cosmos-Reason2/Qwen3-VL backbone and a flow-matching action transformer. The model is trained using LeRobot and is available for training and deployment under the Apache 2.0 license.
NVIDIA has released a cross-embodiment foundation model for humanoid robots, leveraging a Cosmos-Reason2/Qwen3-VL backbone and a flow-matching action transformer. The model is trained using LeRobot and is available for training and deployment via Hugging Face.
NVIDIA has released the 'gr00t17-lerobot-libero_object-640' model, a cross-embodiment foundation model for humanoid robot reasoning and skills. It uses a Cosmos-Reason2/Qwen3-VL backbone and a flow-matching action transformer, trained with LeRobot and available under the Apache 2.0 license. The model supports training and deployment through LeRobot, with detailed instructions provided.
NVIDIA has released a cross-embodiment foundation model for humanoid robots, leveraging a Cosmos-Reason2/Qwen3-VL backbone and a flow-matching action transformer. The model is trained using LeRobot and is available for training and deployment via the Hugging Face Hub.