Skip to content
#Data and simulation
Nov 5Thu
  1. 未来机器人网 · 微信公众号77

    D-Robotics and University of Tokyo release RawVLA: frozen VLA, only imaging front end trained, low-light dual-arm task success rate rises from 0 to 66.5%

    D-Robotics, together with the University of Tokyo and other institutions, released the RawVLA technical approach, which adds a learnable neural ISP module before the VLA policy while keeping the VLA model frozen, using action error to train the imaging front end in reverse. On real robots, four dual-arm tasks were tested with π0.5; the average success rate in low light rose from 0% with the default ISP to 66.5%, versus 75.0% in normal light, where the default ISP achieved 76.5%.

    Why it matters: The approach brings the imaging front end into the learnable pipeline and provides comparative data under low light and normal light, offering a reference for vision pipeline design.

  2. 未来机器人网 · 微信公众号40

    Wang Xingxing at Wuzhen: 2027 likely the tipping point for embodied intelligence; humans' core value is setting goals and scoring for AI

    At the 2026 World Internet Conference Wuzhen Summit, Unitree founder Wang Xingxing judged that next year is likely the tipping point for the robotics industry or robot models, based on the rapid capability gains of foundation models such as GPT-6. He said the biggest current bottleneck is aligning models with real machines in the "last few millimeters" of precision assembly and flexible manipulation. Unitree is advancing a physical AI robot self-evolution system: AI agents autonomously search papers and open-source solutions to generate control code, which is validated in simulation and then deployed for real-machine testing, followed by closed-loop iteration combining automated evaluation and human scoring.

Nov 4Wed
  2. Guyuehome55

    Review: Three paths for embedding physical priors into robot learning, from classical mechanics to VLA and WAM

    The article reviews three paths for embedding physical priors into robot learning: data and representation, model architecture, and training loss, and emphasizes that where a prior is placed determines its constraining effect. Using grasping a soft object as an example, it explains the layered roles of kinematics, dynamics, contact, and material properties in the task, while noting that physical priors are not a panacea and need to match the task.

  4. Guyuehome58

    How DiffTaichi's differentiable physics simulation uses gradients to drive control optimization

    The article analyzes the adjoint gradient mechanism of DiffTaichi's differentiable physics simulation, using loss gradients to directly optimize actuator control sequences in material point method soft body, mass-spring, and fluid scenarios. It shows that gradient methods can optimize soft-body motion forms that reinforcement-learning-style sampling methods struggle to reach.

  5. Guyuehome60

    Reproducing PPO Reinforcement Learning-Based Quadrotor Racing Flight Control

    The author loads an open-source pretrained policy in Isaac Sim to reproduce quadrotor racing on a figure-8 track, matching the official result with 25 gates and 3.12 laps. The content breaks down the three attitude phases of high-speed turns and uses a three-seed controlled experiment to test the real benefits of domain randomization.

  6. 未来不远 · 微信公众号76

    Future Not Far releases Self-Evolving WAM self-evolving world-action model

    Future Not Far released the Self-Evolving WAM self-evolving world-action model, integrating joint world-action modeling, interactive experience learning, and training-effect feedback into a single system. In update attempts from round 0 to round 30, all four groups of household tasks showed gains in autonomous full-completion success rates: folding clothes rose from 54.6% to 96.8%, hanging clothes from 46.2% to 91.2%, making pasta from 42.7% to 89.5%, and stir-fried pork with chili peppers from 38.2% to 68.9%.

    Why it matters: The original text gives the success-rate changes for four groups of household tasks from round 0 to round 30 updates, showing the practical benefits of the self-evolving training loop.

  7. OFweek Robotics62

    Fei-Fei Li's World Labs launches world model Atlas, generating robot simulation environments from phone photos

    World Labs launched its next-generation world model Atlas on September 1. It is currently not open source and is available for early trials only to a small number of partners. Atlas takes 1 to 6 reference images and a camera trajectory as input, can generate videos up to 1 minute long at 1440p, and exports point clouds and 3D Gaussian splatting, used to reconstruct real scenes as robot simulation training environments. In the official demo, 24 frames taken from a phone video were used to reconstruct a warehouse-sized space, allowing a virtual robot to walk through it and generating camera images and depth data; however, there is no public data yet on complex physical behavior or large-scale operating costs.

  9. 量子位42

    Kinetix AI Showcases Full-Stack Matrix Including SMASH Ping-Pong Robot and 117-DOF Humanoid KAIBot at Hong Kong FinTech Week 2026

    Kinetix AI presented its full-stack embodied intelligence matrix at Hong Kong FinTech Week 2026, including the high-dynamic humanoid ping-pong system SMASH, the 117-DOF humanoid robot body KAIBot, the 37-DOF dexterous hand KAI Hand, and the first-person data collection headband KAI Halo.

  10. 锐曼REEMAN38

    Planning a multi-floor autonomous forklift project? REEMAN Ironhide elevator integration checklist

    From a supplier perspective, REEMAN provides a five-item checklist of mechanical prerequisites, communication architecture, and responsibility division for autonomous forklift and elevator integration. For a single elevator, a local LoRa direct connection to the elevator control module is used; for multiple elevators, MQTT is used with the fleet system queuing and dispatching. The elevator control module only coordinates calls and status and does not replace the elevator's native safety controls. In the Maoming factory case, Ironhide transports about 1,000 kg of wire rod per trip over a one-way distance of about 100 m, completing 65–85 cross-floor elevator tasks per day.

  11. Robotics & Automation News71

    LiberAI launches Liber-0 world action model, pretrained on 2.2 million hours of human experience

    LiberAI has released Liber-0, a world action model pretrained on more than 2.2 million hours of human experience, combining large-scale human video, hand motions recorded with data gloves, and robot demonstrations, with the goal of giving robots generalizable manipulation capabilities. In an eight-task evaluation, a policy trained with 25% robot demonstrations plus 100% glove demonstrations scored 75.6 on progress, higher than the 72.8 scored using only all robot demonstrations. In a 60-task zero-shot evaluation, Liber-0 achieved a 36% success rate.

  12. HuggingFace Daily Papers85

    PrimeSim: A Generalist World Model Simulator from Large-Scale Cross-Embodiment Demonstrations

    The research team proposes PrimeSim, an action-conditioned multiview world model simulator, midtrained on robot demonstration data covering 22 platforms and 30,110 hours, and has released the model weights, inference code, and training code.

    Why it matters: This work presents a world model simulator trained on demonstrations spanning 22 platforms and 30,000 hours, and releases the weights and training code, facilitating comparison of its long-horizon stability and action-following capability.

  13. 锐曼REEMAN40

    REEMAN autonomous forklift and stretch wrapper integration: task signals, pallet handover, and docking conditions

    REEMAN autonomous forklifts perform a task-level handshake with turntable stretch wrappers via an external wireless module: the forklift places the pallet and, after retreating from the rotating area, sends a start signal; the wrapper completes rotary film wrapping and automatic film cutting, then returns a completion signal, after which the forklift re-identifies and aligns to pick up the load.

  14. 雷科技56

    Robots aren't selling yet, but the business of selling training data is already booming

    Leikeji observed that companies providing collection, annotation, and training services for humanoid robot training data are emerging, with approaches including centralized training grounds, crowdsourced collection, and full-chain services. The report notes that bringing a single robot skill to a deliverable level may require 2,000 to 5,000 hours of training data. Mifeng Technology, under AgiBot, launched crowdsourced collection in September and had reached 40 cities by October, with participants earning 10 to 30 yuan per hour.

  17. HuggingFace Daily Papers72

    DreamFree: A world action model without future video generation, reusing video diffusion priors via representation distillation

    DreamFree proposes a world action model (WAM) that does not require generating future videos. Through representation distillation, it turns a pretrained video diffusion model into a predictive visual encoder that predicts the frozen video model's representation using only the current observation; an action module then extracts action-relevant information.

  19. 机器人前瞻 · 微信公众号72

    Tsinghua and Berkeley open-source embodied agent uni-robot-agent; Qwen3.8-Flash-Next scores above GPT-6 Astra on RoboDojo

    Tsinghua University and UC Berkeley jointly proposed and open-sourced uni-robot-agent, a cross-environment unified embodied agent harness for connecting tool interfaces across different robots and simulation environments. Based on this framework, Qwen3.8-Flash-Next scored 31.39 on the RoboDojo benchmark, exceeding the publicly reported 28.97 for GPT-6 Astra, at about 1/20 of Astra's cost.

  20. 图灵人工智能 · 微信公众号36

    Interview with Princeton's Liu Shilong: From Grounding DINO to self-evolving agents, humans are super routers

    Liu Shilong, a postdoctoral researcher at Princeton (joining Columbia University in fall 2027 as a tenure-track assistant professor), has shifted his research focus to self-evolving agents, with robotics and scientific discovery as application scenarios. He is a core author of the open-world detection model Grounding DINO, which has over 2 million monthly downloads on Hugging Face, and whose open-source project has accumulated over 30,000 GitHub stars.

Nov 3Tue
  2. 机器人产业应用 · 微信公众号72

    IROS 2026 Best Paper LT-Mem explained: using hierarchical spatiotemporal memory to solve robotic temporal forgetting

    The paper LT-Mem from the APRL lab at Daegu Gyeongbuk Institute of Science and Technology (DGIST) in South Korea won the IROS 2026 Best Paper award. The work proposes a volatile-perception spatiotemporal memory framework that uses three memory layers—Live, Delta, and Meta—and three volatility-based update strategies—coverage, retention, and multi-hypothesis—to solve the temporal forgetting problem in which new observations overwrite old memories when a robot revisits the same space multiple times.

  3. 新智元 · 微信公众号38

    How Tencent Cloud AI Infra Supports Embodied Intelligence: Manifold Space, Machine Science, and Zeroth's Compute, Storage, and Network Practices

    At the 2026 Tencent Global Digital Ecosystem Summit, Tencent Cloud showcased AI Infra deployments in embodied intelligence. Manifold Space used a cluster of 512 domestic chips to train stably, improving training performance nearly 3x in about two weeks. Machine Science built a unified data lake in Nanjing with COS, enabling cross-region direct reads of the same data between Beijing and Nanjing and saving tens of millions of yuan that would have gone to building high-performance storage in both locations. Zeroth used COS to handle PB-scale data across multiple campuses in Shanghai, North America, and Europe, with MetaInsight helping engineers quickly extract high-value segments.

  4. 人形大讲堂 · 微信公众号72

    Zhicheng AI completes funding of hundreds of millions of yuan, with state-owned capital jointly betting on JEPA physical world model

    Zhicheng AI announced the completion of a funding round of hundreds of millions of yuan, jointly invested by Hangzhou Capital, Fuzhe Technology, Caitong Capital, and Linglan Capital; the funds will be used for embodied intelligence and world model R&D and commercialization. The company was founded in March 2024 by Hu Luhui, former chief engineering lead at Meta. Its Chengling physical world model adopts the JEPA architecture, and version 0.1 was open-sourced under the MIT license in May 2026. The XR1 composite wheeled industrial robot has entered real production line operations at NIO's factory, while cooperation with Volkswagen remains in the scenario testing stage at the final assembly section.

  5. 深蓝AI · 微信公众号71

    Shanghai Jiao Tong University's Zou Danping team proposes DiffDreamer, combining world models and differentiable simulation for agile quadrotor control

    Zou Danping's team at Shanghai Jiao Tong University proposes DiffDreamer, which couples a learned world model with differentiable simulation for agile quadrotor control; the paper has been accepted to RAL 2026. The method uses the world model to predict external dynamics and differentiable simulation to propagate the robot's physical state. In aggressive tracking tasks, distance error drops to 0.381 m, versus 0.793 m for the strongest baseline; in dynamic gate crossing at gate speeds of 1.0/1.5/2.0 m/s, the number of gates passed is 16/12/8, versus 12/4/1 for MPC.

  6. Guyuehome58

    Reproducing Deep Reinforcement Learning Control for In-Hand Object Reorientation with the Revo3 Dexterous Hand

    This article reproduces three official pretrained policies for the Revo3 dexterous hand in Isaac Lab for in-hand object reorientation control. The cube random-target reorientation runs 8 times per episode, while the cylinder uses a 22.5° incremental chain, advancing once every 3.4 steps.

  8. Xbotics具身智能实验室 · 微信公众号78

    GE-Sim 2.0 and π0.5 closed-loop simulation test: Why does reward hacking occur in VLA training?

    The Xbotics Embodied Intelligence Lab built a closed-loop simulation pipeline using the AgiBot GE-Sim 2.0 world model and the OpenPI π0.5 policy. In testing, the world model generated about 4.33 frames per second, the pure-policy closed loop ran at about 3.20 frames per second, and after adding MiMo online scoring, the rate dropped to about 0.88 frames per second.

    Why it matters: The author uses a four-stage test to break down the frame-rate bottlenecks and causes of reward hacking in world-model closed-loop simulation, which can be transferred to the design of VLA evaluation workflows.

  9. 幻尔科技38

    SO-ARM101 vs. NexArm Open-Source 6-Axis LeRobot Arm Buying Guide: Which One to Choose

    Huan'er Technology compares two open-source 6-axis robotic arms that natively support the Hugging Face LeRobot ecosystem: the SO-ARM101 has a 200 g payload, a 400 mm reach, and a maximum servo torque of 30 kg·cm, making it suitable for beginners learning teleoperation and imitation learning; the NexArm has a 500 g payload, a 500 mm reach, and 65 kg·cm torque, features an aviation-grade metal body, and can be expanded with peripherals such as the K230 vision module and a Mecanum-wheel chassis.

  10. PsiBot/灵初智能58

    Lingchu Intelligence's embodied robot enters Li-Ning East China smart logistics center for trial operation

    Lingchu Intelligence and Hengyan Technology are jointly delivering the Li-Ning East China Smart Logistics Center project. The embodied robot, equipped with Lingchu's self-developed model, has been deployed on site and entered trial operation, taking on routine work and fulfilling commercial orders. The solution targets the bin-picking scenario in footwear and apparel warehousing, where shoeboxes are mixed in storage and item poses are random. The model can autonomously break down multi-step goals and complete recognition, pose estimation, dexterous grasping, product sorting, and pushing full boxes, without requiring large-scale modification of existing infrastructure. During the trial operation, Lingchu will work with Li-Ning's warehouse operations team on stress testing and validation of abnormal conditions, optimizing bin-picking success rates and stability during long periods of continuous operation.

  11. PsiBot/灵初智能71

    Lingchu Intelligence and collaborators propose FAFM, a continuous action generation method for heterogeneous-frequency data, accepted to NeurIPS 2026

    Lingchu Intelligence, together with Beihang University, Peking University, and The Chinese University of Hong Kong, proposed a frequency-aware flow matching method called FAFM; the paper was accepted to the NeurIPS 2026 main conference. The method uses the discrete cosine transform (DCT-II) to represent actions as frequency-domain coefficients corresponding to real time, performs flow matching in coefficient space and constrains the first-order time derivative, requires no additional learnable parameters, and can be used independently or integrated into VLA models such as π₀ and π₀.₅.

    Why it matters: The paper presents a frequency-domain action representation that adds no learnable parameters and can be integrated into existing VLA models, offering a transferable approach for joint training on heterogeneous-frequency data.

  12. PsiBot/灵初智能71

    PKU-Spirit AI Joint Lab's RoboFFT Accepted to CoRL2026, Uses Reinforcement Learning Post-Training to Generate Robotic Policies

    RoboFFT, proposed by the PKU-Spirit AI Joint Lab, has been accepted to CoRL2026. It is a reinforcement learning post-training method for generative robotic policies that reuses the forward noising process from diffusion model and flow model pretraining to construct an approximate probability ratio for PPO-style clipped updates.

    Why it matters: The method reuses the forward noising loss of diffusion and flow models to construct policy update signals, offering a reusable approach for post-training generative policies.

  13. 雷科技69

    Xiaomi wins two Wuzhen Summit awards; humanoid robot completes 8-hour validation at car factory workstation

    Xiaomi Group received the Outstanding Contribution Award at the 2026 World Internet Conference, and its "key technologies for a general-purpose robot manipulation foundation model for large-scale deployment" received the Leading Science and Technology Award. Xiaomi also disclosed that its humanoid robot took on an 8-hour shift for the first time at a center-console side-cover sequencing workstation in a car factory and achieved autonomous anomaly correction; an embodied foundation model based on Xiaomi-Robotics-1 began trial operation at Xiaomi Youth Apartments, completing long-horizon task validation of over 30 consecutive minutes, covering 3 rooms and 50+ tasks.

  15. Ken Goldberg30

    KEC combines Fable and Astra for parameter tuning to achieve real2sim

    KEC combines Fable with Astra to rapidly tune complex simulation parameters via RSS, including geometry, dynamics, deformation, materials, lighting, textures, and camera extrinsics, so that the simulation matches a given input video. This work targets the capability gap in agentic robotics for extending to new environments, new objects, and new tasks.

  16. Guyuehome40

    Deep Vision Reinforcement Learning for Autonomous Quadrotor Flight: AirGym Control Hierarchies and Depth Observation Design Overview

    For quadrotor obstacle avoidance, AirGym incorporates depth perception, low-level control, and sensor noise into a single training pipeline, and outlines five control hierarchies, depth observation designs, and reward structures. The article also records the engineering finding that Blackwell GPUs cannot run Isaac Gym.