Skip to content
Oct 28Wed
  1. HuggingFace Daily Papers66

    PDRVLA: A Vision-Language-Action Model with Physical Dynamics Representations

    PDRVLA introduces a lightweight Physical Dynamics Module that extracts physical tokens from vision-language features to condition action generation, without requiring online future-video generation. The module learns by directly predicting future RGB, depth, and optical-flow features, with auxiliary supervision from a frozen pretrained video world model. Training uses only public data, expanded through human-to-robot synthesis, subtask annotation, and embodied chain-of-thought (ECoT) annotation.

  2. NASA17

    Curiosity Mars rover Sols 5043-5049: Preparing to drill at 'Nazca'

    After completing the Basque Lakes drilling campaign, NASA's Curiosity Mars rover returned to a spot a few meters from the 'Chiu Chiu' rock block to prepare for its next drilling campaign at the newly named 'Nazca' target. Before the pre-drill payload test, the team scheduled MAHLI quantitative relief model observations and a ChemCam LIBS 150-shot laser depth profile experiment to characterize the 'disc-and-needle' crystal morphology. The plan also includes nighttime imaging of Phobos to estimate the size of Martian cloud particles.

  3. 钛媒体22

    Musk discusses AI control, xAI and the future of robotics on Isaacson's podcast The Builders

    In the first episode of Walter Isaacson's podcast The Builders, Musk acknowledged that humanity's desire to firmly control a superintelligence a million times smarter may be wishful thinking, and argued that AI should pursue truth and maintain curiosity to avoid logic conflicts spiraling out of control. He called AI and robotics the way out of America's debt predicament, said AI aimed at the physical world is one to two orders of magnitude more important than purely digital AI, and that once robots can manipulate atoms, the value of intelligence will grow exponentially. He also proposed that frontier AI companies communicate weekly and open new models to competitors for vulnerability testing one week before release.

  4. 钛媒体22

    Worth 280 billion yuan, DeepSeek founder Liang Wenfeng still rents in Beijing

    On October 27, 41-year-old DeepSeek founder Liang Wenfeng appeared for the first time on the 2026 Hurun Rich List, ranking seventh with 280 billion yuan in wealth; on the same day, it was reported that he has long rented housing in Beijing. DeepSeek launched its first funding round in April this year and completed settlement at the end of June, with a valuation exceeding 350 billion yuan; the second round's pre-money valuation was about 500 billion yuan, and a new round of about 100 billion yuan has been completed. The company also plans to submit an application for a STAR Market listing as early as the end of 2026, and Liang Wenfeng has recently held intensive contacts with embodied intelligence entrepreneurs, evaluating team formation or investment arrangements.

  5. OFweek Robotics36

    Today's robotics sector moves are more interesting than a limit-up

    On October 28, divergence in the robotics sector intensified: Unitree Robotics rose more than 10% in early trading before closing down more than 4%, while the underlying index of the robotics ETF rose more than 3%. Tesla's Q3 earnings report confirmed that the Optimus production line at the Fremont factory has started, with the first 10,000 to 20,000 units used for its own training, and deliveries to external customers targeted for around the end of 2027. Unitree's trailing price-to-earnings ratio is 465 times; it delivered more than 5,500 units in full-year 2025, mainly purchased by research institutions and universities. UBTech's humanoid robot sales in the first half of the year were 921 units, a decrease of about 10.9% from the second half of 2025.

  7. 36Kr Embodied Intelligence47

    PHYMI launches its first Physical Agent product, Sagi, positioned as an "action partner"

    PHYMI's first Physical Agent product, Sagi, has officially debuted, positioned as the company's first "action partner." The product carries self-developed intelligence, is designed for the open, dynamic real world, focuses on goal-oriented long-range mobility and manipulation, and collaborates with people around high-frequency daily needs such as pickup and delivery and carrying. The launch selected three scenarios—camping, office, and community collaboration—to show Sagi taking part in hauling equipment, delivering materials, and carrying and dispatching items.

  8. 钛媒体20

    2026 Hurun Rich List: Zhang Yiming at 670 billion yuan, Liang Wenfeng at 280 billion yuan, as AI reshapes the wealth rankings

    The 2026 Hurun Rich List, released on October 27, shows that ByteDance founder Zhang Yiming's wealth rose by 200 billion yuan to 670 billion yuan, making him China's richest person for the second time; DeepSeek founder Liang Wenfeng debuted on the list at 280 billion yuan, entering at seventh place. Entrepreneurs focused primarily on AI now account for 10.4% of those listed, nearly 80 robotics entrepreneurs made the list, and Unitree Robotics' Wang Xingxing debuted with 69.5 billion yuan.

  11. Yaskawa70

    Yaskawa Electric and GAZOO Racing to Bring AI Autonomous Welding Robot Technology to GR KART Manufacturing Site

    Yaskawa Electric and Toyota GAZOO Racing have jointly developed AI-driven autonomous robot production technology, with deployment beginning in September 2026 at the Gamagori GR KART workshop. The technology builds on the SFA method (Sequence Freezing Arc Welding) jointly developed by the two parties in 2024, and uses MOTOMAN NEXT robots equipped with GPUs and AI to autonomously recognize the state of pipe frames, complete welding, confirm and correct quality, and analyze appearance images, welding sounds, and various data in an integrated manner.

    Why it matters: Yaskawa and GAZOO Racing are extending their SFA welding experience into an AI autonomous welding line, and it has already gone into actual production at the GR KART workshop, offering a view of how welding automation moves from validation to the mass-production floor.

  12. 36Kr Embodied Intelligence60

    XPeng IRON Robot's Solid-State Battery Rolls Off Line in Small-Batch Production

    XPeng Group Vice President Gu Jie said the solid-state battery for the XPeng IRON robot has rolled off the line in small-batch production, and the robot is currently in the pre-SOP stage. At its 2025 Tech Day, XPeng announced that IRON will be the industry's first to use an all-solid-state battery, and the company plans to achieve large-scale production of high-level humanoid robots by the end of 2026, with the robots gradually entering XPeng stores and commercial scenarios worldwide from 2027 to handle shopping guidance, explanation, and other services.

  13. 芯东西73

    Google releases world model Genie 4, capable of simulating aerodynamics and collisions with native audio generation

    Google released its next-generation world model Genie 4 on October 28, capable of simulating physical processes such as aerodynamics and collisions, and natively generating music, ambient sound, and interactive dialogue. The model combines capabilities from the Gemini series, with Gemini 3.8 Flash breaking down multi-step tasks in a loop, and Google also demonstrated it combined with Gemini Robotics to simulate an Apptronik Apollo humanoid robot playing checkers.

  15. 雷科技35

    Qingtianzu appears at ROBOTECH 2026: exploring new paths for robot commercialization through scenario-based services

    At ROBOTECH 2026 in Shenzhen, Qingtianzu, as an official public service strategic partner, created a "Robot Carnival" and deployed a public service formation of 1 AgiBot Expedition A3 and 2 Lingxi X2 robots to handle visitor guidance and route direction. The event also gathered the AgiBot Expedition A3 and the Zhongqing T800 for humanoid robot boxing matches, with quadruped robots participating in the "Cool Dog Race," while small humanoid robots such as the Qiyuan Q1 and Gaoqing Xiaopai interacted on the same stage. The exhibition organizer paid to procure its solutions, and Qingtianzu says it is advancing normalized operations in cultural tourism scenarios through content productization and standardized delivery.

  16. Guyuehome58

    How Much Does TEB Differ from DWA? 18 Gazebo Mecanum-Wheel Robot Simulations Compared

    The author ran 18 experiments comparing the TEB and DWA planners in a Gazebo mecanum-wheel robot simulation, with the title stating a 1.65x difference between them. The author also notes that when the planners moved sideways throughout, their lateral-movement rates differed by 100x, yet this was completely invisible on the trajectory plots.

  18. arXiv cs.RO62

    SCRIM: Quantifying State Criticality in Imitation Learning

    The authors propose SCRIM, a state criticality measure that uses an adversarial perturbation policy and a value critic to estimate the worst-case degradation in downstream value caused by small perturbations to expert actions, scoring each demonstration state. These scores can be used to reweight behavior-cloning losses or to select compact, high-value subsets of demonstrations, reducing training data while maintaining performance. The method is evaluated on a variety of simulated and real robot manipulation tasks, and the paper was accepted to CoRL 2026.

  20. arXiv cs.RO62

    Goal2Pixel: A Vision-Language Navigation Method That Grounds Goals to Pixels

    Goal2Pixel proposes a pure pixel paradigm that reformulates vision-and-language navigation in continuous environments (VLN-CE) as navigable pixel grounding: the model predicts navigable pixels on the image plane, which are then back-projected into 3D waypoints for forward movement, while turning and stopping are encoded by auxiliary left, right, and bottom regions.

  21. arXiv cs.RO62

    AWARE: Action-Entropy-Weighted and Attention-Guided Retrieval of Observation Evidence for Vision-Language-Action Models

    The authors propose AWARE, a training-free, plug-and-play module for Vision-Language-Action (VLA) models that tracks layer-wise uncertainty via element-wise action entropy and retrieves observation evidence into the feed-forward networks' key-value memory through associative memory lookup.

  22. arXiv cs.RO60

    Semantic World Models: Using Vision-Language Models to Predict Semantic Information for Robot Planning

    The paper proposes Semantic World Models, which model the world model as visual question answering about the semantic information of future frames rather than reconstructing future pixels, so that vision-language models can be obtained through supervised fine-tuning on image-action-text data. The authors validated it on simulated robot tasks, where it outperformed typical reconstruction-based action-conditioned world models.

  23. arXiv cs.RO60

    QUAN: Quaternion Approximate Networks Improve Image Classification and Oriented Object Detection

    The paper proposes Quaternion Approximate Networks (QUAN), which use Hamilton product decomposition to approximate quaternion convolution with real-valued operations, paired with custom CUDA kernels to achieve rotation-equivariant image classification and object detection. The authors also propose Independent Quaternion Batch Normalization (IQBN) to stabilize training and extend quaternion operations to spatial attention mechanisms.

  24. arXiv cs.RO72

    BiDex-100: A Bimanual Dexterous Manipulation Benchmark Covering 100+ Simulation Tasks

    The authors propose BiDex-100, a simulation benchmark for bimanual dexterous manipulation comprising 106 tasks organized along five capability dimensions: generalization, instruction following, limited demonstrations, long horizon, and memory. Each task is paired with an out-of-domain variant featuring controlled visual and spatial changes for evaluating robustness. The benchmark supports two embodiments and provides 8,080 success-validated teleoperated demonstrations totaling 6.25 million frames. The authors evaluate seven recent vision-language-action models under a unified protocol, establishing baselines for capability-specific performance, robustness, and data efficiency.

  26. arXiv cs.RO69

    IDTD: Information-Decoupled Trajectory Design for Sim-to-Real System Identification

    The authors propose an information-decoupled trajectory design framework, IDTD, which constructs an exploration objective from a Schur complement score derived from the Fisher information matrix, then normalizes by parameter, selects the optimal trajectory segment for each parameter, and aggregates the scores logarithmically, to mitigate unreliable identification caused by indistinguishable parameters.

  28. arXiv cs.RO58

    Compact Force Sensor and Tension-Aware Outer-Loop Control for Dual-UAV Cable-Suspended Payload Transport

    This study proposes a dual-UAV payload transport framework that uses a custom compact force sensor at the cable anchor point to measure interaction forces, with static and dynamic tests evaluating its linearity, hysteresis, repeatability, and crossload. Control uses a cascade structure: inner loops handle vehicle stabilization, while outer loops compensate for the measured forces, validated in simulations and indoor experiments under position uncertainty. Payload-drop and constrained-space tests compared with literature-based distributed methods show improved stability, coordination, and disturbance rejection.

  29. arXiv cs.RO52

    When Is Test-Time Physical Diagnosis Effective? Six Ordered Empirical Conditions and a Failure Analysis of Frozen Policies

    Zhengshu Zhang proposes in an arXiv paper that for diagnosis to improve behavior when a robot faces unfamiliar physical conditions, six ordered empirical conditions must hold: a meaningful reference, identifiability of the physical condition, use of the acquired evidence, decision value, selection value relative to a fixed alternative, and safe realization.

  30. arXiv cs.RO50

    Constraining World-Model Planning with Legal Reasoning: Validation of a Defeasible Deontic Logic Planning Stack on a Simulated Robotic Arm

    The paper proposes a legal planning stack that uses Defeasible Deontic Logic (DDL) to constrain a motion planner, leveraging a learned world model for planning and to supply legal context, thereby enabling ex ante intervention before an illegal action is executed. The stack was deployed in a scenario where a simulated robotic arm pushes blocks on a 3x3 grid. Results show that the legally constrained agent complied with norms at a markedly higher rate than the unconstrained one, and that modeling perceptual uncertainty further raised the compliance rate; legal reasoning ran efficiently at runtime, its rulings were auditable, and the stack adapted to exogenous signals and endogenous rule changes.

  31. arXiv cs.RO65

    Gradient-Free Neural Hamilton-Jacobi Reachability for Scalable Safety-Critical Control

    The authors propose a discrete-time neural reachability framework that learns backward reachable tubes (BRTs) and backward reach-avoid tubes (BRATs) through Bellman-Isaacs value propagation, and uses gradient-free value probes to construct approximate teacher actions, turning adversarial actor learning into supervised policy learning.

  32. arXiv cs.RO66

    Learning-Based Informative Path Planning for Multi-Robot Systems: A Spatio-Temporal Gaussian Process Kalman Filter Framework

    The authors propose a grid-based spatio-temporal Gaussian process-Kalman filter framework for learning-based multi-robot informative path planning, representing anonymous target presence as a single latent variable field over a discrete workspace grid rather than maintaining one Gaussian process per target.

  33. arXiv cs.RO60

    Temporal Tactile Encoding and Compliant Control for Intent-Aware Robot-to-Human Bimanual Handover

    This study combines a VLA model with a compliant controller, fine-tuning the model using RGB observations, temporally encoded tactile feedback, and proprioception from human demonstrations to achieve robot-to-human bimanual handover. In a human-subject study, the system was compared against two baselines: one without tactile feedback and one with tactile feedback but no compliant control. Results show that compliant control and tactile history are complementary, with performance clearly outperforming the baselines. The project page is publicly available.

  34. arXiv cs.RO60

    From Language to Navigation Goals: A Vision-Language Approach for Semantic Navigation of Mobile Robots Using RGB-D Perception

    The study proposes a language-driven navigation framework that enables mobile robots to understand natural-language requests such as "go to the mailbox," identify the referenced object, estimate its position using RGB-D data, and generate a navigation goal that is then executed by the ROS 2 Nav2 navigation stack.

  35. arXiv cs.RO70

    From Grasping to Dexterous Manipulation: Large-Scale Grasp Pretraining for Dexterous Manipulation

    The research team proposes pretraining the low-level controller of a hierarchical imitation learning framework with large-scale dexterous grasp data, then fine-tuning it on downstream task demonstrations to support functional dexterity in articulated tool use. They built a 355k-trajectory grasp-pretraining dataset from large-scale dexterous grasp annotations and introduced DexCraft, a simulation benchmark with six articulated tool-use tasks. In simulation and real-world experiments, the method outperforms end-to-end diffusion policy baselines and hierarchical policies trained from scratch, improving real-world full-task success by 33.3 percentage points over DP3.

  36. arXiv cs.RO61

    LOPAL: Local Performance-Aware Active Learning from Imperfect Demonstrations

    LOPAL is a local performance-aware active learning method that uses a Gaussian Mixture Model (GMM) to jointly encode demonstration trajectories and their local quality assessments, thereby generating trajectories that outperform the imperfect demonstrations. The method also uses a shared autonomy (SA) mechanism to actively request user corrections in areas with insufficient data, while the robot autonomously executes its learned behavior during this process.

  37. arXiv cs.RO78

    FTP-1: A Generalist Foundation Tactile Policy Across Tactile Sensors

    The research team proposes FTP-1, a generalist foundation tactile policy pretrained across multiple tactile sensors and embodiments. It uses heterogeneous encoders to map image-, array-, and state-based tactile signals into unified morphology-aware latent tokens, which are jointly modeled by a shared tactile Transformer expert.

    Why it matters: This work unifies multiple tactile signals into a shared policy and reports changes in success rates for cross-sensor transfer, offering a reference for tactile manipulation research.

  38. arXiv cs.RO70

    FAWAM: Force-Aware World Action Model for Closed-Loop Contact-Rich Manipulation

    The authors propose FAWAM, a force-aware world action model that incorporates force information at three levels: perception, prediction, and closed-loop execution. It first encodes historical 6-axis force/torque signals to modulate action generation, then jointly predicts future actions and end-effector wrenches to explicitly model contact evolution, and uses the predicted wrench trajectory as an execution-time reference while correcting actions online with real-time force feedback. In real-world experiments across multiple contact-rich tasks, FAWAM achieved an average success rate 36.25% higher than vision-only baselines and 21.25% higher than existing force-aware baselines.

  39. arXiv cs.RO64

    GTP-FA: A Two-Stage Grasp-Then-Plan Manipulation Framework with Failure Attribution

    The authors propose GTP-FA, a two-stage "grasp-then-plan" framework for robotic manipulation: it first generates grasp candidates, then performs downstream motion planning conditioned on the selected grasp. The framework learns a failure attribution model that outputs a distribution over failure modes for failed trajectories, and uses it to separately optimize the grasp scoring and planning modules. Simulation and real-robot experiments show that GTP-FA improves overall task success rates over baselines including RL, IL, diffusion policies, and VLA, and the paper has been accepted to NeurIPS 2026.

  40. arXiv cs.RO71

    RePO-VLA: Recovery-Driven Policy Optimization for Vision-Language-Action Models

    RePO-VLA proposes a recovery-driven policy optimization framework that assigns distinct roles to success, recovery, and failure trajectories for long-horizon, contact-rich manipulation tasks. The method first uses recovery-aware initialization to slice corrective segments into new episodes, then learns a progress-aware semantic value function, and uses value-conditioned refinement to make the policy prefer high-progress actions.