Physical AI 前沿雷达

AutoBio+: Benchmarking Dexterous Manipulation with Dynamic Experimental Phenomena in Realistic Laboratories

2026-10-12T23:40:00Z

主线: 仿真与评测（sim2real / 闭环）
证据: [R]
来源: arXiv 2610.13643 · 2026-10-09
关键数字: 14 tasks
信号: 含闭环成功率 · 含真机实验

Automating biological laboratories requires robots to perform dexterous tool interactions under realistic visual conditions while handling dynamic experimental phenomena, presenting significant challenges for robotic manipulation.

2026-10-13 雷达全文

Forward Dynamics Models

2026-10-12T23:40:00Z

主线: 基座模型（VLA / WAM）
证据: [R]
来源: arXiv 2610.13117 · 2026-10-09
关键数字: 99.4% · 13%
信号: 含量化指标 · 含闭环成功率 · 含真机实验

Vision-language-action (VLA) models and world action models (WAMs) have advanced robotic policy learning by building on pretrained vision-language models (VLMs) and video generative models (VGMs), respectively.

2026-10-13 雷达全文

Evo-RL: Iterative Robot Reinforcement Learning from Selective Experience

2026-10-12T23:40:00Z

主线: 训练与自我改进（RL / 后训练）
证据: [R]
来源: arXiv 2610.13345 · 2026-10-09
关键数字: 1,000 hours
信号: 含量化指标 · 含闭环成功率 · 含真机实验 · 有开放代码或权重

Existing real-world reinforcement learning methods improve robot policies by combining human corrections with autonomous rollouts.

2026-10-13 雷达全文

3D Gaussian Splatting in practice: Industry use cases

2026-10-12T23:40:00Z

主线: 仿真与评测（sim2real / 闭环）
证据: [O]
来源: AWS Physical AI Blog · 2026-10-06

Gaussian Splatting is dramatically lowering the cost and complexity of photorealistic 3D, making once impractical use cases like near real time capture, robotics, and simulation viable across media, construction, manufacturing, and real…

2026-10-13 雷达全文

Closing the reality gap: Why real-time execution has become the new competitive advantage

2026-10-12T23:40:00Z

主线: 边缘与实时性（端侧推理 / 控制频率）
证据: [M]
来源: The Robot Report · 2026-10-12
信号: 含延迟或控制频率

Daifuku's Cecile Dick-Calmes said success is no longer measured solely by automation density or equipment utilization.

2026-10-13 雷达全文

费莫一刘念邱：智驾无图的经验，也能迁移到机器人 | 物理AI 50人

2026-10-12T23:40:00Z

主线: 本体与供应链（硬件 / 成本）
证据: [M]
来源: 雷峰网 Leiphone · 2026-10-12

刘念邱很难被归进典型的具身创业者模板里。他不是从机器人实验室出来的学者，也不是过去两年才进入 Physical AI 的连续创业者。

2026-10-13 雷达全文

产业出题，科技答题：2026 IEIC国际具身智能技能大赛，用真实岗位任务考核机器人落地能力

2026-10-12T23:40:00Z

主线: 数据引擎（采集 / 清洗 / scaling）
证据: [M]
来源: 雷峰网 Leiphone · 2026-10-12

实景比拼赋能科创，助力人形机器人技术规模化落地！编辑丨李希为深入贯彻落实工业和信息化部、国务院国资委《 2026年度人形机器人与具身智能实景实训专项行动》部署，加快推动我国机器人与具身智能技术创新突破与规模化落地，国际具身智能技能大赛检阅赛将于2026年10月23日至25日在上海举行。

2026-10-13 雷达全文

百度世界2026将于11月12日在北京举办，主题为「下一幕，智能组织」

2026-10-12T23:40:00Z

主线: 训练与自我改进（RL / 后训练）
证据: [M]
来源: 雷峰网 Leiphone · 2026-10-12

10月12日，百度官方宣布，百度世界2026将于11月12日在北京·国家会议中心二期举办。目前，大会官网已开启售票通道（https://baiduworld2026.events.net.cn/tickets）。

2026-10-13 雷达全文

LiteNWM: Efficient Latent World Models for Onboard Visual Navigation in the Wild

2026-10-11T23:40:00Z

主线: 仿真与评测（sim2real / 闭环）
证据: [R]
来源: arXiv 2610.12368 · 2026-10-08
关键数字: 17.56% · 16.2% · 43.3% · 83.3%
信号: 含量化指标 · 含闭环成功率 · 含真机实验

Direct visual navigation policies generate trajectories efficiently but do not explicitly evaluate their future consequences.

2026-10-12 雷达全文

RoboAware: Learning to Coordinate Embodied Skills from Counterfactual Outcomes

2026-10-11T23:40:00Z

主线: 仿真与评测（sim2real / 闭环）
证据: [R]
来源: arXiv 2610.11480 · 2026-10-08
关键数字: 100 tasks · 77.0% · 90.0% · 73.8%
信号: 含量化指标 · 含闭环成功率

Embodied coding agents can combine modular robot skills with frozen end-to-end policies, yet effective composition requires anticipating which policy family will succeed in the current physical state.

2026-10-12 雷达全文

Generative Neural Retargeting for Human-to-Robot Dexterous Manipulation

2026-10-11T23:40:00Z

主线: 数据引擎（采集 / 清洗 / scaling）
证据: [R]
来源: arXiv 2610.12440 · 2026-10-08
信号: 含闭环成功率

Human demonstrations are a scalable data source for learning dexterous manipulation, but the embodiment gap prevents human motion from being executed directly on robots.

2026-10-12 雷达全文

LeWAM: A JEPA World Action Model with Diffusion-Steering-Based MPC

2026-10-11T23:40:00Z

主线: 基座模型（VLA / WAM）
证据: [R]
来源: arXiv 2610.12407 · 2026-10-08
信号: 含闭环成功率

World action models (WAMs) predict actions and future observations, typically from a reconstruction-based representation that carries noisy, redundant information which can complicate downstream predictions.

2026-10-12 雷达全文

SpatialHarness: Test-Time Spatial Scaffolding for Fine Robotic Manipulation

2026-10-11T23:40:00Z

主线: 训练与自我改进（RL / 后训练）
证据: [R]
来源: arXiv 2610.12457 · 2026-10-08
关键数字: 26.7% · 66.7% · 0% · 100%
信号: 含量化指标 · 含闭环成功率 · 含真机实验

Frontier multimodal foundation models (e.g., GPT-6 Astra) have recently shown strong potential for direct robotic control, yet their performance on fine manipulation remains limited.

2026-10-12 雷达全文

Real-Time Motion Planning with Dynamic Hazards: Classical vs. Learning-Based Methods

2026-10-11T23:40:00Z

主线: 边缘与实时性（端侧推理 / 控制频率）
证据: [R]
来源: arXiv 2610.12249 · 2026-10-08
信号: 含闭环成功率 · 含延迟或控制频率

We study real-time motion planning in dynamic hazard fields through a controlled comparison between classical planning and learning-based methods.

2026-10-12 雷达全文

WAND: Learning Robust Navigation under Complex Wind Disturbances and Dense Obstacles for Quadrotors

2026-10-11T23:40:00Z

主线: 边缘与实时性（端侧推理 / 控制频率）
证据: [R]
来源: arXiv 2610.11809 · 2026-10-08
信号: 含闭环成功率 · 含延迟或控制频率

Robust navigation in cluttered environments remains a fundamental challenge for quadrotors, particularly when strong wind disturbances arise, which perturb vehicle dynamics, limit control authority, and substantially increase collision…

2026-10-12 雷达全文

CSF: Contextual Safety Filtering for Motion Generators

2026-10-11T23:40:00Z

主线: 安全、权限与合规
证据: [R]
来源: arXiv 2610.12467 · 2026-10-08
关键数字: 90% · 100%
信号: 含量化指标 · 含真机实验

Text-conditioned motion generators produce trackable whole-body motion, but they have no notion of scene-dependent safety: the same action may target an object or a person.

2026-10-12 雷达全文

RESETTLE: Robotic Recovery through Disagreement-Triggered Retrieval and Efficient Corrective Control

2026-10-10T23:40:00Z

主线: 仿真与评测（sim2real / 闭环）
证据: [R]
来源: arXiv 2610.12185 · 2026-10-08
关键数字: 8.70% · 6.28% · 6.83% · 74.04%
信号: 含量化指标 · 含真机实验 · 有开放代码或权重 · 含延迟或控制频率

Reliable robotic manipulation requires timely intervention to correct emerging deviations and restore progress after execution errors.

2026-10-11 雷达全文

Being-M0.7: A Latent World-Action Model for Humanoid Robots

2026-10-10T23:40:00Z

主线: 本体与供应链（硬件 / 成本）
证据: [R]
来源: arXiv 2610.11283 · 2026-10-08
关键数字: 10,000 hours
信号: 含量化指标 · 含闭环成功率 · 含真机实验

Humanoid loco-manipulation requires coordinated locomotion and manipulation informed by future scene evolution and whole-body motion, yet learning these capabilities is constrained by scarce robot demonstrations.

2026-10-11 雷达全文

Residual Modeling Closes the Regression and Generative Policy Gap in Robot Learning

2026-10-10T23:40:00Z

主线: 仿真与评测（sim2real / 闭环）
证据: [R]
来源: arXiv 2610.12231 · 2026-10-08
信号: 含闭环成功率 · 含真机实验

Learning from demonstration has enabled impressive robot behaviors. A common choice for policy learning is to use diffusion or flow matching (Flow-Policies), which often outperforms direct action regression trained with mean squared error…

2026-10-11 雷达全文

ARC: A Reasoning Recipe for Robot Foundation Models

2026-10-10T23:40:00Z

主线: 基座模型（VLA / WAM）
证据: [R]
来源: arXiv 2610.12386 · 2026-10-08
信号: 含闭环成功率 · 含真机实验

The prevailing approach to improving robot foundation models (RFMs) relies on larger models, more robot demonstrations, and costly training at scale.

2026-10-11 雷达全文

SkillWeave: Weaving Heterogeneous Demonstrations into Long-Horizon Manipulation Skills

2026-10-10T23:40:00Z

主线: 数据引擎（采集 / 清洗 / scaling）
证据: [R]
来源: arXiv 2610.12046 · 2026-10-08
关键数字: 27% · 65% · 87%
信号: 含量化指标 · 含闭环成功率 · 含真机实验

Dexterous manipulation requires both large-scale task progression and precise contact-rich interaction, making it challenging to collect demonstrations that effectively support both regimes.

2026-10-11 雷达全文

FAITH: Feasibility-Aware Safety-Filtered RL for High-Dimensional Systems

2026-10-10T23:40:00Z

主线: 安全、权限与合规
证据: [R]
来源: arXiv 2610.12432 · 2026-10-08
关键数字: 99.95% · 97%
信号: 含量化指标 · 含真机实验

Safe reinforcement learning commonly places safety and task performance in the same policy objective, where they can introduce competing updates.

2026-10-11 雷达全文

Rewiring Semantics, Dynamics, and Control: A Simple yet Effective Action-Centric Tri-Stream Transformer

2026-10-10T23:40:00Z

主线: 基座模型（VLA / WAM）
证据: [R]
来源: arXiv 2610.11416 · 2026-10-08
信号: 含真机实验

Vision-Language-Action (VLA) models have emerged as a prominent framework for complex robotic manipulation, building on the strong semantic understanding of pretrained Vision-Language Models (VLMs).

2026-10-11 雷达全文

Why humanoid robot demos still fail the generalization test

2026-10-10T23:40:00Z

主线: 本体与供应链（硬件 / 成本）
证据: [M]
来源: The Robot Report · 2026-10-09
信号: 含真机实验

How better data and human judgment can help robots move beyond demos toward reliable real-world performance.

2026-10-11 雷达全文

RoboRender: Robot-Oriented Video Generation for Visual Sim-to-Real Transfer

2026-10-10T07:15:00Z

主线: 仿真与评测（sim2real / 闭环）
证据: [R]
来源: arXiv 2610.09254 · 2026-10-07
关键数字: 71% · 7.1x · 3.6x
信号: 含量化指标 · 含闭环成功率 · 含真机实验

Simulation enables large-scale, low-cost robot data generation, but policies trained in simulation often fail to transfer to the real world due to the sim-to-real visual discrepancies.

2026-10-10 雷达全文

USDCraft: Geometrically Grounded Programmatic Modeling of Articulated 3D Assets for Simulation

2026-10-10T07:15:00Z

主线: 仿真与评测（sim2real / 闭环）
证据: [R]
来源: arXiv 2610.11322 · 2026-10-08
信号: 含真机实验

Geometrically faithful and functional articulated 3D assets are essential for real-to-sim robot manipulation, where policies trained in simulation must transfer to physical objects.

2026-10-10 雷达全文

Humanoid World Action Model With Joint State--Action Generation

2026-10-10T07:15:00Z

主线: 基座模型（VLA / WAM）
证据: [R]
来源: arXiv 2610.12026 · 2026-10-08
关键数字: 70.6% · 43.3%
信号: 含量化指标 · 含闭环成功率

Humanoid robots are a promising platform for general-purpose manipulation. Recent Vision-Language-Action (VLA) policies learn actions directly from multimodal observations, while World Action Models (WAMs) further incorporate future…

2026-10-10 雷达全文

ManiUnit: A Manipulation Skill Dataset and Benchmark for Long-Horizon Tasks

2026-10-10T07:15:00Z

主线: 系统与编排（长程 / 多策略）
证据: [R]
来源: arXiv 2610.12089 · 2026-10-08
关键数字: 56% · 78.7% · 49.3% · 4.0%
信号: 含量化指标 · 含闭环成功率

Long-horizon mobile manipulation requires a robot to navigate multi-room environments and execute a sequence of manipulation skills under a single natural language instruction.

2026-10-10 雷达全文

Video Friday: Robot Decommissioning Takes a Fun Turn

2026-10-10T07:15:00Z

主线: 本体与供应链（硬件 / 成本）
证据: [M]
来源: IEEE Spectrum Robotics · 2026-10-09

It’s possible that I wasn’t the only one who thought that Figure’s recent stunt was somewhat ridiculous. Devanthro is in fact selling these plastic balls for $1000 each, and they’ll arrive in time for Christmas.

2026-10-10 雷达全文

让触觉像视觉一样可规模化，哥大李昀烛团队 FlexiTac 的开源实践 | IROS 2026

2026-10-10T07:15:00Z

主线: 本体与供应链（硬件 / 成本）
证据: [M]
来源: 雷峰网 Leiphone · 2026-10-10

FlexiTac压阻式传感器可兼顾柔性、低成本与一致性。机器人的视觉能力正快速增强，但触觉仍然是具身智能系统中的薄弱环节。

2026-10-10 雷达全文

这届 IROS 2026 上，具身学术与产业边界正在消融

2026-10-10T07:15:00Z

主线: 数据引擎（采集 / 清洗 / scaling）
证据: [M]
来源: 雷峰网 Leiphone · 2026-10-10

模型、数据、灵巧操作都在工程化，顶会上具身传递出的信号变了9 月 27 日到 10 月 1 日，全球机器人领域最顶级的国际会议之一 IROS 2026 将在美国匹兹堡开幕。如果你前两年也在关注这个会，你会发现那时候大家都在 “卷概念”。

2026-10-10 雷达全文

再裁员5000人！保时捷明确回归燃油车，高端车型计划涨价20%；腾讯拟发行50亿美元离岸债券，或筹钱加码AI；OpenAI全面上线GPT-6

2026-10-10T07:15:00Z

主线: 基座模型（VLA / WAM）
证据: [M]
来源: 雷峰网 Leiphone · 2026-10-09

要闻提示再裁员5000人！保时捷明确回归燃油车，高端车型计划涨价20% 阿里巴巴蔡崇信：开源是欧洲发展 AI 的最佳选择腾讯拟发行50亿美元离岸债券，或筹钱加码AI 超33亿元！

2026-10-10 雷达全文

WARP-VLA: Wrist-Camera Adaptation for View-Robust Policy Execution in Vision-Language-Action Models

2026-10-09T07:28:00Z

主线: 仿真与评测（sim2real / 闭环）
证据: [R]
来源: arXiv 2610.11508 · 2026-10-08
关键数字: 39.2% · 78.3%
信号: 含量化指标 · 含闭环成功率 · 含真机实验 · 有开放代码或权重

Despite recent advances in Vision-Language-Action models (VLAs) for robotic manipulation, their performance remains sensitive to changes in camera configuration.

2026-10-09 雷达全文

REACT: Rolling Denoising and Dual Decoupling for Reactive Robot Control with VLA Models

2026-10-09T07:28:00Z

主线: 仿真与评测（sim2real / 闭环）
证据: [R]
来源: arXiv 2610.12007 · 2026-10-08
信号: 含闭环成功率 · 含真机实验 · 含延迟或控制频率

Flow-based vision-language-action (VLA) models generate action chunks for temporally coherent robot motion, but chunked control creates a fundamental closed-loop trade-off: long chunks provide smooth execution, whereas frequent replanning…

2026-10-09 雷达全文

Leveraging Human-In-The-Loop Demonstrations in Reinforcement Learning for Digital Twin-Driven Robot Flexibility

2026-10-09T07:28:00Z

主线: 数据引擎（采集 / 清洗 / scaling）
证据: [R]
来源: arXiv 2610.12140 · 2026-10-08
关键数字: 100% · 17%
信号: 含量化指标 · 含闭环成功率 · 含真机实验

Growing automation makes collaborative robots work in more variable environments, increasing the need for adaptation.

2026-10-09 雷达全文

Tell Robot What Not to Do: A Negation Understanding Perspective

2026-10-09T07:28:00Z

主线: 基座模型（VLA / WAM）
证据: [R]
来源: arXiv 2610.11952 · 2026-10-08
关键数字: 2.60% · 88.45% · 12.4% · 88.8%
信号: 含量化指标 · 含闭环成功率 · 含真机实验

Instruction following enables robots to perform diverse tasks specified in natural language, making it a fundamental capability for human-robot interaction.

2026-10-09 雷达全文

This Disembodied Hand Is All the Robot You Need

2026-10-09T07:28:00Z

主线: 本体与供应链（硬件 / 成本）
证据: [M]
来源: IEEE Spectrum Robotics · 2026-10-08

A hand attached to the end of an arm is a given for most humans, but what if robots didn’t have to play by the same rules as we do?

2026-10-09 雷达全文

Jabil discusses the pace of humanoid robot development and production

2026-10-09T07:28:00Z

主线: 本体与供应链（硬件 / 成本）
证据: [M]
来源: The Robot Report · 2026-10-08

Jabil, which is working with robot developers such as Apptronik, provides a design perspective on humanoid progress.

2026-10-09 雷达全文

AWS launches open-source Physical AI Toolchain for robotics

2026-10-09T07:28:00Z

主线: 边缘与实时性（端侧推理 / 控制频率）
证据: [M]
来源: The Robot Report · 2026-10-08
信号: 有开放代码或权重

The toolchain combines AWS and NVIDIA tools for synthetic data generation, model training, simulation and validation, edge deployment, and continuous improvement.

2026-10-09 雷达全文

独家 | 清华 AIR 首届博士李健雄创立本溯智能：五源资本领投，押注动作原生具身模型

2026-10-09T07:28:00Z

主线: 训练与自我改进（RL / 后训练）
证据: [M]
来源: 雷峰网 Leiphone · 2026-10-08

常春藤、线性、华山资本跟投，AIR副教授詹仙园任首席科学家。雷峰网·鲸犀独家获悉，清华大学智能产业研究院（AIR）首届博士生李健雄已毕业，并创立具身智能公司本溯智能（BASAL Intelligence，简称 BASAL）。

2026-10-09 雷达全文

Agentic RSR: Real-to-Sim-to-Real through Scene Reconstruction and Execution-Grounded Robot Policies

2026-10-08T07:30:00Z

主线: 仿真与评测（sim2real / 闭环）
证据: [R]
来源: arXiv 2610.10479 · 2026-10-07
关键数字: 80%
信号: 含量化指标 · 含闭环成功率 · 含真机实验

A simulation of a real robot workspace must preserve task-relevant interactions, while policies developed in it must operate on observations available to the real robot.

2026-10-08 雷达全文

RealtimeWAM: How Fast Can I Run My World Action Model?

2026-10-08T07:30:00Z

主线: 仿真与评测（sim2real / 闭环）
证据: [R]
来源: arXiv 2610.10079 · 2026-10-07
关键数字: 63.09 ms · 82.75% · 87.41%
信号: 含量化指标 · 含闭环成功率 · 含真机实验 · 含延迟或控制频率

World Action Models (WAMs) combine visual dynamics modeling with action generation, but their high inference latency limits responsive robot control.

2026-10-08 雷达全文

RFPO: Rectified Flow Policy Optimization for Embodied Control

2026-10-08T07:30:00Z

主线: 边缘与实时性（端侧推理 / 控制频率）
证据: [R]
来源: arXiv 2610.10453 · 2026-10-07
关键数字: 2.4% · 98.5% · 4.39 ms · 0.08 ms
信号: 含量化指标 · 含真机实验 · 有开放代码或权重 · 含延迟或控制频率

Flow-based policies provide an expressive framework for continuous robot control, but their iterative ODE integration incurs substantial inference cost.

2026-10-08 雷达全文

PEARS: Physical-Prior-Guided Efficient Adaptation via Failure Reasoning and Diffusion Steering for Tactile Manipulation

2026-10-08T07:30:00Z

主线: 训练与自我改进（RL / 后训练）
证据: [R]
来源: arXiv 2610.08784 · 2026-10-06
关键数字: 53.2% · 95% · 90%
信号: 含量化指标 · 含闭环成功率 · 含真机实验

Pretrained robotic policies can suffer substantial performance degradation under out-of-distribution (OOD) conditions encountered during deployment, motivating post-training through real-world interaction.

2026-10-08 雷达全文

研究了1933篇 IROS 论文，我们看到了机器人学的六项新变化

2026-10-08T07:30:00Z

主线: 本体与供应链（硬件 / 成本）
证据: [M]
来源: 雷峰网 Leiphone · 2026-10-08

大模型并没有“吃掉”机器人学。9月27日，IROS 2026 将在匹兹堡开幕。

2026-10-08 雷达全文

对话如祺出行COO韩锋：世界模型越强，真实数据才是真壁垒

2026-10-08T07:30:00Z

主线: 数据引擎（采集 / 清洗 / scaling）
证据: [M]
来源: 雷峰网 Leiphone · 2026-10-08

过去一年，“物理AI”几乎成了自动驾驶、车企共同争夺的下一张船票。世界模型、具身大模型、机器人数据……众多公司从不同切入口布局物理AI，试图成为新风口之下的主角，相比之下，如祺出行选择了一条不那么显眼的路径，不自己训练大模型，而是尝试成为模型背后的数据基础设施。

2026-10-08 雷达全文

IROS 2026大奖揭晓：长期记忆拿下最佳论文，人形机器人打网球、端托盘双双获奖

2026-10-08T07:30:00Z

主线: 基座模型（VLA / WAM）
证据: [M]
来源: 雷峰网 Leiphone · 2026-10-08

在机器人领域，一个更大的模型并不能自动解决所有问题。当地时间 9 月 30 日，IROS 2026 主会在美国匹兹堡落下帷幕，本届大会的主要奖项也在最后一天正式揭晓。

2026-10-08 雷达全文

Harmonic Drive to share how integrated actuators can reduce engineering at RoboBusiness

2026-10-08T07:30:00Z

主线: 本体与供应链（硬件 / 成本）
证据: [M]
来源: The Robot Report · 2026-10-07

Attendees will learn how integrated actuators can reduce engineering effort by simplifying system integration.

2026-10-08 雷达全文

Demo: Closed-Loop Sionna-Isaac Sim Co-Simulation Framework for Wireless-Aware Robot Navigation over ROS 2

2026-10-07T07:19:00Z

主线: 仿真与评测（sim2real / 闭环）
证据: [R]
来源: arXiv 2610.08618 · 2026-10-06
关键数字: 16ms · 60 Hz · 7.4%
信号: 含量化指标 · 含闭环成功率 · 含延迟或控制频率

A robot that offloads its control loop to the network carries the receiver with it, so link quality is decided by where it goes.

2026-10-07 雷达全文

StairVLA: Stage-Aware Hierarchical Action Generation for Vision-Language-Action Models

2026-10-07T07:19:00Z

主线: 仿真与评测（sim2real / 闭环）
证据: [R]
来源: arXiv 2610.07756 · 2026-10-06
关键数字: 96.5% · 97.8% · 115.0 ms · 44.2 ms
信号: 含量化指标 · 含闭环成功率 · 含延迟或控制频率

Vision-language-action (VLA) models increasingly rely on diffusion- or flow-matching-based action heads to generate continuous robot actions.

2026-10-07 雷达全文

ESP: Energy-Score Policy for One-Step Multimodal Action Generation

2026-10-07T07:19:00Z

主线: 基座模型（VLA / WAM）
证据: [R]
来源: arXiv 2610.07696 · 2026-10-06
信号: 含闭环成功率 · 含真机实验 · 含延迟或控制频率

Generative action models based on diffusion and flow matching have been increasingly adopted in vision-language-action (VLA) policies for their ability to capture diverse behaviors, including multiple valid action sequences under the same…

2026-10-07 雷达全文

Adapting Vision-Language-Action Models to Unknown Visual Disruptions During Execution

2026-10-07T07:19:00Z

主线: 基座模型（VLA / WAM）
证据: [R]
来源: arXiv 2610.07946 · 2026-10-06
关键数字: 43.9% · 53.2% · 58.7% · 66.0%
信号: 含量化指标 · 含真机实验

Visual disruptions can arise while a robot is executing a task, leaving a vision-language-action (VLA) policy to respond without knowing the disruption type or timing.

2026-10-07 雷达全文

LightVLN: Efficient Aerial Vision-and-Language Navigation with Compact Memory and History-Guided Local Aggregation

2026-10-07T07:19:00Z

主线: 边缘与实时性（端侧推理 / 控制频率）
证据: [R]
来源: arXiv 2610.05024 · 2026-10-04
关键数字: 50.93% · 36.14% · 25.83% · 16 GB
信号: 含量化指标 · 含闭环成功率 · 含延迟或控制频率

Aerial vision-and-language navigation (VLN) enables unmanned aerial vehicles to execute long-horizon natural-language instructions from visual observations in complex three-dimensional environments.

2026-10-07 雷达全文

ActTune: Action-Aware Precision and GPU Operating-Point Adaptation for Energy-Efficient Vision-Language-Action Inference

2026-10-07T07:19:00Z

主线: 边缘与实时性（端侧推理 / 控制频率）
证据: [R]
来源: arXiv 2610.08444 · 2026-10-06
信号: 含闭环成功率 · 含延迟或控制频率

Vision-language-action (VLA) policies repeatedly invoke inference to control robots, making graphics processing unit (GPU) energy a recurring cost of task execution.

2026-10-07 雷达全文

ルネサスが低耐圧GaNデバイス市場に参入、業界トップクラスの性能指数を実現

2026-10-07T07:19:00Z

主线: 本体与供应链（硬件 / 成本）
证据: [M]
来源: MONOist (ITmedia) · 2026-10-07

ルネサスエレクトロニクスは、同社初となる100V耐圧のEモードGaNパワートランジスタ製品群を発売した。AIデータセンターやヒューマノイドロボットなど向けに、超高速スイッチングと優れた放熱性能を実現する。

2026-10-07 雷达全文

Boston Dynamics appoints former Amazon exec Rohit Prasad as new CEO

2026-10-07T07:19:00Z

主线: 本体与供应链（硬件 / 成本）
证据: [M]
来源: The Robot Report · 2026-10-06

Before Boston Dynamics, Prasad served as senior vice president and head scientist for Alexa and artificial general intelligence at Amazon.

2026-10-07 雷达全文

FastOPD: On-Policy Distillation for Lightweight VLA Deployment

2026-10-05T23:40:00Z

主线: 基座模型（VLA / WAM）
证据: [R]
来源: arXiv 2610.02832 · 2026-10-02
关键数字: 84% · 78.1%
信号: 含量化指标 · 含闭环成功率 · 含真机实验 · 含延迟或控制频率

Vision-Language-Action (VLA) foundation models have scaled rapidly to enhance manipulation performance and generalizability, but this scaling incurs high computational costs that render real-world deployment increasingly challenging.

2026-10-06 雷达全文

UniIntervene++: An Adaptive Intervention Agent for Efficient Real-World Reinforcement Learning

2026-10-05T23:40:00Z

主线: 训练与自我改进（RL / 后训练）
证据: [R]
来源: arXiv 2610.03620 · 2026-10-02
关键数字: 89.67% · 0.77% · 94.6%
信号: 含量化指标 · 含闭环成功率 · 含真机实验 · 有开放代码或权重

Online reinforcement learning (RL) enables robot policies to improve through physical interaction, but the assistance they require changes as their competence evolves.

2026-10-06 雷达全文

MobiAgent: Dual-Loop Recursive Policy Self-Improvement for Long-Horizon Mobile Manipulation

2026-10-05T23:40:00Z

主线: 系统与编排（长程 / 多策略）
证据: [R]
来源: arXiv 2610.03476 · 2026-10-02
关键数字: 7.50% · 27.50% · 32.5% · 57.5%
信号: 含量化指标 · 含真机实验

Long-horizon mobile manipulation presents significant challenges due to compounding execution errors and capacity interference between locomotion and arm control.

2026-10-06 雷达全文

World Action Learning via Interaction-Centric Spectral Latent Guidance

2026-10-05T23:40:00Z

主线: 仿真与评测（sim2real / 闭环）
证据: [R]
来源: arXiv 2610.03607 · 2026-10-02
关键数字: 99.20% · 93.80% · 57.7%
信号: 含量化指标 · 含闭环成功率 · 含真机实验

Learning general-purpose robot policies requires large-scale real-world interaction data, yet collecting robot demonstrations remains expensive and difficult to scale.

2026-10-06 雷达全文

SimpleTouch: Can Vision-Language-Action Models Master Contact-Rich Manipulation Without Tactile Policy Pretraining?

2026-10-05T23:40:00Z

主线: 基座模型（VLA / WAM）
证据: [R]
来源: arXiv 2610.02784 · 2026-10-02
关键数字: 50 demonstrations · 77.5% · 45.2% · 66.7%
信号: 含量化指标 · 含闭环成功率 · 含真机实验

Tactile sensing provides essential contact information for robotic manipulation, yet incorporating it into pretrained vision-language-action (VLA) models remains challenging.

2026-10-06 雷达全文

HexVIO: Towards All-Day Stereo-Inertial Tracking Through Commodity DSPs

2026-10-05T23:40:00Z

主线: 边缘与实时性（端侧推理 / 控制频率）
证据: [R]
来源: arXiv 2610.03283 · 2026-10-02
关键数字: 67% · 86% · 30 fps · 0.83 W
信号: 含量化指标 · 含延迟或控制频率

The ability of a device to localize itself within its surroundings is a fundamental prerequisite for spatial computing.

2026-10-06 雷达全文

FCC robot restrictions could accelerate shift to local AI

2026-10-05T23:40:00Z

主线: 安全、权限与合规
证据: [M]
来源: The Robot Report · 2026-10-05

Companies deploying physical AI today need to decide which AI workloads belong on the robot, what can remain in the cloud, and whether the underlying technology stack can meet rising expectations for security, safety, and supply-chain…

2026-10-06 雷达全文

川崎重工とNVIDIAが目指す「次世代デジタルシップヤード構想」とは何か

2026-10-05T23:40:00Z

主线: 仿真与评测（sim2real / 闭环）
证据: [M]
来源: MONOist (ITmedia) · 2026-10-05

川崎重工は、NVIDIAのフィジカルAIとデジタルツイン技術を活用し、「次世代デジタルシップヤード」の実現を目指すプロジェクトを開始した。海外で進むデジタルシップヤード構想と比べて、川崎重工とNVIDIAの取り組みにはどのような違いがあるのだろうのか。

2026-10-06 雷达全文

Is Success All You Need? Investigating the Impact of Input Perturbations on VLA Behaviour in Tabletop Manipulation Tasks

2026-10-04T23:40:00Z

主线: 仿真与评测（sim2real / 闭环）
证据: [R]
来源: arXiv 2610.01351 · 2026-10-01
信号: 含闭环成功率

Vision-Language-Action (VLA) models have achieved high task success rates on robot manipulation task benchmarks.

2026-10-05 雷达全文

WBAG: A Whole-Body and Attached-Geometry Safety Framework for Vision-Language-Action Manipulation

2026-10-04T23:40:00Z

主线: 仿真与评测（sim2real / 闭环）
证据: [R]
来源: arXiv 2610.01083 · 2026-10-01
信号: 含闭环成功率

Vision-language-action (VLA) policies have demonstrated impressive capabilities in generalizable robotic manipulation, but their deployment in the real world remains challenging due to potential collisions involving different parts of the…

2026-10-05 雷达全文

NarrativeFlow: Flow-Based Vision-Language-Action Model Using Robot Velocity Fields

2026-10-04T23:40:00Z

主线: 基座模型（VLA / WAM）
证据: [R]
来源: arXiv 2610.00981 · 2026-10-01
信号: 含闭环成功率 · 含真机实验

We focus on language-conditioned flow-based manipulation, where robot flows (robot velocity fields) serve as embodiment-agnostic, motion-centric representations for leveraging data collected from multiple robot platforms.

2026-10-05 雷达全文

Divide-and-Remember: Recursive Action-Relevant Memory for Long-Horizon VLA Policies

2026-10-04T23:40:00Z

主线: 基座模型（VLA / WAM）
证据: [R]
来源: arXiv 2610.00982 · 2026-10-01
信号: 含闭环成功率 · 有开放代码或权重

Vision-language-action (VLA) models struggle on history-dependent manipulation tasks, where the current observation alone does not determine the action, and the policy needs a memory of the history.

2026-10-05 雷达全文

PROMO: Preference-conditioned Multi-Objective Reinforcement Learning for Quadrupedal Robots

2026-10-04T23:40:00Z

主线: 训练与自我改进（RL / 后训练）
证据: [R]
来源: arXiv 2610.01260 · 2026-10-01
关键数字: 30.4% · 38.7% · 59.0%
信号: 含量化指标 · 有开放代码或权重

Quadrupedal locomotion requires balancing conflicting objectives such as command tracking, stability, and energy efficiency, yet conventional reinforcement learning (RL) hardcodes these priorities into a fixed scalar reward at training…

2026-10-05 雷达全文

TOAST: Stochastic Robot Action Tokenization for Autoregressive Vision-Language-Action Models

2026-10-04T23:40:00Z

主线: 边缘与实时性（端侧推理 / 控制频率）
证据: [R]
来源: arXiv 2610.00899 · 2026-10-01
信号: 含闭环成功率

Autoregressive Vision-Language-Action models often represent continuous robot actions as discrete token sequences, enabling action prediction with standard next-token objectives.

2026-10-05 雷达全文

UniWAM: Unified World-Action Model

2026-10-04T23:40:00Z

主线: 数据引擎（采集 / 清洗 / scaling）
证据: [R]
来源: arXiv 2610.02054 · 2026-10-01

Vision-language-action models benefit from the understanding and reasoning capabilities of pretrained vision-language models, but action-only supervision provides limited grounding in world dynamics.

2026-10-05 雷达全文

InterEvolve: Test-Time Evolution of Reward Programs for Humanoid Loco-Manipulation

2026-10-04T23:40:00Z

主线: 系统与编排（长程 / 多策略）
证据: [R]
来源: arXiv 2610.02196 · 2026-10-01

We study test-time evolution for humanoid loco-manipulation: solving tasks that a controller was never trained for by repurposing its existing skills, improving from its own attempts, and retaining what it learns, without retraining.

2026-10-05 雷达全文

ReCo: Response-Consistent Locomotion with Policy-Aware MPC for Legged Manipulation

2026-10-03T23:40:00Z

主线: 仿真与评测（sim2real / 闭环）
证据: [R]
来源: arXiv 2610.01612 · 2026-10-01
关键数字: 28.7% · 27.4%
信号: 含量化指标 · 含闭环成功率 · 含真机实验

Continuous legged manipulation requires accurate end-effector tracking while the base keeps walking. Combining reinforcement learning (RL) with model predictive control (MPC) suits this task: the learned policy provides robust locomotion,…

2026-10-04 雷达全文

DuoMind: Enabling Distributed Multi-Robot Coordination with Semantic Communication

2026-10-03T23:40:00Z

主线: 系统与编排（长程 / 多策略）
证据: [R]
来源: arXiv 2610.02161 · 2026-10-01
信号: 含闭环成功率

Vision-language models (VLMs) and vision-language-action models (VLAs) have recently driven rapid progress in general-purpose robots, yet most progress has focused on single-robot settings.

2026-10-04 雷达全文

ChunkVLA-AM: Parallel Action Chunking for Vision-Language-Action Robot Control in Additive Manufacturing

2026-10-03T23:40:00Z

主线: 基座模型（VLA / WAM）
证据: [R]
来源: arXiv 2610.01856 · 2026-10-01
关键数字: 92.9%
信号: 含量化指标 · 含闭环成功率 · 含真机实验

Vision-language-action (VLA) models unify visual perception, language understanding, and action generation, offering new opportunities for automation in additive manufacturing (AM).

2026-10-04 雷达全文

IronMind: Scaling Humanoid Dexterous Manipulation via Camera-Space Ego-Centric Pretraining

2026-10-03T23:40:00Z

主线: 数据引擎（采集 / 清洗 / scaling）
证据: [R]
来源: arXiv 2609.39403 · 2026-09-30
关键数字: 10,000 hours · 55.0% · 11.7% · 5,000 hours
信号: 含量化指标 · 含闭环成功率

Egocentric human video offers a scalable data source for dexterous manipulation, yet using it to train humanoid robots presents two challenges: (1) an embodiment gap, as human hands differ structurally from robot end-effectors and…

2026-10-04 雷达全文

EmbodiRSI: Recursive Self-Improvement for Data-Efficient Robot Adaptation

2026-10-03T23:40:00Z

主线: 仿真与评测（sim2real / 闭环）
证据: [R]
来源: arXiv 2609.38905 · 2026-09-30
关键数字: 50.4% · 83.5% · 83.1% · 75.0%
信号: 含量化指标 · 含真机实验

Adapting robot manipulation policies to new tasks and environments remains highly data-intensive, while the data needed for further improvement depends on the policy's current capabilities and failure modes.

2026-10-04 雷达全文

Towards Agile Vision-Based Multi-UAV Flight: Revisiting State Estimation

2026-10-03T23:40:00Z

主线: 边缘与实时性（端侧推理 / 控制频率）
证据: [R]
来源: arXiv 2609.39611 · 2026-09-30
关键数字: 40% · 57% · 300 ms
信号: 含量化指标 · 含闭环成功率 · 含真机实验 · 含延迟或控制频率

Agile multi-UAV flight requires accurate and low-latency onboard estimation of the kinematic states of neighboring UAVs for collision avoidance, motion coordination, etc.

2026-10-04 雷达全文

GroundAnything: Reconciling Parallel Decoding with Precise Visual Grounding at Flash Speed

2026-10-03T23:40:00Z

主线: 训练与自我改进（RL / 后训练）
证据: [R]
来源: arXiv 2609.39600 · 2026-09-30
关键数字: 72.42% · 71.35% · 61.75% · 53.32%
信号: 含量化指标 · 含真机实验 · 含延迟或控制频率

Autoregressive (AR) grounding models serialize spatial predictions, introducing sequential latency and imposing a causal order on output tokens.

2026-10-04 雷达全文

DODGER: Safety-Guided Reinforcement Learning for Robot Navigation Among Dynamic Obstacles

2026-10-03T23:40:00Z

主线: 安全、权限与合规
证据: [R]
来源: arXiv 2609.38873 · 2026-09-30
信号: 含真机实验

Robots operating in human-centered environments must safely navigate among multiple dynamic obstacles to avoid collisions with people and surrounding infrastructure.

2026-10-04 雷达全文

Token-World: World Modeling in Vision-Language Model Token Space for Robot Manipulation

2026-10-02T23:40:00Z

主线: 仿真与评测（sim2real / 闭环）
证据: [R]
来源: arXiv 2610.00575 · 2026-09-30
信号: 含闭环成功率 · 含延迟或控制频率

A common approach to world-model simulation for vision-language-action (VLA) systems is to predict future RGB observations and then re-encode them into policy inputs, introducing an indirect interface between simulation and downstream…

2026-10-03 雷达全文

DexPolicy: Scheduled Exploration for Trajectory-Guided Dexterous Manipulation

2026-10-02T23:40:00Z

主线: 训练与自我改进（RL / 后训练）
证据: [R]
来源: arXiv 2610.00360 · 2026-09-30
关键数字: 49.4% · 68.1% · 14.1% · 45.4%
信号: 含量化指标 · 含闭环成功率 · 有开放代码或权重

Reinforcement learning (RL) for dexterous manipulation must discover finger-object contacts and then control the object precisely; the action noise that serves the first goal can interfere with the second.

2026-10-03 雷达全文

Kinematic MeanFlow: One-Step Action Generation Policy for Robotic Foundation Models

2026-10-02T23:40:00Z

主线: 边缘与实时性（端侧推理 / 控制频率）
证据: [R]
来源: arXiv 2610.00864 · 2026-10-01
关键数字: 67.5% · 74.4% · 30.3% · 54.9%
信号: 含量化指标 · 有开放代码或权重 · 含延迟或控制频率

In this paper, we study how to achieve one-step action generation in Robotic Foundation Models (RFMs), aiming to overcome the high inference latency of multi-step flow matching.

2026-10-03 雷达全文

IndoorBEV: A Lightweight Real-Time LiDAR BEV Perception System for Indoor Mobile Robots

2026-10-02T23:40:00Z

主线: 边缘与实时性（端侧推理 / 控制频率）
证据: [R]
来源: arXiv 2610.00355 · 2026-09-30
关键数字: 0.6M parameters · 169.6 ms · 200 ms
信号: 含量化指标 · 含真机实验 · 有开放代码或权重 · 含延迟或控制频率

Efficient indoor LiDAR perception is challenging because mobile robots must understand cluttered three-dimensional environments under strict latency and memory constraints.

2026-10-03 雷达全文

Isaac Lab on AWS: From Simulation to Registered Policy

2026-10-02T23:40:00Z

主线: 仿真与评测（sim2real / 闭环）
证据: [O]
来源: AWS Physical AI Blog · 2026-10-02
信号: 含真机实验

Introduction: The missing layer in Physical AI on AWS Physical AI systems, those that perceive, reason, and act in the physical world, are moving fast from research into production deployments.

2026-10-03 雷达全文

Video Friday: Albatross Falls, Spins, Self-Rights, and Sails Away

2026-10-02T23:40:00Z

主线: 本体与供应链（硬件 / 成本）
证据: [M]
来源: IEEE Spectrum Robotics · 2026-10-02

We’ve developed ALBATROSS, a hybrid aerial–marine robot that can be released from the air, autorotate down to the water without a parachute, passively self-right after landing, and then reuse the same rigid wings as sails for autonomous…

2026-10-03 雷达全文

Runway introduces Praxis-1 world action model for robotics

2026-10-02T23:40:00Z

主线: 基座模型（VLA / WAM）
证据: [M]
来源: The Robot Report · 2026-10-02
信号: 有开放代码或权重

When Runway does release Praxis-1 publicly, it plans to ship it with open weights rather than as a closed model.

2026-10-03 雷达全文

Inside Omron’s next-generation LD mobile robots

2026-10-02T23:40:00Z

主线: 安全、权限与合规
证据: [M]
来源: The Robot Report · 2026-10-02

In mobile robotics, software, safety, and fleet management are becoming just as important as the robot itself, according to an OMROM project manager.

2026-10-03 雷达全文

DSDyn-VLA: A Dual-Stream Dynamic Manipulation Framework with Motion Perception, Future Awareness, and Realtime Correction

2026-10-01T23:40:00Z

主线: 基座模型（VLA / WAM）
证据: [R]
来源: arXiv 2609.39198 · 2026-09-30
信号: 含闭环成功率 · 含真机实验 · 有开放代码或权重 · 含延迟或控制频率

While Vision-Language-Action (VLA) models excel in static tasks, they struggle in dynamic environments where objects are in motion (e.g., conveyor belt manipulation).

2026-10-02 雷达全文

DiffWAM: A Fast and Efficient Navigation World Action Model

2026-10-01T23:40:00Z

主线: 边缘与实时性（端侧推理 / 控制频率）
证据: [R]
来源: arXiv 2609.39763 · 2026-09-30
关键数字: 74.40%
信号: 含量化指标 · 含闭环成功率 · 含真机实验 · 含延迟或控制频率

Pretrained video foundation models encode rich semantic and spatiotemporal priors for embodied navigation, yet converting these priors into UAV motion typically requires expensive future-video synthesis and geometric reconstruction.

2026-10-02 雷达全文

Faster and Better? Benchmark Bugs and Design Limitations Distort the Evaluation of Vision-Language-Action Acceleration

2026-10-01T23:40:00Z

主线: 仿真与评测（sim2real / 闭环）
证据: [R]
来源: arXiv 2609.37771 · 2026-09-29
信号: 含闭环成功率 · 有开放代码或权重 · 含延迟或控制频率

Simulated manipulation benchmarks are the standard tool for evaluating vision-language-action (VLA) policies and the acceleration methods that reduce their inference latency for on-robot deployment.

2026-10-02 雷达全文

Toward Real-Time VLAs: Stage-Aware Two-Step Flow Denoising and System-Level Evaluation

2026-10-01T23:40:00Z

主线: 基座模型（VLA / WAM）
证据: [R]
来源: arXiv 2609.39822 · 2026-09-30
关键数字: 61.557 ms · 21.956 ms
信号: 含量化指标 · 含闭环成功率 · 含延迟或控制频率

Vision-language-action (VLA) models face a timing gap between low-rate inference and high-rate robot execution.

2026-10-02 雷达全文

Atlas Robot’s New Hand May Outperform Humanlike Designs

2026-10-01T23:40:00Z

主线: 本体与供应链（硬件 / 成本）
证据: [M]
来源: IEEE Spectrum Robotics · 2026-10-01

Today, Boston Dynamics is announcing a redesigned hand for its Atlas robot . The previous generation could do some crazy superhuman things with its three fingers, but it was never intended for mass production in the tens or hundreds of…

2026-10-02 雷达全文

Boston Dynamics drops pinkie on new humanoid hand

2026-10-01T23:40:00Z

主线: 本体与供应链（硬件 / 成本）
证据: [M]
来源: The Robot Report · 2026-10-01

Boston Dynamics engineers taped their pinkies and ring fingers together for a day to learn what they weren’t able to do.

2026-10-02 雷达全文

Your Robot’s Safety Functions Already Work. What If the Input Lies?

2026-10-01T23:40:00Z

主线: 安全、权限与合规
证据: [M]
来源: The Robot Report · 2026-10-01

A robot can follow its safety rules and still act on manipulated information. See what this means for robot safety and how teams can strengthen their testing and evidence.

2026-10-02 雷达全文

ANYbotics launches Shift to streamline robot fleet operations, scale inspections

2026-10-01T23:40:00Z

主线: 系统与编排（长程 / 多策略）
证据: [M]
来源: The Robot Report · 2026-10-01

ANYbotics designed Shift to help maintenance teams run industrial inspections, detect failing equipment, and turn data into work orders.

2026-10-02 雷达全文

WorldLine: Action-Driven Visual Simulation for Robotic Manipulation

2026-09-30T23:40:00Z

主线: 仿真与评测（sim2real / 闭环）
证据: [R]
来源: arXiv 2609.38059 · 2026-09-29
关键数字: 10,000 hours · 2,000 hours · 74%
信号: 含量化指标 · 含闭环成功率 · 含真机实验

Real-world robot learning is constrained by the cost of collecting experience and evaluating candidate behaviors.

2026-10-01 雷达全文

RoboFin3D: A Sim-to-Real Platform for Robotic Surface Finishing

2026-09-30T23:40:00Z

主线: 仿真与评测（sim2real / 闭环）
证据: [R]
来源: arXiv 2609.37560 · 2026-09-29
关键数字: 77.15% · 84.47% · 97.41%
信号: 含量化指标 · 含真机实验

Grinding and sanding are fundamental processes in industrial robotic surface finishing. However, physical trials are expensive and consume workpieces, making reproducible experiments difficult.

2026-10-01 雷达全文

Cooperative Multi-Agent Vision-Language-Action Models via Reinforced Fine Tuning

2026-09-30T23:40:00Z

主线: 训练与自我改进（RL / 后训练）
证据: [R]
来源: arXiv 2609.36588 · 2026-09-29
关键数字: 11 tasks
信号: 含闭环成功率 · 含真机实验

We study reinforcement learning (RL) methods for cooperative multi-agent Vision-Language-Action (VLA) models. This problem is challenging because VLAs are pretrained on large-scale single-agent data and therefore lack the fine-grained…

2026-10-01 雷达全文

Urgent Actions Go First: Urgency-Aware Denoising for Real-Time VLA Control

2026-09-30T23:40:00Z

主线: 边缘与实时性（端侧推理 / 控制频率）
证据: [R]
来源: arXiv 2609.37772 · 2026-09-29
关键数字: 1.89x
信号: 含闭环成功率 · 含真机实验 · 含延迟或控制频率

Diffusion and flow-matching Vision-Language-Action (VLA) policies generate action chunks through iterative denoising, incurring substantial inference latency that severely limits real-time robotic control.

2026-10-01 雷达全文