知识脉络总索引 (INDEX)
这是 research/ 的入口。先看本文件，再决定深入哪条线索。
组织原则：research/STRUCTURE.md。核心是「线索（thread）优先」，论文是线索的注脚。
📊 文献基础（核实完成）
- 论文库：
data/curated/papers.jsonl—— 971 篇，7 大方向（含 dexterous），37% 开源 - 数据集库：
data/curated/datasets.jsonl—— 44 个数据集/仿真/benchmark - 引用关系图：
data/curated/citations.dot（papers.jsonl 共 3240 条 builds_on/leads_to 标签，其中 1872 条解析为库内规范化边，覆盖 796 个节点） - 各方向原始综述：
data/raw/*_notes.md（6 份）
元数据核实状态：✅ 全部完成。145 篇 arxiv API 核实 + WebSearch 补充，累计修正 ~35 处真错误（含 971 篇论文 arxiv id 全错指向天文/物理论文、BFM 作者张冠李戴、多篇 venue 跨文件不一致）。剩余 0 篇标记待查。详见 log/CHANGELOG.md。
查询技巧：
# 按关键词找论文
grep -i "关键词" research/data/curated/papers.jsonl | python3 -c "import json,sys; [print(json.loads(l)['title'], json.loads(l).get('arxiv','')) for l in sys.stdin]"
🧵 技术演进线索（threads/）—— 核心
每条线索是一篇深度 md，回答「这条线在解决什么、怎么演进的、开放问题是什么」。已完成 14 条技术演进线索。
| 编号 | 线索 | 核心问题 | 状态 | 字符 | 文件 |
|---|---|---|---|---|---|
| T01 | Sim-to-Real 与在线适应 | 仿真策略怎么迁移真机、怎么实时适应未知环境 | 🟡 | 23k | T01-sim2real-locomotion.md |
| T02 | 数据集基础设施演化 ⭐ | OpenX/DROID/AMASS/Isaac Lab 怎么把机器人推到大数据预训练范式 | 🟢 | 17k | T02-dataset-infrastructure.md |
| T03 | Reward 设计的演化 | 手工 → motion prior → LLM 自动 | 🟡 | 10k | T03-reward-design-evolution.md |
| T04 | Motion Prior: AMP → BFM ⭐ | 把动捕数据沉淀成机器人运动常识（用户重点） | 🟢 | 25k | T04-motion-prior-amp-to-bfm.md |
| T05 | Diffusion Policy 谱系 | 从 BC 到多模态动作分布建模 | 🟢 | 16k | T05-diffusion-policy-lineage.md |
| T06 | VLA 五段演进 | 用大模型当控制策略 | 🟢 | 20k | T06-vla-five-stages.md |
| T07 | 人形 teleop 技术栈 | 实时遥控 + 数据采集管道 | 🟢 | 17k | T07-humanoid-teleop-stack.md |
| T08 | 动作生成 physics-aware | 从纯 kinematic 到真机可执行 | 🟢 | 17k | T08-action-gen-physics.md |
| T09 | World Model 作 simulator | 学一个世界模型当数据源/评估器/规划器 | 🟢 | 12k | T09-world-model-as-simulator.md |
| T10 | VLA + WBC 分层整合 | 人形控制栈的当前最热开放问题 | 🔴 最热 | 16k | T10-vla-wbc-integration.md |
| T11 | 灵巧手操作 ⭐ | 多指灵巧操作（抓/转/捏/工具），DoF 高 + 触觉缺失 + sim-real gap | 🔴 最热 | 16k | T11-dexterous-manipulation.md |
| T12 | 敏捷 Locomotion（parkour 链 + 剧烈动作） | 跳/爬/翻滚等高动态动作怎么学、为什么和平稳走路是两个世界、剧烈动作 sim-real 是否本质无解 | 🟢 | 24k | T12-agile-locomotion.md |
| T13 | 综述视角（鸟瞰整合） | 基于 8 篇综述 + 12 线索的全景俯瞰、5 领域共识、5 未来方向 | 🟢 | 10k | T13-survey-of-embodied-rl.md |
| T14 | 抓取（Grasping） | 从几何规划到学习到生成，抓取 vs 灵巧操作的边界 | 🟢 | 9k | T14-grasping.md |
图例：🟢 活跃（仍在快速演进） / 🟡 趋稳（主要问题已解决） / 🔴 最热（当前研究焦点）
候选线索（下一批可深挖，尚未立项）：人形 loco-manipulation 长程任务 / 安全 RL / deformable terrain
📄 论文索引（papers/）
覆盖所有有可访问页面的论文（方程笔记 / 直觉解读 / 教学页任一即可），按重要度分层：里程碑（有方程笔记）→ 枢纽（被引密集）→ 其它。每条均可点击直达对应深度版本。
本索引列出有专门解读或教学页面的 335 篇论文；知识库全部 971 篇可通过搜索按元数据查找。
里程碑（43 篇 · 有方程笔记的奠基性工作）
| 论文 | 可用版本（点击进入） |
|---|---|
| AMP: Adversarial Motion Priors for Stylized Physics-Based Character Control (2021) | 方程笔记直觉解读 |
| OpenVLA: An Open-Source Vision-Language-Action Model (2024) | 完整教学方程笔记直觉解读 |
| ASE: Large-Scale Reusable Adversarial Skill Embeddings for Physically Simulated Characters (2022) | 入门完整教学方程笔记直觉解读 |
| BeyondMimic: From Motion Tracking to Versatile Humanoid Control via Guided Diffusion (2025) | 完整教学方程笔记直觉解读 |
| Human Motion Diffusion Model (MDM) (2022) | 完整教学方程笔记直觉解读 |
| Extreme Parkour with Legged Robots (2023) | 完整教学方程笔记直觉解读 |
| RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Control (2023) | 完整教学方程笔记直觉解读 |
| OmniH2O: Universal and Dexterous Human-to-Humanoid Whole-Body Teleoperation and Learning (2024) | 完整教学方程笔记直觉解读 |
| Open X-Embodiment: Robotic Learning Datasets and RT-X Models (2023) | 完整教学方程笔记直觉解读 |
| π0: A Vision-Language-Action Flow Model for General Robot Control (2024) | 完整教学方程笔记直觉解读 |
| Diffusion Policy: Visuomotor Policy Learning via Action Diffusion (2023) | 完整教学方程笔记直觉解读 |
| DeepMimic: Example-Guided Deep Reinforcement Learning of Physics-Based Character Skills (2018) | 完整教学方程笔记直觉解读 |
| CALM: Conditional Adversarial Latent Models for Directable Virtual Characters (2023) | 入门方程笔记直觉解读 |
| DreamerV3: Mastering Diverse Domains through World Models (2023) | 完整教学方程笔记直觉解读 |
| Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardware (ACT / ALOHA) (2023) | 完整教学方程笔记直觉解读 |
| ASAP: Aligning Simulation and Real-World Physics for Learning Agile Humanoid Whole-Body Skills (2025) | 完整教学方程笔记直觉解读 |
| Behavior Foundation Model for Humanoid Robots (2025) | 完整教学方程笔记直觉解读 |
| Cosmos World Foundation Model Platform for Physical AI (2025) | 完整教学方程笔记直觉解读 |
| pi0.5: a Vision-Language-Action Model with Open-World Generalization (2025) | 方程笔记直觉解读 |
| HumanPlus: Humanoid Shadowing and Imitation from Humans (2024) | 入门完整教学方程笔记直觉解读 |
| ANYmal Parkour: Learning Agile Navigation for Quadrupedal Robots (2023) | 入门方程笔记直觉解读 |
| Learning robust perceptive locomotion for quadrupedal robots in the wild (ANYmal-C in the wild) (2022) | 完整教学方程笔记直觉解读 |
| Learning Dexterous In-Hand Manipulation (2018) | 完整教学方程笔记直觉解读 |
| SONIC: Supersizing Motion Tracking for Natural Humanoid Whole-Body Control (2025) | 完整教学方程笔记直觉解读 |
| DrEureka: Language Model Guided Sim-To-Real Transfer (2024) | 完整教学方程笔记直觉解读 |
| Generalizable Humanoid Manipulation with Improved 3D Diffusion Policy (iDP3) (2024) | 完整教学方程笔记直觉解读 |
| H2O: Learning Human-to-Humanoid Real-Time Whole-Body Teleoperation (2024) | 完整教学方程笔记直觉解读 |
| HOVER: Versatile Neural Whole-Body Controller for Humanoid Robots (2024) | 完整教学方程笔记直觉解读 |
| MaskedMimic: Unified Physics-Based Character Control Through Masked Motion Inpainting (2024) | 入门方程笔记直觉解读 |
| Perpetual Humanoid Control for Real-time Simulated Avatars (2023) | 完整教学方程笔记直觉解读直觉解读 |
| RMA: Rapid Motor Adaptation for Legged Robots (2021) | 完整教学方程笔记直觉解读 |
| 3D Diffusion Policy: Generalizable Visuomotor Policy Learning via Simple 3D Representations (DP3) (2024) | 完整教学方程笔记直觉解读 |
| Visual Dexterity: In-Hand Reorientation of Novel and Complex Object Shapes (2023) | 完整教学方程笔记直觉解读 |
| GraspNet-1Billion: A Large-Scale Benchmark for General Object Grasping (2020) | 入门完整教学方程笔记直觉解读 |
| QT-Opt: Scalable Deep Reinforcement Learning for Vision-Based Robotic Manipulation (2018) | 完整教学方程笔记直觉解读 |
| GR00T N1: An Open Foundation Model for Generalist Humanoid Robots (2025) | 完整教学方程笔记直觉解读 |
| Bunny-VisionPro: Real-Time Bimanual Dexterous Teleoperation for Imitation Learning (2024) | 完整教学方程笔记直觉解读 |
| AnyGrasp: Robust and Efficient Grasp Perception in Spatial and Temporal Domains (2023) | 完整教学方程笔记直觉解读 |
| Learning Whole-Body Humanoid Locomotion via Motion Generation and Motion Tracking (2026) | 入门方程笔记直觉解读 |
| DexMimicGen: Automated Data Generation for Bimanual Dexterous Manipulation (2024) | 方程笔记直觉解读 |
| Eureka: Human-Level Reward Design via Coding Large Language Models (2023) | 完整教学方程笔记直觉解读 |
| Isaac Gym: High Performance GPU-Based Physics Simulation For Robot Learning (2021) | 完整教学方程笔记直觉解读 |
| DexPilot: Vision Based Teleoperation of Dexterous Robotic Hand-Arm System (2019) | 完整教学方程笔记直觉解读 |
枢纽（204 篇 · 被引密集的核心工作（含直觉解读））
| 论文 | 可用版本（点击进入） |
|---|---|
| RT-1: Robotics Transformer for Real-World Control at Scale (2022) | 入门直觉解读 |
| PhysDiff: Physics-Guided Human Motion Diffusion Model (2023) | 入门直觉解读 |
| RDT-1B: a Diffusion Foundation Model for Bimanual Manipulation (2024) | 入门直觉解读 |
| MotionGPT: Human Motion as a Foreign Language (2023) | 入门直觉解读 |
| HOMIE: Humanoid Loco-Manipulation with Isomorphic Exoskeleton Teleoperation (2025) | 入门直觉解读 |
| Expressive Whole-Body Control for Humanoid Robots (ExBody) (2024) | 入门直觉解读 |
| ControlVAE: Model-based Learning of Generative Controllers for Physics-based Characters (2022) | 入门直觉解读 |
| TACTO: A Fast, Flexible, and Open-source Simulator for High-Resolution Vision-based Tactile Sensors (2022) | 入门直觉解读 |
| ManiSkill3: GPU Parallelized Robotics Simulation and Rendering for Generalizable Embodied AI (2024) | 入门直觉解读 |
| DreamWaQ: Learning Robust Quadrupedal Locomotion With Implicit Terrain Imagination via Deep Reinforcement Learning (2023) | 入门直觉解读 |
| GMD: Guided Motion Diffusion for Controllable Human Motion Synthesis (2023) | 入门直觉解读 |
| Bailando: 3D Dance Generation by Actor-Critic GPT with Choreographic Memory (2022) | 入门直觉解读 |
| Code as Policies: Language Model Programs for Embodied Control (2022) | 入门直觉解读 |
| R3M: A universal visual representation for robot manipulation (2022) | 入门直觉解读 |
| VIP: Towards universal visual reward and representation via value-implicit pre-training (2022) | 入门直觉解读 |
| A Style-Based Generator Architecture for Generative Adversarial Networks (StyleGAN) (2018) | 直觉解读 |
| Generative Adversarial Imitation Learning (GAIL) (2016) | 入门直觉解读 |
| DexImit: Learning Bimanual Dexterous Manipulation from Monocular Human Videos (2026) | 入门直觉解读 |
| Perceptive Humanoid Parkour: Chaining Dynamic Human Skills via Motion Matching (PHP) (2026) | 入门直觉解读 |
| FAST: Efficient Action Tokenization for Vision-Language-Action Models (2025) | 入门直觉解读 |
| Humanoid Locomotion and Manipulation: Current Progress and Challenges in Control, Planning, and Learning (2025) | 入门直觉解读 |
| Parkour in the Wild: Learning a General and Extensible Agile Locomotion Policy Using Multi-Expert Distillation and RL Fine-Tuning (2025) | 入门直觉解读 |
| WholeBodyVLA: Towards Unified Latent VLA for Whole-Body Loco-Manipulation Control (2025) | 入门直觉解读 |
| Behavior Generation with Latent Actions (VQ-BeT) (2024) | 入门直觉解读 |
| EgoMimic: Scaling Imitation Learning via Egocentric Video (2024) | 入门直觉解读 |
| FlowPolicy: Enabling Fast and Robust 3D Flow-based Policy via Consistency Flow Matching for Robot Manipulation (2024) | 入门直觉解读 |
| HumanVLA: Towards Vision-Language Directed Object Rearrangement by Physical Humanoid (2024) | 入门直觉解读 |
| Octo: An Open-Source Generalist Robot Policy (2024) | 入门直觉解读 |
| DINOv2: Learning Robust Visual Features Without Supervision (2023) | 直觉解读 |
| HomeRobot: Open-Vocabulary Mobile Manipulation (2023) | 入门 |
| Humans in 4D: Reconstructing and Tracking Humans with Transformers (HMR 2.0 / 4DHumans) (2023) | 入门直觉解读 |
| MoMask: Generative Masked Modeling of 3D Human Motions (2023) | 入门直觉解读 |
| PaLM-E: An Embodied Multimodal Language Model (2023) | 入门直觉解读 |
| PhysHOI: Physics-Based Imitation of Dynamic Human-Object Interaction (2023) | 入门直觉解读 |
| Q-Transformer: Scalable Offline Reinforcement Learning via Autoregressive Q-Functions (2023) | 入门直觉解读 |
| RoboCat: A Self-Improving Generalist Agent for Robotic Manipulation (2023) | 入门直觉解读 |
| T2M-GPT: Generating Human Motion from Textual Descriptions with Discrete Representations (2023) | 入门直觉解读 |
| UniDexGrasp: Universal Robotic Dexterous Grasping via Learning Diverse Proposal Generation and Goal-Conditioned Policy (2023) | 入门直觉解读 |
| Universal Humanoid Motion Representations for Physics-Based Control (2023) | 入门 |
| Where are we in the search for an artificial visual cortex for embodied intelligence? (VC-1) (2023) | 入门直觉解读 |
| EDGE: Editable Dance Generation From Music (2022) | 入门直觉解读 |
| Inner Monologue: Embodied Reasoning through Planning with Language Models (2022) | 入门直觉解读 |
| High-Resolution Image Synthesis with Latent Diffusion Models (LDM / Stable Diffusion) (2021) | 直觉解读 |
| SimPoE: Simulated Character Control for 3D Human Pose Estimation (2021) | 入门直觉解读 |
| Character Controllers Using Motion VAEs (2020) | 入门直觉解读 |
| AMASS: Archive of Motion Capture as Surface Shapes (2019) | 入门直觉解读 |
| Large Scale GAN Training for High Fidelity Natural Image Synthesis (BigGAN) (2018) | 直觉解读 |
| Proximal Policy Optimization Algorithms (PPO) (2017) | 入门直觉解读 |
| Cross-Hand Latent Representation for Vision-Language-Action Models (2026) | 直觉解读 |
| Dream-Tac: A Unified Tactile World Action Model for Contact-Rich Robot Manipulation (2026) | 入门直觉解读 |
| EAGLE: Embodiment-Aware Generalist Specialist Distillation for Unified Humanoid Whole-Body Control (2026) | 入门直觉解读 |
| General Humanoid Whole-Body Control via Pretraining and Fast Adaptation (2026) | 入门直觉解读直觉解读 |
| Genie Sim 3.0: Open-Source Embodied AI Simulation Platform (2026) | 直觉解读 |
| HumanoidMimicGen: Data Generation for Loco-Manipulation via Whole-Body Planning (2026) | 入门直觉解读 |
| Learning Action Priors for Cross-embodiment Robot Manipulation (2026) | 入门直觉解读 |
| LeRobot: An Open-Source Library for End-to-End Robot Learning (2026) | 入门直觉解读 |
| OmniXtreme: Breaking the Generality Barrier in High-Dynamic Humanoid Control (2026) | 入门直觉解读 |
| Real-Time Whole-Body Teleoperation of a Humanoid Robot Using IMU-Based Motion Capture with Sim2Sim and Sim2Real Validation (2026) | 入门直觉解读直觉解读 |
| RISE: Self-Improving Robot Policy with Compositional World Model (2026) | 入门直觉解读 |
| ULTRA: Unified Multimodal Control for Autonomous Humanoid Whole-Body Loco-Manipulation (2026) | 直觉解读 |
| UltraDexGrasp: Learning Universal Dexterous Grasping for Bimanual Robots with Synthetic Data (2026) | 入门直觉解读 |
| Unified Walking, Running, and Recovery for Humanoids via State-Dependent Adversarial Motion Priors (2026) | 入门直觉解读 |
| Xiaomi-Robotics-1: Scaling Vision-Language-Action Models with over 100K Hours of Real-World Trajectories (2026) | 入门直觉解读 |
| A Survey of Behavior Foundation Model: Next-Generation Whole-Body Control System of Humanoid Robots (2025) | 入门直觉解读 |
| CLONE: Closed-Loop Whole-Body Humanoid Teleoperation for Long-Horizon Loco-Manipulation (2025) | 入门直觉解读 |
| InterMimic: Towards Universal Whole-Body Control for Physics-Based Human-Object Interactions (2025) | 入门直觉解读 |
| X-VLA: Soft-Prompted Transformer as Scalable Cross-Embodiment Vision-Language-Action Model (2025) | 入门直觉解读 |
| A Survey on Vision-Language-Action Models for Embodied AI (2024) | 入门直觉解读 |
| CogACT: A Foundational Vision-Language-Action Model for Synergizing Cognition and Action in Robotic Manipulation (2024) | 入门直觉解读 |
| DexCap: Scalable and Portable Mocap Data Collection System for Dexterous Manipulation (2024) | 入门直觉解读 |
| GR-2: A Generative Video-Language-Action Model with Web-Scale Knowledge for Robot Manipulation (2024) | 入门直觉解读 |
| Mobile ALOHA: Learning Bimanual Mobile Manipulation using Low-Cost Whole-Body Teleoperation (2024) | 入门直觉解读 |
| MotionGPT-2: A General-Purpose Motion-Language Model for Motion Generation and Understanding (2024) | 入门 |
| ReinDiffuse: Crafting Physically Plausible Motions with Reinforced Diffusion Model (2024) | 入门直觉解读 |
| RoboCasa: Large-Scale Demonstration Data for Everyday Manipulation (2024) | 直觉解读 |
| StableMoFusion: Towards Robust and Efficient Diffusion-based Motion Generation Framework (2024) | 入门直觉解读 |
| TD-MPC2: Scalable, Robust World Models for Continuous Control (2024) | 入门直觉解读 |
| TinyVLA: Towards Fast, Data-Efficient Vision-Language-Action Models for Robotic Manipulation (2024) | 入门直觉解读 |
| Consistency Models (2023) | 直觉解读 |
| DexArt: Benchmarking Generalizable Dexterous Manipulation with Articulated Objects (2023) | 入门直觉解读 |
| General In-Hand Object Rotation with Vision and Touch (2023) | 入门直觉解读 |
| LIBERO: Benchmarking Knowledge Transfer for Lifelong Robot Learning (2023) | 入门直觉解读 |
| ManiSkill2: A Unified Benchmark for Generalizable Manipulation Skills (2023) | 入门直觉解读 |
| PriorMDM: Human Motion Diffusion as a Generative Prior (2023) | 入门直觉解读 |
| RoboFlamingo: Vision-Language Foundation Models as Effective Robot Imitators (2023) | 入门直觉解读 |
| TWM: Transformer-based World Models Are Happy With 100k Interactions (2023) | 入门直觉解读 |
| Dynamic bipedal maneuvers through sim-to-real reinforcement learning (2022) | 入门直觉解读 |
| Imitate and Repurpose: Learning Reusable Robot Movement Skills from Human and Animal Behaviors (2022) | 入门直觉解读 |
| MotionCLIP: Exposing Human Motion Generation to CLIP Space (2022) | 入门直觉解读 |
| PI-QT-Opt: Predictive information improves multi-task robotic reinforcement learning at scale (2022) | 入门直觉解读直觉解读 |
| ProgPrompt: Generating Situated Robot Task Plans using Large Language Models (2022) | 入门直觉解读 |
| TEMOS: Generating Diverse Human Motions from Textual Descriptions (2022) | 入门直觉解读 |
| Tracking People by Predicting 3D Body, Pose and Location (PHALP) (2022) | 入门直觉解读 |
| CLIPort: What and where pathways for robotic manipulation (2021) | 入门直觉解读 |
| Masked Autoencoders Are Scalable Vision Learners (MAE) (2021) | 直觉解读 |
| Language Models are Few-Shot Learners (GPT-3) (2020) | 直觉解读 |
| Dream to Control: Learning Behaviors by Latent Imagination (2019) | 直觉解读 |
| DReCon: Data-Driven Responsive Control of Physics-Based Characters (2019) | 入门直觉解读 |
| Variational Discriminator Bottleneck: Improving Imitation Learning, Inverse RL, and GANs by Constraining Information Flow (2019) | 入门直觉解读直觉解读 |
| AC-VLA: Robust Out-of-Distribution Action Execution via Compositional Learning (2026) | 入门 |
| Adaptive Action Chunking at Inference-time for Vision-Language-Action Models (2026) | 入门 |
| Beyond Binary: Sim-to-Real Dexterous Manipulation with Physics-Grounded Contact Representation (2026) | 入门直觉解读 |
| BFM-Zero: Behavior Foundation Model with Zero-Shot RL via Forward-Backward Representations (2026) | 入门 |
| ChunkFlow: Towards Continuity-Consistent Chunked Policy Learning (2026) | 入门 |
| CoLA-Flow Policy: Temporally Coherent Imitation Learning via Continuous Latent Action Flow Matching for Robotic Manipulation (2026) | 入门 |
| Cosmos 3: An Open Frontier Omnimodal World Foundation Model for Physical AI (2026) | 直觉解读 |
| DenseReward: Dense Reward Learning via Failure Synthesis for Robotic Manipulation (2026) | 入门 |
| DreamZero: World Action Models are Zero-shot Policies (2026) | 入门直觉解读 |
| FoMoVLA: Bridging Visual Foresight and Motion Guidance for Vision-Language-Action Models (2026) | 入门 |
| Harness VLA: Steering Frozen VLAs into Reliable Manipulation Primitives via Memory-Guided Agents (2026) | 入门 |
| HELP: Human-Efficient Large-Scale Robot Post-Training with Rollout Segmentation (2026) | 入门 |
| Human-as-Humanoid: Enabling Zero-Shot Humanoid Learning from Ego-Exo Human Videos with Human-Aligned Embodiments (2026) | 入门 |
| HumDex: Humanoid Dexterous Manipulation Made Easy (2026) | 入门直觉解读 |
| Imagining the Sense of Touch: Touch-Informed Manipulation via Imagined Tactile Representations (2026) | 入门直觉解读 |
| KiMoDO: Scaling Controllable Human Motion Generation (2026) | 入门直觉解读 |
| LAP: Language-Action Pre-Training Enables Zero-shot Cross-Embodiment Transfer (2026) | 入门直觉解读 |
| LifeLong-RFT: Continual Learning VLA Models via Reinforcement Fine-Tuning (2026) | 入门 |
| Lights, Camera, Malfunction: When Illumination Robustness Leaves VLA Models Blind to Color (2026) | 入门 |
| Locomotion Beyond Feet (2026) | 入门直觉解读 |
| LongBench: Evaluating Robotic Manipulation Policies on Real-World Long-Horizon Tasks (2026) | 入门 |
| Mem-World: Memory-Augmented Action-Conditioned World Models for Persistent Robot Manipulation (2026) | 直觉解读 |
| MOSAIC: Bridging the Sim-to-Real Gap in Generalist Humanoid Motion Tracking and Teleoperation with Rapid Residual Adaptation (2026) | 入门 |
| MotionBricks: Scalable Real-Time Motions with Modular Latent Generative Model and Smart Primitives (2026) | 入门直觉解读 |
| OSCAR: Omni-Embodiment Action-Conditioned World Model for Robotics (2026) | 入门直觉解读 |
| RADAR: Closed-Loop Robotic Data Generation via Semantic Planning and Autonomous Causal Environment Reset (2026) | 入门 |
| ReActor: Reinforcement Learning for Physics-Aware Motion Retargeting (2026) | 入门直觉解读 |
| RealDexUMI: A Wearable Universal Manipulation Interface for Dexterous Robot Learning (2026) | 入门直觉解读 |
| Riemannian Motion Generation via Manifold-Valued Flow Matching (2026) | 入门直觉解读 |
| RIO: Flexible Real-Time Robot I/O for Cross-Embodiment Robot Learning (2026) | 入门 |
| RoboTTT: Context Scaling for Robot Policies (2026) | 入门 |
| Scalable and General Whole-Body Control for Cross-Humanoid Locomotion (XHugWBC) (2026) | 入门直觉解读 |
| SIMPLE: Simulation-Based Policy Learning and Evaluation for Humanoid Loco-manipulation (2026) | 入门 |
| Spherical Latent Motion Prior for Physics-Based Simulated Humanoid Control (SLMP) (2026) | 入门直觉解读 |
| TeleDex: Accessible Dexterous Teleoperation (2026) | 入门 |
| Toward Reliable Sim-to-Real Predictability for MoE-based Robust Quadrupedal Locomotion (2026) | 入门 |
| Towards Human-like Physical Intelligence: Lifelong Vision-Language-Action Learning for Robotic Manipulation (2026) | 入门 |
| Trajectory-Consistent Flow Matching for Robust Visuomotor Policy Learning (2026) | 入门 |
| TrustVLA: Mechanism-Guided Inference-Time Defense Against Vision-Language-Action Backdoors (2026) | 入门 |
| VISTA: Vision-Grounded and Physics-Validated Adaptation of UMI data for VLA Training (2026) | 入门 |
| World Action Models: The Next Frontier in Embodied AI (2026) | 入门直觉解读 |
| A Survey on Vision-Language-Action Models: An Action Tokenization Perspective (2025) | 入门直觉解读 |
| A Unified and General Humanoid Whole-body Controller for Versatile Locomotion (2025) | 入门 |
| Dexterous Manipulation through Imitation Learning: A Survey (2025) | 入门直觉解读 |
| DOGlove: Dexterous Manipulation with a Low-Cost Open-Source Haptic Force Feedback Glove (2025) | 入门 |
| DreamPolicy: A Scalable Unified Policy for Versatile Humanoid Locomotion (2025) | 入门 |
| GMR: General Motion Retargeting for Humanoid Motion Tracking (2025) | 入门 |
| HiLo: Learning Whole-Body Human-like Locomotion with Motion Tracking Controller (2025) | 入门 |
| HumanoidExo: Scalable Whole-Body Humanoid Manipulation via Wearable Exoskeleton Teleoperation (2025) | 入门 |
| LangWBC: Language-directed Humanoid Whole-body Control via End-to-end Learning (2025) | 入门 |
| ManiFlow: A General Robot Manipulation Policy via Consistency Flow Matching (2025) | 入门 |
| OmniRetarget: Interaction-Preserving Data Generation for Humanoid Whole-Body Loco-Manipulation and Scene Interaction (2025) | 入门 |
| OpenVLA-OFT: Fine-Tuning Vision-Language-Action Models with Effective and Efficient Multi-Token Action Tokens (2025) | 入门 |
| UniVLA: Unified Vision-Language-Action Model (2025) | 入门 |
| VLA-RFT: Vision-Language-Action Reinforcement Fine-tuning with Pre-trained Reward (2025) | 入门 |
| XR-1: Towards Versatile Vision-Language-Action Models via Learning Unified Vision-Motion Representations (2025) | 入门 |
| Zero-Shot Whole-Body Humanoid Control via Behavioral Foundation Models (2025) | 入门 |
| Advanced Expressive Humanoid Whole-Body Control (ExBody2) (2024) | 入门 |
| ALOHA Unleashed: A simple recipe for robot dexterity (2024) | 入门 |
| DexGraspNet 2.0: Learning Generative Dexterous Grasping in Large-scale Synthetic Cluttered Scenes (2024) | 入门 |
| DreMa: Dream to Manipulate — Compositional World Models Empowering Robot Imitation Learning with Imagination (2024) | 入门 |
| DROID: A Large-Scale In-The-Wild Robot Manipulation Dataset (2024) | 入门 |
| Harmon: Whole-body Motion Generation of Humanoid Robots from Language Descriptions (2024) | 入门 |
| ILSA: Incremental Learning for Robot Shared Autonomy (2024) | 入门 |
| Learning Visuotactile Skills with Two Multifingered Hands (2024) | 入门 |
| ManiCM: Real-time 3D Diffusion Policy via Consistency Model for Robotic Manipulation (2024) | 入门 |
| Open-TeleVision: Teleoperation with Immersive Active Visual Feedback (2024) | 入门 |
| Scaling Proprioceptive-Visual Learning with Heterogeneous Pre-trained Transformers (HPT) (2024) | 入门 |
| Simulated Humanoid Benchmark for Whole-Body Locomotion and Manipulation (HumanoidBench) (2024) | 入门 |
| TeleMoMa: A Modular and Versatile Teleoperation System for Mobile Manipulation (2024) | 入门 |
| AirExo: Low-Cost Exoskeletons for Learning Whole-Arm Manipulation in the Wild (2023) | 入门 |
| AnyTeleop: A General Vision-Based Dexterous Robot Arm-Hand Teleoperation System (2023) | 入门 |
| Learning Physically Simulated Tennis Skills from Broadcast Videos (2023) | 入门 |
| LIV: Language-image representations and rewards for robotic control (2023) | 入门 |
| Motion-X: A Large-scale 3D Expressive Whole-body Human Motion Dataset (2023) | 入门 |
| Real-World Humanoid Locomotion with Reinforcement Learning (2023) | 入门 |
| Safety Gymnasium: A Unified Safe Reinforcement Learning Benchmark (2023) | 入门 |
| ViNT: A foundation model for visual navigation (2023) | 入门 |
| VoxPoser: Composable 3D Value Maps for Robotic Manipulation with Language Models (2023) | 入门 |
| GAPartNet: Cross-Category Domain-Generalizable Object Perception and Manipulation via Generalizable and Actionable Parts (2022) | 入门 |
| Hora: In-Hand Object Rotation via Rapid Motor Adaptation (2022) | 入门 |
| IRIS: Transformers are Sample-Efficient World Models (2022) | 直觉解读 |
| Masked Visual Pre-training for Motor Control (2022) | 入门 |
| MotionBERT: Unified Pretraining for Human Motion Perception (2022) | 入门 |
| Perceiver-Actor: A Multi-Task Transformer for Robotic Manipulation (PerAct) (2022) | 入门 |
| QuestSim: Human Motion Tracking from Sparse Sensors with Simulated Avatars (2022) | 入门 |
| Robotic Telekinesis: Learning a Robotic Hand Imitator by Watching Humans on YouTube (2022) | 入门 |
| BABEL: Bodies, Action and Behavior with English Labels (2021) | 入门 |
| CALVIN: A Benchmark for Language-Conditioned Policy Learning for Long-Horizon Robot Manipulation Tasks (2021) | 入门 |
| Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning (2021) | 入门直觉解读 |
| Learning Visually Guided Latent Actions for Assistive Teleoperation (2021) | 入门 |
| Offline Reinforcement Learning as One Big Sequence Modeling Problem (Trajectory Transformer) (2021) | 入门 |
| What matters in learning from offline human demonstrations for robot manipulation (RoboMimic) (2021) | 入门 |
| CARL: Controllable Agent with Reinforcement Learning for Quadruped Locomotion (2020) | 入门 |
| Score-based Generative Modeling through Stochastic Differential Equations (2020) | 入门 |
| Learning agile and dynamic motor skills for legged robots (2019) | 入门直觉解读 |
| Neural Probabilistic Motor Primitives for Humanoid Control (2019) | 入门 |
| Solving Rubik's Cube with a Robot Hand (2019) | 入门 |
| ChauffeurNet: Learning to Drive by Imitating the Best and Synthesizing the Worst (2018) | 入门 |
| HG-DAgger: Interactive Imitation Learning with Human Experts (2018) | 入门 |
| Kickstarting Deep Reinforcement Learning (2018) | 入门 |
| ThriftyDAgger: Budget-aware novelty and risk gating for interactive imitation learning (2018) | 入门 |
| Constrained Policy Optimization (2017) | 入门 |
| Benchmarking Deep Reinforcement Learning for Continuous Control (2016) | 入门 |
| Learning Dexterous Manipulation Policies from Experience and Imitation (2016) | 入门 |
其它（88 篇 · 有教学或解读页的其余工作）
| 论文 | 可用版本（点击进入） |
|---|---|
| Ambient Diffusion Policy: Imitation Learning from Suboptimal Data in Robotics (2026) | 入门 |
| Behavior Foundations for Quadruped Robots: ABot-C0 Technical Report (2026) | 入门 |
| DexVerse: A Modular Benchmark for Multi-Task, Multi-Embodiment Dexterous Manipulation (2026) | 入门 |
| Dual-Process Atomic Skill Learning: Decoupling Semantic Reasoning and Real-Time Control (2026) | 入门 |
| E-TTS: A New Embodied Test-Time Scaling Framework for Robotic Manipulation (2026) | 入门 |
| From World Action Models to Embodied Brains: A Roadmap for Open-World Physical Intelligence (2026) | 入门 |
| FTP-1: A Generalist Foundation Tactile Policy Across Tactile Sensors for Contact-Rich Manipulation (2026) | 入门 |
| Generalizable VLA Finetuning via Representation Anchoring and Language-Action Alignment (2026) | 入门 |
| Humanoid Manipulation Interface: Humanoid Whole-Body Manipulation from Robot-Free Demonstrations (2026) | 入门 |
| JoyNexus: Service-Oriented Multi-Tenant Post-Training for VLA Models (2026) | 入门 |
| NavCMPO: Critic-Guided MeanFlow Policy Optimization for Adaptive Navigation (2026) | 入门 |
| Never Too Late for Force: Accelerating VLA Post-Training with Reactive Force Injection (2026) | 入门 |
| One Demonstration Is Enough for Real-World Robotic Reinforcement Learning (2026) | 入门 |
| Perceptive Behavior Foundation Model: Adapting Human Motion Priors to Robot-Centric Terrain (2026) | 入门 |
| PointWorld: Scaling 3D World Models for In-The-Wild Robotic Manipulation (2026) | 入门 |
| Reflex: Real-Time VLA Control through Streaming Inference (2026) | 入门 |
| See like a Robot: Robot-Centric Pointmaps for Vision-Language-Action Models (2026) | 入门 |
| SegDiff: Segmented Trajectory Diffusion for Consistent and Adaptive Robot Manipulation (2026) | 入门 |
| Stubborn: A Streamlined and Unified Reinforcement Learning Framework for Robust Motion Tracking and Fall Recovery for Humanoids (2026) | 入门 |
| Training Vision-Language-Action Models with Dense Embodied Chain-of-Thought Supervision (2026) | 入门 |
| VIA: Visual Interface Agent for Robot Control (2026) | 入门 |
| VistaVLA: Geometry- and Semantic-Aware 3D Gaussian-Grounded VLA for Robotic Manipulation (2026) | 入门 |
| WOLF-VLA: Whole-Body Humanoid Optimal Locomotion Framework for Vision-Language-Action Learning (2026) | 入门 |
| WSA$_1$: a 3D-Centric World-Spatial-Action Model for Generalizable Robot Control (2026) | 入门 |
| X-Loco: Towards Generalist Humanoid Locomotion Control via Synergetic Policy Distillation (2026) | 入门 |
| Zero2Skill: Bootstrapping Robot Skills through Autonomous Data Collection, Training, and Deployment (2026) | 入门 |
| π0.7: A Steerable Generalist Robotic Foundation Model (2026) | 入门 |
| SmolVLA: A Vision-Language-Action Model for Affordable and Efficient Robotics (2025) | 入门 |
| Unleashing Large-Scale Video Generative Pre-training for Visual Robot Manipulation (2023) | 入门 |
| Habitat: A Platform for Embodied AI Research (2019) | 入门 |
| One-shot imitation from observing humans via domain-adaptive meta-learning (2018) | 入门 |
| Learning Complex Dexterous Manipulation with Deep Reinforcement Learning and Efficient Planning (2017) | 入门 |
| Action QFormer: Structured Representation Shaping under Action Supervision in Vision-Language-Action Models (2026) | 入门 |
| Action-to-Action Flow Matching (2026) | 入门 |
| ADP: Adversarial Dynamics Priors for Physically Grounded Humanoid Locomotion (2026) | 入门 |
| AeroAct: Action-Centered World-Action Models for Language-Conditioned Quadrotor Flight (2026) | 入门 |
| AnyDexRT: Calibration-Free Dexterous Hand Retargeting with Few-Shot Human Guidance (2026) | 入门 |
| D-SafeMPC: Diffusion-Driven Safe Model Predictive Control with Discrete-Time Control Barrier Functions (2026) | 入门 |
| DexCompose: Reusing Dexterous Policies for Multi-Task Manipulation with a Single Hand (2026) | 入门 |
| From Grasps to Dexterity: Large-Scale Grasp Pretraining for Dexterous Manipulation (2026) | 入门 |
| GaitSpan: Growing Humanoid Locomotion from Walking to Running (2026) | 入门 |
| Handroid: Bridging Dexterous Hand and Humanoid (2026) | 入门 |
| Image2Sim: Scaling Embodied Navigation via Generative Neural Simulator (2026) | 入门 |
| Jetson-PI: Towards Onboard Real-Time Robot Control via Foresight-Aligned Asynchronous Inference (2026) | 入门 |
| Learning Dexterous Manipulation Using Contact Wrench Guidance From Human Demonstration (2026) | 入门 |
| MIDAS Hand: Modular low-Impedance Direct-drive Anthropomorphic Sensing Hand (2026) | 入门 |
| More Structure, Not More Capacity: Object-Centric Representations for Visuomotor Imitation Learning (2026) | 入门 |
| Multi-Gait Learning for Humanoid Robots Using Reinforcement Learning with Selective Adversarial Motion Prior (2026) | 入门 |
| Physics-Guided Biomechanical Gait Adaptation for Humanoid Locomotion on Extreme Sloped Terrains (2026) | 入门 |
| Pix2Act: Image-Space Manipulation Policies with Equivariant Augmentation (2026) | 入门 |
| RoboSnap: One-Shot Real-to-Sim Scene Generation for Generalizable Robot Learning and Evaluation (2026) | 入门 |
| Scaling Behavior Foundation Model for Humanoid Robots (2026) | 入门 |
| Self-Correcting VLA: Online Action Refinement via Sparse World Imagination (2026) | 入门 |
| Source-Lifted Flow Matching for Intervenable Multimodal Imitation (2026) | 入门 |
| TouchWorld: A Predictive and Reactive Tactile Foundation Model for Dexterous Manipulation (2026) | 入门 |
| Towards Predictive, Aligned, and Scalable Robot Learning (2026) | 入门 |
| UniTacVLA: Unified Tactile Understanding and Prediction in Vision Language Action Models (2026) | 入门 |
| Worlds in One Demo: A Synthetic Data Engine for Learning Open-World Mobile Manipulation (2026) | 入门 |
| WorldSample: Closed-loop Real-robot RL with World Modelling (2026) | 入门 |
| BC-Z: Zero-Shot Task Generalization with Robotic Imitation Learning (2022) | 入门 |
| SAPIEN: A SimulAted Part-based Interactive ENvironment (2020) | 入门 |
| RLBench: The Robot Learning Benchmark & Learning Environment (2019) | 入门 |
| A Minimalist Retargeting-Guided Reinforcement Learning Recipe for Dexterous Manipulation (2026) | 入门 |
| CLEAR: Closed-Loop Reinforcement Learning at Scale for End-to-End Autonomous Driving (2026) | 入门 |
| Current as Touch: Proprioceptive Contact Feedback for Compliant Dexterous Manipulation (2026) | 入门 |
| Dexora: Open-source VLA for High-DoF Bimanual Dexterity (2026) | 入门 |
| Dexterous Point Policy: Learning Point-based Dexterous Hand Policies from Human Demonstrations (2026) | 入门 |
| Exploratory, Communicative, and Deployable: Vision-Driven Embodied Agents for Open-World Mobile Manipulation (2026) | 入门 |
| ExToken: Structured Exploration for Efficient Vision-Language-Action Reinforcement Fine-tuning (2026) | 入门 |
| GraspGen-X: Cross-Embodiment 6-DOF Diffusion-based Grasping (2026) | 入门 |
| GraspGraphNet: Graph-Structured Multi-Embodiment Dexterous Grasp Generation (2026) | 入门 |
| Hand-in-the-Loop: Improving VLA Policies for Dexterous Manipulation via Seamless Hand-Arm Intervention (2026) | 入门 |
| In vivo feasibility study of humanoid robots in surgery (2026) | 入门 |
| Labimus: A Simulation and Benchmark for Humanoid Dexterous Manipulation in Chemical Laboratory (2026) | 入门 |
| LAMP: Latent Motion Prior-Guided Real-World Learning for Dexterous Hand Manipulation (2026) | 入门 |
| MPC-Injection: Biasing Off-Policy Locomotion RL Toward Controller-Induced Behavior Basins (2026) | 入门 |
| NeuralActuator: Neural Actuation Modeling for Robot Dynamics and External Force Perception (2026) | 入门 |
| OASIS: From Simulation Data Collection to Real-World Humanoid Loco-Manipulation (2026) | 入门 |
| OmniTacTune: Policy-Agnostic Real-World RL for Tactile Residual Adaptation of Visual Policies (2026) | 入门 |
| RoboVista: Evaluating Vision Language Models for Diverse Robot Applications (2026) | 入门 |
| RynnWorld-Teleop: An Action-Conditioned World Model for Digital Teleoperation (2026) | 入门 |
| SILO: Simulation-in-the-Loop Sim-to-Real Transfer for Multi-Stage Cable Routing (2026) | 入门 |
| Switch: Learning Agile Skills Switching for Humanoid Robots (2026) | 入门 |
| T-Rex: Tactile-Reactive Dexterous Manipulation (2026) | 入门 |
| TAC-LOCO: Unified Whole-Body Control for Quadrupedal TACtile-Informed LOCO-Manipulation (2026) | 入门 |
| Towards Human-level Dexterous Teleoperation (2026) | 入门 |
| Walk the PLANC: Physics-Guided RL for Agile Humanoid Locomotion on Constrained Footholds (2026) | 入门 |
| Video PreTraining (VPT): Learning to Act by Watching Unlabeled Online Videos (2022) | 入门 |
🧠 概念辨析（concepts/）
已完成 13 篇概念辨析，覆盖数学基础、方法论、工程实务。
| 文件 | 内容 | 字符 |
|---|---|---|
| C-sim2real-methods.md | DR / System ID / Teacher-Student / RMA 四法辨析 | 20k |
| C-action-representations.md | torque/PD/token/flow/chunk/latent 七种 action 表示 | 7.5k |
| C-rl-algorithms.md | PPO/SAC/TD3/DDPG/DreamerV3 五大算法家族（含 PG 定理推导、on/off-policy 数学定义、机器人适用场景表） | 13.6k |
| C-mdp-pomdp-formulation.md | MDP/POMDP/机器人形式化（五元组、Bellman、belief state、history-based policy、与 LQR/MPC 的统一视角） | 12k |
| C-imitation-learning-theory.md | IL 理论基础（BC MLE、covariate shift $O(\epsilon T^2)$、DAgger、GAIL/AMP 对抗式 IL、多模态动作、distribution vs trajectory matching） | 12.6k |
| C-optimization-and-loss-functions.md | 优化与损失函数谱系（likelihood-ratio、Bellman 残差、actor-critic variance reduction、contrastive、flow matching、PBRS reward shaping 不变性） | 14.3k |
| C-multimodal-disambiguation.md | 「多模态」三义辨析（多模态感知/输入 vs 多峰动作分布 vs 多源数据；MSE 失败的数学、模态对齐、判别手册） | 11k |
| C-distillation-methods.md | 各种 distillation 辨析（privileged/teacher-student vs policy 压缩 vs feature 中间表征 vs online 共演化；AMP、RMA、Consistency Policy、HOVER 定位） | 15k |
| C-diffusion-flow-matching.md | Diffusion / Flow Matching / Score Matching 的数学统一（同一连续时间 ODE/SDE 的三种路径；π0 为何选 flow） | 5k |
| C-humanoid-hardware-platforms.md | 人形硬件平台四档对比（剧烈运动 / 学术主流 / 开源自研 / 闭源工业；DoF、力矩、价格、开源程度） | 10k |
| C-representation-learning.md | 机器人表征学习三路线（通用视频预训练 / 仿真域随机化 / 视觉-动作联合 world model） | 8k |
| C-rl-to-control-bridge.md | RL 与传统控制的统一视角（LQR/MPC/WBC vs RL；目标函数、模型来源、求解方式三轴对齐） | 8k |
| C-safety-and-deployment.md | 真机部署的安全与工程（空载→悬空→低速→正常→长期→无人值守六层递进；硬件保护、故障恢复） | 9k |
📅 持续收集日志（log/）
log/CHANGELOG.md—— 按日期的变更日志（详见 2026-07-18 的多轮推进）
🔬 知识库分析层（站点投影）
分析层（对论文做引用分析 / 统计 / 查漏 / 交叉学习）——可复现脚本，改数据→重跑：
scripts/analyze_kb.py→data/curated/analytics.md：七维体检（元数据完整度、影响力工作榜、引用图健康、跨方向引用矩阵、覆盖度、查漏补缺汇总）。scripts/normalize_citations.py→citations_normalized.jsonl(2147 边 / 867 节点) +influence.jsonl(含 PageRank)：把 builds_on/leads_to 的模糊短名规范化为 canonical id。- 影响力 hub（被引最多）：OmniH2O · OpenVLA · AMP · HOVER · RT-2 · MDM · PHC · OpenX · Diffusion Policy · π0。
站点投影（面向学习者）：site/ 下
analytics.html—— 知识库分析报告（影响力榜 / 缺口 / 跨方向矩阵）citation-graph.html—— 交互式引用图（节点大小=PageRank，颜色=方向，悬停看详情）search.html—— 1,488 个条目（离线站内搜索）- 导航单一数据源
data/curated/nav_manifest.json（侧边栏全部由它派生）
🗺️ 全景
roadmap.md—— 全景脉络总览（已有，会随线索深化持续更新）
如何贡献新内容
- 新发现一篇重要论文：在
research/papers/下写一篇P-名称-年份.md深度笔记，更新对应线索threads/TXX，在log/CHANGELOG.md记一条 - 发现一条新线索：建
threads/TXX-*.md（参考_template.md），更新本 INDEX - 辨析一个概念：写
concepts/C-*.md - 更新已有线索：直接编辑，但在
log/CHANGELOG.md记录改了什么、为什么
site/ 的 HTML 渲染可以隔一段时间批量做（有脚本），不是每次都要同步。