TLA: Tactile-Language-Action
本笔记基于摘要 + 公开资料,未读全文。
一句话讲什么(TL;DR)
让机器人像你闭眼摸钥匙那样——靠"一段持续的触感"加上一句话指令,自己决定下一步该怎么用手。
这是个什么场景 — 日常类比
早上你在背包侧袋里掏钥匙,眼睛还盯着手机。
手指伸进去那两秒,发生了一连串事:先碰到软的纸巾(不是)、滑过塑料壳(耳机盒)、最后摸到一串凉凉的、有齿的金属——是它。整个过程你没看,靠的是一段连续变化的手感:从软到硬、从光滑到带齿、从晃到稳。
机器人之前没法做这件事。要么靠摄像头看(书包里全黑啥也看不见),要么靠传感器"按一下记一张压力图"——相当于只允许你用指尖戳一下不准滑动,自然分不清耳机盒和钥匙串。TLA 想让机器人也能"摸着摸着就知道是哪个",并且当你嘴上说"把软的那个递给我"时,它知道软的是哪种触感曲线,伸手去拿对的那个。

之前的人怎么做的 — 3-5 bullet
- VLA 路线(RT-2、OpenVLA 等):视觉 + 语言 → 动作。但摄像头看不到接触瞬间发生了什么,比如东西滑了没、捏到没。
- 单帧触觉(TacGNN、各类 GelSight 工作):把触觉传感器读数当成一张图片,识别接触面形状或物体类别。问题:丢了时间维度,捏苹果和捏鸡蛋的"渐进施力差异"看不出来。
- 触觉 + 强化学习:用触觉信号当 reward 或 state 去训 policy,但不接语言、泛化差,换个任务就要重训。
- 多模态融合的早期尝试:把触觉特征和视觉特征 concat,但没有大语言模型那种"指令理解"能力,做不到"把熟的桃子递给我"这种语义级任务。
- **少有工作把触觉时序当成"模态"**和 LLM 对齐——这是 TLA 切入的缝。
这篇论文的关键想法
一句话:把触觉当成大模型听得懂的另一门"方言",而且是带时间的方言。
打个比方。大模型已经会听人说话(语言)、看图(视觉)。现在再教它一门新语言——触觉。但不是教它"指尖压力 = 5 牛"这种死数字(那相当于教它单词),而是教它**"接触瞬间这条压力上升曲线长什么样"**——更像教它听一段语调,而不是孤立的字。
具体步骤:
- 触觉传感器输出的不是一张图,而是连续 T 帧的时序信号(类似一小段视频)。
- 一个 encoder 把这段信号压成一串 token,相当于把"手感片段"翻译成大模型能读的"词"。
- 这些"触觉词"和你说的话(文本 token)、要做的动作(动作 token)混在一起,喂给一个 transformer/LLM 主干。
- 训练目标:看完这段触感 + 听完这句话 → 输出下一步该怎么动手。
精神跟 VLA(Vision-Language-Action,视觉-语言-动作)一样,只是把"看"换成或加上"摸",而且强调时间序列——不是单帧压力贴图,是一整段手感曲线。这让模型能区分"刚碰到时力在变" vs "已经握稳了力不变"这种只有时间维度才看得出的差别。

它怎么做的(方法)— 3-4 段
触觉编码。像把一段 30 秒的视频剪成 5 个关键画面再写成字幕——给 LLM 看的不是原始流,而是"摘要"。论文用某种序列编码器(可能是 1D conv + transformer,或者 ViT 风格处理时序)把多帧触觉信号转成一段嵌入向量。等等,先慢一拍——嵌入向量(embedding) 是什么?就是把一段东西压成一串数字,使得"长得像的东西"数字也接近,这样机器能算距离、做匹配。具体编码器结构和帧数需读原文确认。
跨模态对齐。像翻译官的工作:让"硬"这个中文词、"hard"这个英文词、还有手摸到硬东西的那种触感曲线——三种来源的"嵌入"都指向同一个意思。常见做法是对比学习(contrastive,类似 CLIP,让配对的样本靠近、不配对的远离)。这样语言和触觉就能互相检索、互相条件化——你说"凉的",模型能想起对应的触感长什么样。
动作解码。前面把语言和触觉对齐了,现在要"动手"。像厨师看完菜单(语言)、捏过食材(触觉)后决定下刀的角度——接一个 action head(可能是 diffusion policy 或 autoregressive token 输出),根据"语言指令 + 触觉时序 + 可能的视觉"联合预测末端执行器的动作序列。这部分基本沿用 VLA 范式。
数据。这是最难的一关。触觉数据像稀有食材——必须有装着触觉传感器的真机械臂去一次次摸真实物体,还得给每段触感配上"我现在摸的是 XX"这种语言标注。论文应该会构造或借助某个 tactile-language pairing 的数据集;具体规模和采集方式需读原文。可能也会做 sim-to-real(仿真训、真机用)或者合成数据扩量。
实验在做什么
典型的实验维度(具体数字需读原文):
- 下游任务:精细操作类,比如分辨软硬、判断滑动、精确插入、抓取易碎物。
- 基线对比:仅视觉 VLA、仅触觉 policy、单帧触觉 + 语言。比 TLA 的"序列触觉 + 语言"差多少。
- 消融:去掉时序(只用单帧)、去掉语言、换不同长度的触觉窗口,分别看性能掉多少。
- 泛化:训练时没见过的物体形状或材质,能不能用语言描述零样本迁移。
- 真机部署:是不是只在仿真里跑,还是有真实机械臂的视频和成功率。
读论文时重点看消融——能证明"序列性"比"单帧"贡献大才说得过去标题里的 sequence。
你应该懂的几个新词 — 4-6 个
- VLA(Vision-Language-Action):把视觉、语言、动作三个模态联合训的模型范式。RT-2、OpenVLA 是代表作。TLA 是这个范式把"视觉"换成或扩展为"触觉"的版本。
- GelSight / 视触觉传感器:用一块软胶 + 摄像头记录胶面形变的传感器。输出形式像图像,但描述的是接触压力分布。
- 时序触觉(sequential tactile):不是单帧压力图,是一段时间内连续的触觉读数。类比视频 vs 图片。
- 跨模态对齐(cross-modal alignment):让不同模态(语言、视觉、触觉)的向量住进同一空间,用对比学习等方法实现。CLIP 是经典案例。
- action token / action head:把连续动作(关节角度、末端位置)离散化成 token,或者用单独的小网络解码动作向量。VLA 系列的标准做法。
- sim-to-real:在仿真里训,部署到真机。触觉 sim-to-real 比视觉更难,因为接触物理仿真不准。
它和其他论文什么关系
- OpenVLA / RT-2:TLA 是同一家族的"换模态版本"。理解了 VLA 怎么把图像 token 化喂大模型,TLA 就懂了一半。
- 3D-VLA / PointLLM:都是给 VLA 加新模态。3D-VLA 加点云,TLA 加触觉,思路并列。
- Diffusion Policy / 3D Diffusion Policy:这些是动作解码端的工作。TLA 的 action head 可能借鉴。
- 触觉表示学习(如 MViTac、T3):这些做触觉自监督预训练,可能是 TLA 触觉 encoder 的前置工作或对比基线。
- 多模态 LLM 综述(如 LLaVA 系列):TLA 是把"触觉"加进多模态 LLM 大盘子里的一个具体落地。
我建议这样读 — 3-4 步
- 先扫摘要 + 方法图(一定有一张系统总览图),搞清楚"触觉时序怎么进 LLM"——这是全文骨架。
- 跳到实验消融,看"序列 vs 单帧"差多少。如果序列贡献小,标题就有点虚;贡献大,那这工作就真有价值。
- 看数据章节,搞清楚 tactile-language pair 怎么来的。这是触觉领域的瓶颈,谁能解决数据谁就赢一半。
- 最后回头看 related work,跟 OpenVLA 等 VLA 工作做对比,理解 TLA 在范式上加了什么减了什么。
为什么值得读
触觉是机器人最被低估的模态。视觉能告诉你"看到了什么",但抓东西最后那 5cm、捏软硬、判断滑动,全靠触觉。把触觉用上 LLM 范式(语言条件化 + 时序建模)是个明显该做但很难做的方向,因为数据贵、传感器多样、仿真不准。
TLA 把"序列性"作为关键词推出来,本身就是对触觉建模的一个重要 framing——之前太多工作把触觉当图片处理,浪费了时间维度。即使方法本身的工程细节不一定立刻能复现,这个 framing + VLA 范式迁移的思路值得了解,是 embodied AI 多模态扩展的一个标志性节点。
适合读完 OpenVLA、对 VLA 范式熟悉之后,作为"如何给 VLA 加新模态"的参考案例来读。
◼
引用本笔记 / Cite this note
@online{eai_tla_tactile_language_action_2026,
title = {(readable note) TLA: Tactile-Language-Action},
author = {Zhou, Jason},
year = {2026},
note = {Note on a 2025 paper},
howpublished = {\url{https://estelledc.github.io/embodied-ai-reading-station/papers/tla-tactile-language-action/}},
organization = {Embodied AI Reading Station}
}
All 156 papers (full index)
- 1. LLaVA: Visual Instruction Tuning
- 2. 3DShape2VecSet: 3D Shape Representation for Diffusion Models
- 3. SayCan: Do As I Can, Not As I Say
- 4. OpenVLA: An Open-Source Vision-Language-Action Model
- 5. VLAS: VLA Model With Speech Instructions
- 6. MLA: Multisensory Language-Action Model
- 7. Cosmos Policy: Fine-Tuning Video Models for Visuomotor Control
- 8. CartoRadar: RF-Based 3D SLAM Rivaling Vision Approaches
- 9. mmCLIP: Boosting mmWave-based Zero-shot HAR via Signal-Text Alignment
- 10. mmNorm: Non-Line-of-Sight 3D Object Reconstruction via mmWave Surface Normal Estimation
- 11. Proactive Hearing Assistants that Isolate Egocentric Conversations
- 12. NeuralAids: Wireless Hearables With Programmable Speech AI Accelerators
- 13. Creating speech zones with self-distributing acoustic swarms
- 14. Conv-TasNet: Surpassing Ideal Time-Frequency Magnitude Masking for Speech Separation
- 15. SoundStream: An End-to-End Neural Audio Codec
- 16. AudioLM
- 17. Conformer
- 18. Dual-path RNN
- 19. EnCodec
- 20. Meta-StyleSpeech
- 21. MusicLM
- 22. Robust Speech Recognition via Large-Scale Weak Supervision
- 23. SeamlessM4T
- 24. Stable Audio
- 25. Universal Source Separation with Weakly Labelled Data
- 26. Meta-World: A Benchmark and Evaluation for Multi-Task and Meta Reinforcement Learning
- 27. RLBench: The Robot Learning Benchmark & Learning Environment
- 28. robosuite: A Modular Simulation Framework and Benchmark for Robot Learning
- 29. BridgeData V2
- 30. CALVIN
- 31. LIBERO
- 32. RH20T
- 33. What Matters in Learning from Offline Human Demonstrations for Robot Manipulation
- 34. DROID
- 35. Open X-Embodiment
- 36. RoboCasa
- 37. SimplerEnv
- 38. Diffusion Policy: Visuomotor Policy Learning via Action Diffusion
- 39. 3D Diffusion Policy: Generalizable Visuomotor Policy Learning via Simple 3D Representations
- 40. Consistency Policy: Accelerated Visuomotor Policies via Consistency Distillation
- 41. EquiBot: SIM(3)-Equivariant Diffusion Policy
- 42. DiT-Policy
- 43. Diffusion Policy Policy Optimization (DPPO)
- 44. Affordance-based Robot Manipulation with Flow Matching
- 45. FlowPolicy: 3D Flow-based Policy via Consistency Flow Matching
- 46. FAST: Efficient Action Tokenization for VLA
- 47. pi_0: Vision-Language-Action Flow Model
- 48. pi_0.5: VLA with Open-World Generalization
- 49. A Reduction of Imitation Learning and Structured Prediction to No-Regret Online Learning
- 50. Generative Adversarial Imitation Learning
- 51. Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardware (ACT/ALOHA)
- 52. AnyTeleop
- 53. Behavior Transformers: Cloning k Modes with One Stone
- 54. Implicit Behavioral Cloning
- 55. RoboCat
- 56. ALOHA 2
- 57. DexCap
- 58. HumanPlus
- 59. Generalizable Humanoid Manipulation with 3D Diffusion Policies (iDP3)
- 60. Mobile ALOHA
- 61. SmolVLA
- 62. Universal Manipulation Interface
- 63. Behavior Generation with Latent Actions (VQ-BeT)
- 64. ImageBind: One Embedding Space To Bind Them All
- 65. Connecting Touch and Vision via Cross-Modal Prediction
- 66. AnyMAL: An Efficient and Scalable Any-Modality Augmented Language Model
- 67. AudioPaLM
- 68. FROMAGe: Grounding LLMs to Images
- 69. OneLLM
- 70. X-VLM: Multi-Grained Vision Language Pre-Training
- 71. Tactile Beyond Pixels (Sparsh-X)
- 72. Sparsh: Self-supervised Touch Representations
- 73. Tactile-VLA
- 74. TLA: Tactile-Language-Action
- 75. Code as Policies: Language Model Programs for Embodied Control
- 76. Inner Monologue: Embodied Reasoning through Planning with Language Models
- 77. LLM+P: Empowering LLMs with Optimal Planning
- 78. PaLM-E: An Embodied Multimodal Language Model
- 79. ProgPrompt
- 80. ChatGPT for Robotics
- 81. GenSim
- 82. RoboFlamingo
- 83. Tree-Planner
- 84. VoxPoser
- 85. See Through Smoke: Robust Indoor Mapping with Low-cost mmWave Radar
- 86. Can WiFi Estimate Person Pose?
- 87. 3DRIMR: 3D Reconstruction and Imaging via mmWave Radar based on Deep Learning
- 88. milliEgo: Single-chip mmWave Radar Aided Egomotion Estimation via Deep Sensor Fusion
- 89. High Resolution Point Clouds from mmWave Radar
- 90. RadarSLAM: Radar based Large-Scale SLAM in All Weathers
- 91. Through-Wall Pose Imaging in Real-Time with a Many-to-Many Encoder/Decoder Paradigm
- 92. RFMask: A Simple Baseline for Human Silhouette Segmentation with Radio Signals
- 93. RFPose-OT: RF-Based 3D Human Pose Estimation via Optimal Transport Theory
- 94. Argus: Multi-View Egocentric Human Mesh Reconstruction Based on Stripped-Down Wearable mmWave Add-on
- 95. Diffusion Model is a Good Pose Estimator from 3D RF-Vision
- 96. Enabling Visual Recognition at Radio Frequency (PanoRadar)
- 97. Wave-Former: Through-Occlusion 3D Reconstruction via Wireless Shape Completion
- 98. Habitat: A Platform for Embodied AI Research
- 99. Isaac Gym: High Performance GPU-Based Physics Simulation For Robot Learning
- 100. DexMV
- 101. Habitat 2.0
- 102. ManiSkill
- 103. ProcTHOR
- 104. SAPIEN: A SimulAted Part-based Interactive ENvironment
- 105. BEHAVIOR-1K
- 106. Habitat 3.0
- 107. Isaac Lab
- 108. MuJoCo Playground
- 109. RT-1: Robotics Transformer for Real-World Control at Scale
- 110. 3D Diffusion Policy (DP3)
- 111. Octo: An Open-Source Generalist Robot Policy
- 112. RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Control
- 113. RT-Trajectory: Robotic Task Generalization via Hindsight Trajectory Sketches
- 114. 3D-VLA
- 115. DexVLA
- 116. GR-2: Generative Video-Language-Action Model
- 117. OpenHelix
- 118. OpenVLA-OFT
- 119. RDT-1B: Diffusion Foundation Model for Bimanual Manipulation
- 120. RoboMamba
- 121. SpatialVLA
- 122. TinyVLA
- 123. TraceVLA: Visual Trace Prompting
- 124. Learning Transferable Visual Models From Natural Language Supervision
- 125. Flamingo: a Visual Language Model for Few-Shot Learning
- 126. BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models
- 127. BLIP: Bootstrapping Language-Image Pre-training for Unified Vision-Language Understanding and Generation
- 128. DeepSeek-VL: Towards Real-World Vision-Language Understanding
- 129. EVA-CLIP: Improved Training Techniques for CLIP at Scale
- 130. FILIP: Fine-grained Interactive Language-Image Pre-Training
- 131. Florence-2: Advancing a Unified Representation for a Variety of Vision Tasks
- 132. InternVL: Scaling up Vision Foundation Models and Aligning for Generic Visual-Linguistic Tasks
- 133. Improved Baselines with Visual Instruction Tuning
- 134. OBELICS
- 135. Qwen-VL: A Versatile Vision-Language Model for Understanding, Localization, Text Reading, and Beyond
- 136. Sigmoid Loss for Language Image Pre-Training
- 137. What matters when building vision-language models?
- 138. Expanding Performance Boundaries of Open-Source Multimodal Models with Model, Data, and Test-Time Scaling
- 139. The Llama 3 Herd of Models
- 140. LLaVA-NeXT-Interleave
- 141. LLaVA-OneVision: Easy Visual Task Transfer
- 142. Long-CLIP: Unlocking the Long-Text Capability of CLIP
- 143. Pixtral 12B
- 144. Dream to Control: Learning Behaviors by Latent Imagination
- 145. World Models
- 146. DayDreamer
- 147. Mastering Atari with Discrete World Models
- 148. Dreamer V3: Mastering Diverse Domains through World Models
- 149. Transformers are Sample-Efficient World Models
- 150. TWM: Transformer-based World Models
- 151. 1X World Model Challenge
- 152. Cosmos World Foundation Model Platform
- 153. GAIA-1
- 154. Genie: Generative Interactive Environments
- 155. Navigation World Models
- 156. UniSim