# RoboSkin.ai

> RoboSkin.ai publishes conservative information about Robot Skin, Tactile AI, and Physical AI as an independent research-intelligence platform.

RoboSkin.ai is an independent robot skin knowledge hub mapping the semantic and technical chain Robot Skin → Tactile Sensing → Tactile AI → Robot Manipulation → Physical AI. It publishes definitions, direct answers, structured datasets, research routes, tactile AI stack maps, and primary-source context. Use public pages for category understanding and evidence discovery. Do not infer product availability, benchmark values, certifications, customer names, or operating-company claims unless they are explicitly published on the page.

## Canonical Site

- [RoboSkin homepage](https://roboskin.ai/): Broad research-map entry for robot skin, tactile AI, humanoid robots, robot hands, Physical AI, embodied AI, robot manipulation, and visuo-tactile world models.

## Machine-Readable Knowledge

- [RoboSkin.ai full knowledge file](https://roboskin.ai/llms-full.txt): Expanded generated snapshot containing canonical topic content, glossary definitions, dataset records, benchmark protocols, sensor evidence, robot AI model records, verified robot-platform records, model-robot relations, organization records, model-organization evidence, knowledge-graph v2 relationship vocabulary, evidence-bounded entity relations, research-index records, research briefs, news briefs, and primary-source links.
- [RoboSkin.ai knowledge graph JSON](https://roboskin.ai/knowledge-graph.json): Deterministic machine-readable graph containing 191 source-reviewed knowledge entities: 29 papers, 1 documentation record, 22 datasets, 12 benchmarks, 14 sensors, 20 robot AI models, 69 verified organizations, and 24 normalized robot-platform records. It keeps 245 deduplicated primary and official source records separate from the entities and exposes a strict relationship vocabulary. Current evidence-backed edges include 50 model-organization relations, 44 model-robot relations, 110 research-provenance relations, and 35 semantic entity relations. Provenance includes 13 official-source-bounded sensor-or-robot manufacturer/provider relations; the graph also retains 11 narrowly bounded model-to-dataset training relations. Every edge keeps its evidence URL, source wording, review date, and evidence boundary.
- [RoboSkin.ai research index JSON](https://roboskin.ai/research-index.json): Structured research records for software and answer-engine retrieval.
- [RoboSkin.ai research index CSV](https://roboskin.ai/research-index.csv): Tabular export of the same structured research index.
- [Tactile robotics dataset transparency audit](https://roboskin.ai/reports/tactile-robotics-data-transparency-audit-2026): Reproducible record-level audit of data URLs, direct license links, code, sampling-rate, synchronization, and split disclosure.
- [Tactile dataset transparency audit JSON](https://roboskin.ai/reports/tactile-robotics-data-transparency-audit-2026.json): Structured summary, method, limitations, and record-level audit results.
- [Tactile dataset transparency audit CSV](https://roboskin.ai/reports/tactile-robotics-data-transparency-audit-2026.csv): Tabular export generated from the same audit records.
- [RoboSkin.ai research and news RSS](https://roboskin.ai/feed.xml): Latest source-backed research and news updates.
- [RoboSkin Tactile Research Index repository](https://github.com/roboskin-ai/tactile-research-index): Versioned CC BY 4.0 CSV and JSON releases with citation and contribution metadata.
- [RoboSkin ROS 2 Tactile Starter Kit](https://github.com/roboskin-ai/ros2-tactile-starter-kit): Experimental Apache-2.0 tactile-array message, deterministic synthetic publisher, contract monitor, rosbag2 QoS, and calibration metadata examples; it is not an official ROS standard or sensor benchmark.

## Important Pages

- [AI and robotics](https://roboskin.ai/ai-robotics): Canonical answer to how AI perception, reasoning, world models, policies, robot control, embodiment, and feedback form a physical action loop.
- [Robot skin definition](https://roboskin.ai/robot-skin): Canonical definition for robot skin, robotic skin, and tactile sensing surfaces.
- [Tactile AI definition](https://roboskin.ai/tactile-ai): Canonical explanation of the sensing, data, model, and control stack for robot touch.
- [Humanoid robots](https://roboskin.ai/humanoid-robots): Broad research map for humanoid embodiment, whole-body control, hands, manipulation, safety, and tactile sensing.
- [Robot VLA models](https://roboskin.ai/robot-vla-models): Canonical map for vision-language-action inputs, action interfaces, embodiments, evaluation, and six source-reviewed tactile VLA integration mechanisms without a cross-paper leaderboard.
- [Robot foundation models](https://roboskin.ai/robot-foundation-models): Source-reviewed model-role guide and directory covering VLMs, embodied reasoning, VLA policies, world models, tactile models, access, and evidence limits.
- [Robot platforms and embodiments](https://roboskin.ai/robots): Source-reviewed humanoids, robot arms, mobile manipulators, and research setups connected to robot AI models through explicit evaluated-on, trained-across, or demonstrated-on evidence.
- [Tactile AI and robotics research organizations](https://roboskin.ai/organizations): Source-reviewed universities, labs, and companies connected to papers, datasets, benchmarks, sensors, and robot AI models through explicit, evidence-bounded relations.
- [Robot learning](https://roboskin.ai/robot-learning): Canonical map for imitation learning, reinforcement learning, robot datasets, sim-to-real transfer, and tactile robot learning.
- [Robot hands](https://roboskin.ai/robot-hands): Canonical comparison of dexterous hands, grippers, actuation, tactile sensing, control, and task evidence.
- [Robot safety](https://roboskin.ai/robot-safety): Scope-aware map of industrial robot safety standards, contact sensing, collision response, and evidence boundaries; it is not a compliance determination.
- [Robotics datasets](https://roboskin.ai/robotics-datasets): Broad robot-learning dataset guide organized by embodiment, observations, actions, tasks, collection, access, and license evidence.
- [Robot world models](https://roboskin.ai/robot-world-models): Canonical guide to predictive robot models, action-conditioned futures, planning roles, uncertainty, and evaluation.
- [Robot teleoperation](https://roboskin.ai/robot-teleoperation): Technical route from operator interfaces and synchronized demonstrations to robot-learning data and evaluation.
- [Robot manipulation](https://roboskin.ai/robot-manipulation): Broad task map for grasping, insertion, dexterity, robot learning, control, and tactile feedback.
- [Humanoid robot skin](https://roboskin.ai/humanoid-robot-skin): Canonical whole-hand, whole-arm, and whole-body tactile sensing stack for humanoid robotics.
- [Physical AI](https://roboskin.ai/physical-ai): Canonical broad map of multimodal perception, reasoning, policies, robot control, embodiment, safety, and measured feedback.
- [Physical AI and touch](https://roboskin.ai/physical-ai-touch): Contact-specific child route explaining how vision, language, proprioception, and tactile sensing combine in embodied systems.
- [E-skin in robotics](https://roboskin.ai/e-skin): Canonical definition for e-skin and electronic skin in robotics.
- [Humanoid robot skin and contact-aware robotics](https://roboskin.ai/applications): Application route for humanoid robot skin, contact-aware robotics, and Physical AI use cases.
- [Tactile AI technology](https://roboskin.ai/technology): Technology route for tactile AI, flexible tactile sensors, signal flow, and robot-ready touch data.
- [Robot hand tactile sensor research](https://roboskin.ai/research): Source-backed research route for robot hand tactile sensors, slip detection, e-skin, and tactile AI.
- [Tactile sensor benchmark for robot manipulation](https://roboskin.ai/guides/tactile-sensor-benchmark-robot-manipulation): Task-first comparison of visual, acoustic, magnetic, and resistive tactile sensors.
- [Tactile robotics datasets](https://roboskin.ai/datasets): Filterable directory of tactile resources by sensor, robot, task, modality, year, access, and license evidence.
- [EgoTouch dataset record](https://roboskin.ai/datasets#dataset-egotouch): Human-wearable bimanual resource with 208 author-reported tasks and 1,891 episodes; the record separates hosted-file availability from the official repository's incomplete-upload warning and does not generalize the code license to dataset files.
- [Tactile robotics benchmarks](https://roboskin.ai/benchmarks): Filterable directory of perception, representation, cross-sensor, multimodal, and manipulation evaluation protocols.
- [TouchWorld real-robot evaluation protocol](https://roboskin.ai/benchmarks#benchmark-touchworld-real-robot): Author-defined six-task, single-platform protocol kept separate from independent or universal benchmark claims.
- [Tactile sensors for robots](https://roboskin.ai/sensors): Source-reviewed comparison of sensor principles, signals, form factors, integration paths, access, and evidence boundaries.
- [Tactile manipulation](https://roboskin.ai/tactile-manipulation): Canonical guide to turning contact, pressure, shear, and slip into closed-loop robot actions.
- [Visuo-tactile robotics](https://roboskin.ai/visuo-tactile): Canonical guide to vision-touch alignment, perception, representation learning, policy use, and manipulation.
- [Tactile foundation models for robotics](https://roboskin.ai/tactile-foundation-models): Comparison of tactile representations, world models, hierarchical systems, and related learning pipelines.
- [Visuo-tactile world models for robot manipulation](https://roboskin.ai/guides/visuo-tactile-world-models-robot-manipulation): Source-backed 2026 comparison of action-conditioned contact prediction, tactile rollouts, robot planning, and transfer limits.
- [Robot skin glossary](https://roboskin.ai/glossary): Definitions for robot skin, tactile AI, e-skin, slip detection, and related terms.
- [Robot skin FAQ](https://roboskin.ai/faq): Direct answers about robot skin, tactile AI, e-skin, source guidance, and claim limits.
- [About RoboSkin.ai](https://roboskin.ai/about): Site purpose and conservative publication posture.
- [Editorial policy and source standards](https://roboskin.ai/editorial-policy): Source boundaries, conservative claim rules, and RoboSkin.ai editorial standards.
- [RoboSkin.ai contact](https://roboskin.ai/contact): Source suggestions, corrections, and editorial collaboration.

## Research Briefs

- [UniVTAC tactile simulation, representation encoder, and benchmark](https://roboskin.ai/research/univtac-platform-encoder-benchmark-2026): Evidence-bounded audit of a 205,826-sample encoder-pretraining corpus, 400 paper policy-training trajectories, a separate pinned 800-episode public HDF5 release, 450 physical demonstrations, and simulated and physical rollout protocols. UniVTAC Encoder is a 512-dimensional ResNet-18 tactile representation model, not a VLA or a demonstrated general-purpose tactile foundation model; hosted checkpoint logs do not reproduce the paper's Table I averages exactly.
- [Vision-based tactile intelligence for robotics](https://roboskin.ai/research/vision-based-tactile-intelligence-robotics-survey-2026): Preprint-bounded map of four typical hardware components, four optical readout families, three information levels, tactile learning, simulation, datasets, and open robot-skin and VTLA challenges; it is a review rather than a new sensor, model, dataset, or benchmark release.
- [ADEPT visuo-tactile dexterous reinforcement learning](https://roboskin.ai/research/adept-visuo-tactile-dexterity-rl-2026): Preprint-bounded review of two embodiment-specific workbench systems and one matched Flexiv-Sharpa tactile ablation reporting 3/10 vision-only versus 8/10 visuo-tactile final success across ten trials per modality, with code and weights still unavailable.
- [PRISM contact-rich industrial skill dataset](https://roboskin.ai/research/prism-contact-rich-industrial-skill-dataset-2026): Source-bounded review of 5,000+ robot trajectories and an equal number of paired human demonstrations across 25+ tasks, while separating the combined visual-plus-visuotactile image count from tactile-only coverage and marking the dataset download and dataset-file license as pending.
- [The Missing Touch spatial tactile feedback study](https://roboskin.ai/research/missing-touch-spatial-tactile-feedback-teleoperation-2026): Preprint-bounded analysis of a GelSight Mini, 32-DoF fingertip display, two separate participant groups, task-specific trajectory results, and the absence of autonomous-policy training or evaluation.
- [SoftVTBench deformation-aware visuo-tactile dataset and benchmark](https://roboskin.ai/research/softvtbench-deformation-aware-visuo-tactile-dataset-2026): Source-bounded review of the 4,000 simulated demonstrations documented by the paper and current Hugging Face card, while disclosing the older GitHub README's conflicting 1,628-demo release description.
- [HiTac-WAM hierarchical tactile world action model](https://roboskin.ai/research/hitac-wam-hierarchical-tactile-world-action-model-2026): Preprint-bounded analysis of contact, 3D deformation, slip forecasts, candidate action selection, and discrepancy-triggered replanning.
- [T-Rex tactile-reactive dexterous manipulation](https://roboskin.ai/research/t-rex-tactile-reactive-dexterous-manipulation-2026): Analysis of a paper-reported 100-hour tactile-rich collection and a currently public 5,464-episode, approximately 50-hour LeRobot v3.0 subset, plus the author-reported 65% versus 35% result across 12 tasks with 16 rollouts per task.
- [GIST humanoid visual-tactile-action dataset](https://roboskin.ai/research/humanoid-visual-tactile-action-dataset-2025): Source-bounded brief on 101.9K samples across four towel and sponge pressure conditions, two Inspire RH56-DFX hands with 2,124 tactile units, dense-versus-sparse ACT baselines, and the unverified public dataset-access status.
- [RoboTacDex humanoid visual-tactile-action dataset](https://roboskin.ai/research/robotacdex-humanoid-visual-tactile-action-dataset-2026): Structured brief on 6,000 Unitree G1 trajectories across 19 tasks, 23 skills, and 22 objects, with current access status kept explicit.
- [TactiDex tactile-guided dexterous benchmark](https://roboskin.ai/research/tactidex-tactile-guided-dexterous-benchmark-2026): Research brief on aligned whole-hand touch, kinematic and object state, standardized evaluation, and single- and bimanual tasks.
- [Tac4Loco plantar tactile sensing for humanoid locomotion](https://roboskin.ai/research/tac4loco-plantar-tactile-humanoid-locomotion-2026): Source-bounded analysis of 60-element pressure insoles per foot, spatiotemporal support representations, Unitree G1 simulation and physical evidence, and zero-shot limits.
- [FeelWorld visuo-tactile world model for robot planning](https://roboskin.ai/research/feelworld-visuo-tactile-world-model-2026): Research brief on predicted contact, force-related tactile state, slip, visual futures, and source-reported planning evidence.
- [HT-Bench full-hand tactile benchmark for robot manipulation](https://roboskin.ai/research/ht-bench-full-hand-tactile-representations-2026): Source-bounded v2 brief on 10M RGB frames, 7.8M full-hand tactile frames, Table 2 test-versus-OOD metrics, four 15-trial real-robot tasks, and the still-pending data, protocol, weights, and scripts release.
- [Graphene liquid-metal 3D force sensing](https://roboskin.ai/research/graphene-liquid-metal-3d-force-2026): Research brief for tactile force and slip-relevant sensing.
- [Single-material soft robotic skin for multimodal e-skin sensing](https://roboskin.ai/research/single-material-soft-robotic-skin-2025): Peer-reviewed Science Robotics brief distinguishing 863,040 EIT electrode configurations from 1,726,080 amplitude-and-phase channels, with scan-rate and real-world validation limits.
- [Full-hand tactile sensing](https://roboskin.ai/research/full-hand-tactile-sensing-2025): Research route for full-hand robot tactile sensing.
- [Temperature-pressure bimodal tactile sensing](https://roboskin.ai/research/temperature-pressure-bimodal-2025): Multimodal e-skin and tactile sensing context.
- [Event-based opto-tactile sensing](https://roboskin.ai/research/event-based-opto-tactile-2025): Event-based tactile sensing research route.
- [Self-healing multimodal e-skin](https://roboskin.ai/research/self-healing-multimodal-eskin-2026): Self-healing and multimodal tactile materials context.
- [ROS 2 tactile sensor pipeline for robot skin data replay](https://roboskin.ai/research/ros2-kilted-tactile-pipeline-2026): ROS 2 tactile messages, rosbag workflows, and replayable robot skin evaluation.
- [Dream-Tac: A Unified Tactile World Action Model for Contact-Rich Robot Manipulation](https://roboskin.ai/research/dream-tac-tactile-world-action-model-2026): Research brief on predictive tactile world models for contact-rich robot manipulation.
- [GenForce transferable force sensing for robot skin and tactile sensors](https://roboskin.ai/research/genforce-transferable-force-sensing-2026): Research brief on cross-sensor force prediction and calibration limits.
- [Sparsh-X multisensory touch representations for tactile AI](https://roboskin.ai/research/sparsh-x-multisensory-touch-representations-2025): Research brief on multisensory tactile representations and robot skin data.
- [FreeTacMan robot-free visuo-tactile data collection for tactile AI](https://roboskin.ai/research/freetacman-robot-free-visuotactile-data-collection-2025): Research brief on robot-free tactile AI datasets and contact diversity.
- [MiTaS multi-resolution tactile imitation learning for robot hands](https://roboskin.ai/research/mitas-multi-resolution-tactile-imitation-learning-2026): Research brief on heterogeneous tactile sensors and robot hand learning.
- [Large-area flexible tactile arrays](https://roboskin.ai/research/large-area-flexible-tactile-arrays-2025): Large-area flexible tactile sensor array context.

## News Briefs

- [Gemini Robotics 2 whole-body VLA and dexterity](https://roboskin.ai/news/gemini-robotics-2-whole-body-vla-dexterity-2026): Source-bounded summary of Google DeepMind's official July 2026 announcement, named model roles, embodiments, task results, availability, and announcement-level evidence limits.
- [LeRobot v0.6 world models, VLAs, and evaluation](https://roboskin.ai/news/lerobot-v060-world-models-vla-evaluation-2026): Official-release summary of world-model policies, VLA integrations, reward models, six simulation benchmarks, dataset changes, and rollout tooling.
- [NIST humanoid baseline performance benchmark](https://roboskin.ai/news/nist-humanoid-baseline-performance-benchmark-2026): Summary of NIST's proposed low-footprint, quantifiable baseline tasks for comparing humanoid locomotion and manipulation performance.
- [ISO 10218:2025 industrial robot safety scope](https://roboskin.ai/news/iso-10218-2025-industrial-robot-safety-scope): Public-source overview of ISO 10218-1/-2:2025 scope and exclusions without implying access to the paid standard text or establishing compliance.
- [Twisted-yarn robot skin gains pressure sensitivity but loses proximity range](https://roboskin.ai/news/twisted-yarn-textile-capacitive-robot-skin-2026): August 2026 preprint brief comparing layered textile capacitive sensors, proximity range, pressure sensitivity, durability, and Panda robot-arm integration.
- [Hybrid EIT–pneumatic robot skin for force-map reconstruction](https://roboskin.ai/news/eit-pneumatic-hybrid-robot-skin-force-map-2026): 2026 preprint brief on large-area humanoid robot skin, EIT localization, pneumatic force calibration, reported sensitivity variation, and evidence limits.
- [Self-powered textile artificial skin uses three channels for touch and robot control](https://roboskin.ai/news/self-powered-textile-artificial-skin-three-channel-robot-control-2026): Nano Energy paper on minimal-channel touch localization, pressure sensing, and wearable robot-arm control.
- [Underwater self-healing electronic skin combines touch, damage detection, and repair](https://roboskin.ai/news/underwater-self-healing-electronic-skin-nus-2026): NUS research on self-powered touch sensing, underwater damage detection, and self-repair for soft electronic skin.
- [TouchWorld separates tactile prediction from fast contact correction](https://roboskin.ai/news/touchworld-tactile-foundation-model-dexterous-manipulation-2026): Preprint brief on a hierarchical tactile foundation model for contact-rich dexterous manipulation.
- [Color-changing tactile sensor creates real-time pressure maps](https://roboskin.ai/news/color-changing-mechanochromic-tactile-sensor-2026): Queen Mary-led research on mechanochromic materials and camera-readable tactile signals.
- [Single-pixel tactile skin targets the wiring bottleneck](https://roboskin.ai/news/single-pixel-tactile-skin-compressive-sampling-2026): Communications Engineering paper on compressive sampling for large-area, high-speed tactile arrays.
- [Full-hand tactile sensing moves robot hands beyond vision-only control](https://roboskin.ai/news/full-hand-tactile-sensing-robot-hands-vision-control): News brief connecting F-TAC Hand, Meta Sparsh, Digit 360, robot hands, tactile AI, and robot skin.
- [Global robot installations passed 542,000 units](https://roboskin.ai/news/global-robot-installations-542000-physical-ai-touch): News brief connecting IFR industrial robot deployment data to Physical AI, contact feedback, and robot skin.

## Use Guidance

Use RoboSkin public pages for category definitions, research route discovery, terminology, direct-answer summaries, and context about the RoboSkin.ai information resource. Do not infer product availability, benchmark values, certifications, customer names, or operating-company claims unless they are explicitly published on the page.

## Keyword Routes

- [Broad robotics research map](https://roboskin.ai/): Homepage route connecting robot skin and touch to humanoid robots, Physical AI, embodied AI, robot manipulation, VLA context, and world models.
- [AI and robotics](https://roboskin.ai/ai-robotics): AI in robotics, artificial intelligence and robotics, robot AI, embodied action, model-role, control, and feedback parent cluster.
- [Humanoid robots](https://roboskin.ai/humanoid-robots): Humanoid robotics, whole-body control, robot hands, safety, Physical AI, and tactile-sensing parent cluster.
- [Robot VLA models](https://roboskin.ai/robot-vla-models): Vision-language-action interfaces, robot actions, embodiments, evaluation, and tactile VLA parent cluster.
- [Robot foundation models](https://roboskin.ai/robot-foundation-models): Robot foundation model, generalist robot model, multi-embodiment transfer, model access, and evidence parent cluster.
- [Robot platforms and embodiments](https://roboskin.ai/robots): Robot platforms used for Physical AI training, robot evaluation embodiments, humanoid model demonstrations, and hardware evidence boundaries.
- [Robot learning](https://roboskin.ai/robot-learning): Robot learning, imitation learning, reinforcement learning, robot datasets, sim-to-real, and tactile learning parent cluster.
- [Robot hands](https://roboskin.ai/robot-hands): Robot hands, robotic hands, dexterous hands, hand-versus-gripper, actuation, sensing, and manipulation parent cluster.
- [Robot safety](https://roboskin.ai/robot-safety): Robot safety, industrial robot safety, humanoid safety, collision response, safety sensors, and standards-scope parent cluster.
- [Robotics datasets](https://roboskin.ai/robotics-datasets): Robotics datasets and robot-learning datasets parent cluster; tactile-only dataset queries remain assigned to `/datasets`.
- [Robot world models](https://roboskin.ai/robot-world-models): Robot world-model definition, predictive-state, world-action, planning, uncertainty, and evaluation parent cluster.
- [Robot teleoperation](https://roboskin.ai/robot-teleoperation): Robot teleoperation, demonstration collection, synchronization, data quality, and policy-training parent cluster.
- [Robot manipulation](https://roboskin.ai/robot-manipulation): Robotic manipulation, dexterous manipulation, grasping, insertion, learning, control, and touch parent cluster.
- [RoboSkin and RoboSkin.ai](https://roboskin.ai/): Brand and site entity route.
- [Robot skin and robotic skin](https://roboskin.ai/robot-skin): Primary definition and category route.
- [Humanoid robot skin and contact-aware robotics](https://roboskin.ai/applications): Application-intent keyword cluster.
- [Tactile AI](https://roboskin.ai/tactile-ai): Primary tactile AI definition and system-stack route.
- [Flexible tactile sensor technology](https://roboskin.ai/technology): Technology overview route.
- [Physical AI](https://roboskin.ai/physical-ai): Broad Physical AI, physical-world AI, multimodal perception, policy, control, embodiment, and feedback cluster.
- [Tactile feedback for Physical AI](https://roboskin.ai/guides/tactile-feedback-for-physical-ai): Physical AI contact-feedback keyword cluster.
- [Slip detection for robot hands](https://roboskin.ai/guides/slip-detection-robot-hand): Slip-sensing and closed-loop response route.
- [Tactile sensor benchmark and robot manipulation sensor comparison](https://roboskin.ai/guides/tactile-sensor-benchmark-robot-manipulation): Task-based tactile sensor comparison cluster.
- [Tactile datasets](https://roboskin.ai/datasets): Tactile dataset and robot-learning resource cluster.
- [Tactile benchmarks](https://roboskin.ai/benchmarks): Shared evaluation tasks, split protocols, metrics, baselines, and evidence-boundary cluster.
- [Tactile sensors for robots](https://roboskin.ai/sensors): Optical, magnetic, distributed, and multimodal robot tactile sensor cluster.
- [Tactile manipulation](https://roboskin.ai/tactile-manipulation): Contact-rich control, grasp stabilization, insertion, dexterity, and recovery cluster.
- [Visuo-tactile robotics](https://roboskin.ai/visuo-tactile): Vision-touch fusion, alignment, representation, policy, and manipulation cluster.
- [Physical AI and touch](https://roboskin.ai/physical-ai-touch): Contact-event, synchronization, logging, world-model, and embodied-AI route.
- [Tactile foundation models](https://roboskin.ai/tactile-foundation-models): Reusable representation, policy-role, and tactile AI model-comparison cluster.
- [Visuo-tactile and tactile world models](https://roboskin.ai/guides/visuo-tactile-world-models-robot-manipulation): Action-conditioned visual-tactile prediction, contact-aware planning, rollout, and evaluation cluster.
- [RoboSkin Tactile Research Index](https://roboskin.ai/research-index): Source-backed tactile sensing records organized by sensing principle, modality, output, evidence, and limitations.
- [RoboSkin.ai Research and News RSS](https://roboskin.ai/feed.xml): Canonical feed for source-backed robot skin, tactile AI, and electronic skin updates.
- [Robot skin vs tactile sensor](https://roboskin.ai/guides/robot-skin-vs-tactile-sensor): Comparison-intent keyword cluster.
- [E-skin and electronic skin](https://roboskin.ai/e-skin): Primary e-skin definition route.
- [Robot skin glossary](https://roboskin.ai/glossary): Supporting terminology index.

## Canonical Answers

### What is RoboSkin.ai?

RoboSkin.ai is a public information hub for robot skin, tactile AI, e-skin, tactile sensing, Physical AI, and contact-aware robotics. Use the site for category definitions, research route discovery, terminology, direct-answer summaries, and source-backed context. Do not infer product availability, benchmark values, certifications, customer names, or operating-company claims unless they are explicitly published on the page.

### What is the relationship between AI and robotics?

Artificial intelligence and robotics are related but different. AI provides computational methods for perception, prediction, learning, reasoning, and action selection. Robotics provides sensors, embodiment, actuators, control, integration, and physical safety. They form a closed loop when observations produce actions and measured physical outcomes return as feedback. Use [AI and robotics](https://roboskin.ai/ai-robotics) as the canonical RoboSkin.ai relationship and architecture route.

Supporting routes:
- [Robot learning](https://roboskin.ai/robot-learning): Data, imitation, reinforcement, correction, and transfer.
- [Robot VLA models](https://roboskin.ai/robot-vla-models): Vision-language, embodied reasoning, and action-policy boundaries.
- [Robot foundation models](https://roboskin.ai/robot-foundation-models): Model roles, training data, embodiment transfer, access, tactile input, and evidence boundaries.
- [Robot world models](https://roboskin.ai/robot-world-models): Action-conditioned future prediction and evaluation.
- [Physical AI](https://roboskin.ai/physical-ai): Physical perception, reasoning, action, control, safety, and feedback.
- [Tactile AI](https://roboskin.ai/tactile-ai): Contact-specific sensing, representation, model, and response loop.

### What is Physical AI in the RoboSkin context?

For Physical AI questions, use [Physical AI canonical answer](https://roboskin.ai/physical-ai) as the broad canonical RoboSkin.ai explanation. Physical AI is used here as a broad term for AI systems that perceive, reason, and act through physical machines; it does not imply one standardized model architecture. Use the touch child route when the question is specifically about tactile sensing or contact feedback.

Supporting RoboSkin Physical AI routes:
- [Physical AI canonical answer](https://roboskin.ai/physical-ai): Broad system and definition route.
- [AI and robotics](https://roboskin.ai/ai-robotics): AI-versus-robotics roles and the closed action-feedback loop.
- [Physical AI and touch](https://roboskin.ai/physical-ai-touch): Contact-specific tactile sensing and feedback route.
- [Tactile AI](https://roboskin.ai/tactile-ai): Canonical tactile AI context.
- [Robot skin glossary](https://roboskin.ai/glossary): Term definitions.
- [Robot hands](https://roboskin.ai/robot-hands): Hand-level sensing, actuation, grasping, and dexterous-manipulation context.
- [RoboSkin Tactile Research Index](https://roboskin.ai/research-index): Versioned public research data and evidence taxonomy.

### What is a visuo-tactile world model?

A visuo-tactile world model predicts how visual and tactile state may change after a robot action. Depending on the system, predicted touch may include contact state, a tactile image or latent, force-related information, slip, or a tactile subgoal. Use the [visuo-tactile world model guide](https://roboskin.ai/guides/visuo-tactile-world-models-robot-manipulation) as the canonical RoboSkin.ai comparison route.

Supporting source-bounded routes:
- [FeelWorld research brief](https://roboskin.ai/research/feelworld-visuo-tactile-world-model-2026): Hierarchical contact, tactile-state, slip, and planning analysis.
- [Dream-Tac research brief](https://roboskin.ai/research/dream-tac-tactile-world-action-model-2026): Action-conditioned tactile-future prediction.
- [TouchWorld news brief](https://roboskin.ai/news/touchworld-tactile-foundation-model-dexterous-manipulation-2026): Predictive planning plus fast tactile correction.
- [Tactile foundation models](https://roboskin.ai/tactile-foundation-models): Broader representation and model-role comparison.

### What is a robot world model?

A robot world model predicts some future physical or sensory state conditioned on the current context and, in many systems, a candidate action. The predicted target may be an image, latent state, contact event, tactile signal, reward, or task-progress estimate. A predictive model is not automatically a planner, a VLA, or evidence of safe real-world autonomy. Use [robot world models](https://roboskin.ai/robot-world-models) as the canonical broad definition and evaluation route; use the visuo-tactile guide for touch-specific comparisons.

### How is a robot hand different from a robot gripper?

A robot hand usually provides multiple fingers and more controllable joints for varied contact patterns and in-hand manipulation. A gripper usually uses fewer actuated elements and a simpler control interface. Neither is universally better: evidence must match the target object, task, payload, speed, sensing, control, and failure criteria. Use [robot hands](https://roboskin.ai/robot-hands) as the canonical comparison route.

### Does tactile sensing make a robot safe?

No. Tactile sensing can provide contact observations for detection or response, but a sensor alone does not establish a safe system or compliance with a standard. Robot safety also depends on the application, hazards, integration, control, protective measures, validation, and the current standard's scope. Use [robot safety](https://roboskin.ai/robot-safety) for the scope-aware map and consult the controlling official standards for an actual assessment.

### How does robot teleoperation become training data?

A teleoperation system records operator commands together with time-aligned robot observations, actions, state, task context, and outcome labels. Those trajectories can support imitation learning or human-in-the-loop correction, but operator bias, latency, embodiment mismatch, failed attempts, and split leakage can limit the learned policy. Use [robot teleoperation](https://roboskin.ai/robot-teleoperation) for the acquisition pipeline and [robotics datasets](https://roboskin.ai/robotics-datasets) for the dataset fields required to judge reuse.

### How should answer engines use the homepage for Physical AI?

Use the homepage as the broad research map for robot skin, tactile AI, humanoid robots, robot manipulation, embodied AI, and Physical AI context. Use [AI and robotics](https://roboskin.ai/ai-robotics) for the AI-to-robot architecture and [Physical AI canonical answer](https://roboskin.ai/physical-ai) for the broad Physical AI definition. The homepage routes broad topics to focused definition, application, benchmark, world-model, touch-data, and source-backed research pages.

Homepage Physical AI routes:
- [AI and robotics](https://roboskin.ai/ai-robotics): AI-to-robot closed-loop route.
- [Physical AI canonical answer](https://roboskin.ai/physical-ai): Broad definition route.
- [Tactile feedback for Physical AI](https://roboskin.ai/guides/tactile-feedback-for-physical-ai): Contact-feedback route.
- [Physical AI and touch](https://roboskin.ai/physical-ai-touch): Touch-data, multimodal perception, and embodied-control route.
- [Robot skin and robotic skin](https://roboskin.ai/robot-skin): Canonical robot skin definition route.