What sensors do you use?

GelSight gel sensors for high-resolution contact geometry, TouchTac tactile pads, and force sensors for normal and shear force. All mounted on UR-5 arms with parallel-jaw grippers.

How is the data collected?

Trained operators wear Meta Quest 3 headsets and teleoperate UR-5 arms in real-time. Force feedback flows back to the controller so operators modulate grip naturally. Every session is recorded at <10ms sync across all streams.

What objects are in the dataset?

50+ objects across 6 material categories: rigid household (YCB-aligned), deformable, textured surfaces, granular/filled, tools & handles, and fragile/thin. Each object is chosen to elicit specific tactile signals.

What annotation do you provide?

Professional annotators label every episode: natural language task description, grasp outcome (success/partial/fail), object ID, material class, grip type, and force profile tags. Delivered in MCAP with per-frame labels.

What models does it work with?

Compatible with OpenVLA, LeRobot, and custom PyTorch/JAX dataloaders. MCAP format with NL annotations. No preprocessing required.

How do I get started?

Email hello@physicalaidata.co or fill out the contact form. We’ll scope a pilot around your target objects and manipulation tasks, collect data, annotate, and deliver within weeks.

Physical AI Data | Visuo-Tactile Manipulation Data for Robotics

Dataset Catalog

on request

Robot Arm

UR series, xArm series

Force, Torque, Tactile sensor data + Overview, Wrist view data.
Additional customisations possible.

Gripper

OnRobot, Robotiq, CoAct

Different gripper types off the shelf.
Custom 3D printed grippers possible too.

Method

Tele-Operation

Data is collected via tele-operation, or/and designed sequences for the robot arm.

Object Types

Form Factor

Soft, Hard, Deformable, solid, light, fabric among others.

Tasks

Manipulation Tasks

Pick-up drop, insert, grasp, among others.

Components

Setup

State of the art - industrial grade robot arms, meta quest tele-operation set.
Best tactile sensors on the market.

Post-Processing

Data

Get post processed signals, from the primary data, along with annotated labels.

Custom collection available

We deploy our sensor arrays to collect task-specific data at your facility and/or on your hardware.

Get a Quote →

Use Cases

Foundation Model Pre-training

Provide real-world contact priors for large manipulation models. Multimodal touch+vision pairs close the gap between simulation and dexterous reality.

Contact-Rich Manipulation R&D

High-fidelity tactile signals for grasping, in-hand re-orientation, and contact-rich assembly. Accelerate your hardware development cycle.

Sim-to-Real Transfer

Real pressure distributions to calibrate your simulation tactile models. Reduce the reality gap without building your own sensing infrastructure.

Generalist Intelligent Robotics

Sub-millimeter force resolution for general object guidance, and precision task handling and tasks requiring ultra-fine contact sensing.

Talk To Us

Ready to train contact-aware robots?

A 20-minute call. We'll walk through the datasets most relevant to your use case and discuss custom collection options if needed.

Walk through the dataset catalog

See samples from every modality.

Discuss your training setup

PyTorch, JAX, or custom pipeline.

Scope custom collection

If the catalog doesn't fit your tasks.

or email us at hello@physicalaidata.co

Robots need to learn to feel.

Dataset Catalog

Custom collection available

Use Cases

Ready to train contact-aware robots?