Core Modalities
Four foundational sensorimotor streams calibrated for physical AI and embodied robotic learning.
RGB-D
Egocentric human data in commercial and residential environments: calibrated stereo RGB with dense metric depth, synced IMU, 6-DoF VIO, and 3D hand pose.
Teleoperation
Ground-truth sensorimotor demonstrations across bimanual and humanoid embodiments. Sub-millimeter kinematic tracking, 200Hz joint torques, 6-DoF end-effector trajectories, and synchronized multi-view wrist cameras—curated for Action Chunking (ACT) and Diffusion Policies.
Tactile & Force
High-density tactile sensing and magnetic kinematic tracking for contact-rich dexterous manipulation. Dense pressure arrays, continuous shear and normal force feedback, and synchronized 25-DoF hand poses—capturing the fine physical mechanics of human touch.
Simulation & Digital Twins
High-fidelity neural digital twins and physics-accurate simulation environments for end-to-end policy pretraining. Photorealistic real2sim reconstruction, multi-sensor synthetic rendering, domain randomization, and closed-loop sim2real validation across residential, industrial, and unstructured domains.
Custom Collection
Tailored sensorimotor data pipelines engineered for your proprietary hardware, sensor rigs, and domain environments.
Scoping Call
A 30-minute technical session to specify your robot embodiment, sensor topologies, target manipulation tasks, and environment requirements.
Written Spec & Quote
We deliver a formal data schema, calibration protocol, episode milestones, and transparent per-hour pricing within 48 hours.
Pilot Batch
A small-scale benchmark dataset collected and delivered to verify your Action Chunking / Diffusion policy training pipeline before full rollout.
Weekly Delivery
Continuous, QA-verified sensorimotor batches streamed weekly directly to your private AWS S3 bucket or Hugging Face enterprise repository.
How Agon measures
and acts on the world.
Every module is a way of capturing or interacting with physical reality. Together they form the signal vocabulary behind the dataset.
Pixel Understanding
Depth & Geometry
Object Tracking
Segmentation
Tactile Force
Sound Waves
Speed & Motion
Point Cloud
Surface Contour
