Smartphone Ego
First-person capture from an approved phone configuration. Scales to volume.
- Approved phone configurations and mounting positions
- Task briefs with initial and terminal state holds
- Video, motion signals, and device metadata
AuraOne / Human Data / Physical AI
Real-world demonstrations, prepared for your AI task.
Keep observed capture, machine enrichment and synthetic coverage distinct, with the relevant timing and permission records.
People record everyday tasks, with consent, from a phone or a wearable. Reviewers check every episode before it ships.
Capture, Synthetic Coverage and Machine Enrichment are scoped per program.
TaskA brief with clear start and end states
ConsentEach person and site agrees before recording
CaptureA phone or wearable records the work
ReviewEvery episode is checked before delivery
What shipsReviewed episodes
Robotics training data
Teleoperation and human-demonstration data.
Teleoperation data records a person remotely driving a robot through a task. Human-demonstration data records people doing the task themselves. Robotics teams use both to teach robots by example, a method called imitation learning.
AuraOne's Physical AI program covers both. Trained operators drive your robot through the task, or people perform it while cameras and wearables capture how. Reviewers check every episode before it ships, and observed, machine-enriched and synthetic data stay distinct. Glossary: teleoperation data
Real-world capture
Teach machines how the world works.
People perform everyday tasks with consent. Trained operators capture the motion, timing, and failure context — not just the picture.
One task. Three moments.
Three-phase episode segmentation
Before · 1 of 3
Two ways to record the work.
Wearable Manipulation
Multi-view capture for tasks where hand visibility and timing matter.
- Head view, one or two wrist views, optional chest view
- Registered stream roles with full camera and kit lineage
- Measured temporal alignment across every accepted view
Phone capture runs in Aura Capture, our human demonstration capture app. Programs that need mounts and cameras can use our robotics capture kits.
- TeleoperationTrained operators drive your robot through the task.Session, operator, and task record
- Human demonstrationPeople perform the task while cameras and wearables capture how.Video, pose, and device metadata
- Multimodal captureVideo, trajectory, and force streams recorded in sync.Aligned streams with timestamps
- Episode reviewReviewers watch every episode. Failures get labeled, not deleted.Disposition, reason, and rework history
- Technical validationDropped frames, clock drift, and sensor faults are caught first.Validation report per episode
- Versioned deliveryAccepted episodes ship with manifests, checksums, and a data card.Manifest, checksums, and data card
Real rooms. Not a lab bench.
We record in the places your robot will actually work.
What you receive
Three layers. One usable dataset.
Raw is never overwritten. Clean removes the dead time. Intelligence explains what happened.
SourceNever overwritten
Raw
The unchanged evidence, exactly as it was collected.
- Original media and sensor artifacts
- Immutable checksums and artifact identity
- Device, camera, and capture metadata
DerivedMapped from Raw
Clean
The accepted task content, with the dead time removed.
- Accepted task intervals and valid streams
- Setup and interruptions excluded, each with a reason
- Reproducible raw-to-clean mappings
DerivedExplains Raw and Clean
Intelligence
The record that explains what happened, and whether it worked.
- Task version, phase boundaries, and terminal state
- One declared outcome per episode
- Quality, alignment, and provenance evidence
What arrives in a delivery
Your contract sets the exact scope. Managed robotics programs collect and review human demonstrations, then coordinate a monitored handoff; queued transcode, copy, or webhook work is not represented as delivered.
- Recordings
- Video, depth, and audio streams recorded in sync
- Motion
- Trajectories, joint states, and force readings
- Context
- Robot, environment, and operator metadata
- Review
- Accept and reject decisions, rework history, failure labels
- Package
- Data card, delivery manifest, and per-file checksums
- Formats
- Video as MP4 or MOV. Records as JSONL, JSON, or a contracted export format
Beyond capture
Use the service the data problem requires.
Tell us the gap, or let AuraOne help identify it. Synthetic Coverage does not require Machine Enrichment. Machine Enrichment does not require Synthetic Coverage. Each capability must be contracted and production-qualified for its own program scope before use.
Combine real-world teleoperation and egocentric capture with local episode inspection, automated trajectory checks, and simulation-backed coverage.
Generate what is missing.
Start from a buyer-defined or AuraOne-assisted gap to generate, validate, review, and deliver an immutable coverage batch.
- No enrichment prerequisite
- Accepted Coverage Batch is the commercial delivery
Understand what is already there.
Add machine suggestions to existing real episodes and route uncertainty through human review.
- No synthetic generation prerequisite
- Output classes set by contract
Connect them when the evidence supports it.
Enriched episodes can point to a measurable gap, and Synthetic Coverage can then fill it. The full loop is useful, but never a prerequisite for either capability.
Synthetic Coverage
Fill a known gap with verified examples.
When a real-world failure mode needs broader variation, define the scenario and target count to generate synthetic trajectories complete with a distribution report and lineage manifest.
Start with the failure you already know.
A coverage program can begin from a buyer brief. Automatic gap discovery is not required.
- Low-light manipulation variants
- Transparent or reflective containers
- Recovery episodes after grasp failures
- Target counts, formats, and acceptance thresholds
Use source evidence to refine the target.
When the program includes suitable source data and analysis, AuraOne can help identify measurable underrepresented conditions. The evidence and method stay attached to the gap.
- Slice and condition coverage
- Failure and recovery distribution
- Scenario constraints and exclusions
- Traceable source-to-gap rationale
- Define the gapBuyer-defined or AuraOne-assisted
- Specify the scenarioConditions, counts, formats, rights
- Generate coverageApproved providers and recorded lineage
- Validate and reviewQuality gates plus human decisions
- Accept the batchImmutable dataset, manifest, and receipt
Lineage and distribution manifest included.
Every synthetic expansion batch ships with its scenario specification, physics and sensor parameters, and a distribution report comparing synthetic rollouts against your real-world failure episodes.
Machine Enrichment
Turn raw Physical AI video into reviewable structure.
Each machine suggestion carries a confidence score and its source episode. Confidence thresholds route questionable outputs into an append-only human review record with preserved machine authorship. Automated kinematics and trajectory checks run across every episode file alongside multi-camera visual review.
A machine-readable account of what happened.
Output classes are configured for your program’s sensor suite and evaluation schema.
- Episode and interval descriptions
- Task phases and temporal event boundaries
- Actions, objects, failures, and recoveries
- Confidence, provider, model, and prompt lineage
Uncertain output becomes a review task.
Existing annotation, queue, persistence, and review machinery carries machine suggestions into an auditable human decision.
- Program-specific confidence thresholds
- Accept, correct, reject, or quarantine decisions
- Append-only corrections with immutable machine authorship
- Durable links back to the source episode
- Ingest real episodesAccepted, rights-cleared Capture episodes
- Record machine suggestionsDescriptions, phases, actions, objects
- Route uncertaintyConfidence thresholds decide review
- Human reviewAccept, correct, reject, or quarantine
- Deliver enrichmentAnnotations with source and review lineage
Machine Enrichment stands on its own.
Every automated kinematic finding is logged with its exact frame range and sensor channel so reviewers can jump straight to the anomaly. Enrichment can run independently or alongside Synthetic Coverage.
Verification & provenance
Explicit lineage across recorded, automated, and synthetic robotics data.
Every artifact in a physical-AI program is tagged with its origin so your team never mistakes a machine flag or a synthetic rollout for a recorded episode.
AuraOne verifies the batch.
Coverage, quality, review, rights, provenance, format, and delivery evidence meet the contracted criteria.
The buyer accepts the delivery.
An authorized buyer accepts the immutable batch. This completes the normal commercial delivery path.
The customer reports downstream impact.
If the customer evaluates the batch in a model or robot, AuraOne can attach the reported result without presenting it as AuraOne-controlled proof.
AuraOne runs a separately authorized evaluation.
Only a controlled, authorized environment can produce this evidence class. It is optional and outside normal batch completion.
Program governance and quality standards
- Recorded sensor data, automated kinematic flags, and synthetic rollouts are stored in separate, clearly tagged lineage tracks.
- Held-out evaluation episodes remain strictly isolated from training batches to ensure clean policy benchmarking.
- Hardware calibration, camera intrinsics, and frame-sync tolerances are verified at the start of every capture cohort.
- All episode manifests export in portable open schemas compatible with standard robot-learning frameworks.
What the data lets you do
Each outcome depends on the scope your program agreed to.
| Outcome | Work | What you receive | Program fit |
|---|---|---|---|
| Train a manipulation policy | Collect successful demonstrations of one skill across varied objects and settings. | Episode count, task coverage, and accept rate. | Coverage applies to the environments and objects in the brief. |
| Build a failure set | Keep the attempts that went wrong and label how they went wrong. | Failure labels, source episodes, and reviewer rationale. | Reuse depends on your rights scope and retention policy. |
| Benchmark a robot release | Replay a fixed episode set against a new policy and compare outcomes. | Held-out episodes, scoring criteria, and per-episode results. | Results apply to the scoped episode set and criteria version. |
| Expand into a new environment | Run the same skill in homes, warehouses, or retail floors. | Environment coverage, location consent, and privacy zones. | Each site requires its own consent and privacy review. |
Start with the gap
Bring the data problem, not a predetermined workflow.
Tell us what is missing, show us the real episodes that need structure, or scope both. We will define the applicable capability, evidence boundary, and acceptance criteria with you.
Guides
Related reading
Physical AI Has a Human Data ProblemPhysical AI teams need more than video: task intent, demonstration context, failure labels and review records that make robotics training data useful.Robotics Training Data: Collection, Review and HandoffPlan robotics demonstrations around task definitions, sensor context, quality review, source rights and dataset handoff, with program-specific scope.Robotics Studio Open: Local VLA Dataset ReviewReview declared robotics episode metadata, preserve unknown values and export local evidence with Robotics Studio Open’s explicit input and delivery boundaries.