Autonomy application

QVET platform application / Autonomy

A living view of the road.

Vehicle-mounted sensing turns synchronized observations into versioned WorldGraph products for autonomy, mapping, and robotics teams.

See how it worksOpen desktop demo

Solana coordinates customer-funded mapping missions and contributor payments. Environment data stays off-chain.

Daylight street scene with vehicles, pedestrians, tracked objects, and sparse spatial returns.
15,000+real ride sessions in the mobility AI training foundation
Backed byPlug and Play
Powered byRashi

Multimodal capture

Context for people. State for machines.

Camera imagery can support operator review while synchronized spatial returns carry geometry and movement into the machine pipeline.

Daylight street scene paired with sparse spatial returns.
Same scene / spatial returns
The same daylight street scene with tracked road users and vehicles.
Same scene / detected movement
Night road research visualization with a camera view beside sparse spatial returns.
NightLow-light camera context with sparse spatial returns.
Evening road research visualization with camera and spatial sensing views.
EveningThe same machine-readable view across changing light.
Analysis interface showing camera imagery, spatial returns, and detected road users.
AnalysisCaptured observations become aligned tracks and geometry.

Environment intelligence

One sensing network. Four useful outputs.

WorldGraph is a versioned spatial model of persistent features, movement patterns, and freshness. It is not a photorealistic city mesh.

Illustrative machine-readable view of street geometry and persistent physical features.
Illustrative WorldGraph view

  • Spatial structure

    Road edges, curbs, barriers, and persistent landmarks in a consistent machine coordinate frame.

  • Change and freshness

    Versioned tiles show where observations are recent and where new coverage is needed.

  • Movement patterns

    Aggregated stopping, braking, acceleration, and turn events without driver identity records.

  • Requested refresh

    Customers can direct new collection toward stale or operationally important spatial cells.

How it works

Capture once. Build several products.

  1. 01

    Fleet sensing

    A mobile pod timestamps synchronized sensor and pose data, then securely uploads each capture segment.

  2. 02

    Cloud processing

    QVET aligns each observation with vehicle pose, accounts for ego motion, and separates moving objects from persistent returns.

  3. 03

    WorldGraph

    Persistent features, movement events, source history, and freshness scores become versioned data products.

  4. 04

    Enterprise delivery

    Customers receive scoped WorldGraph tiles, freshness signals, requested-area exports, and machine APIs.

Illustrative compact mobile sensing module.
Mobile sensing hardware direction

System boundary

Lightweight on the vehicle. Intensive in the cloud.

The mobile pod handles sensing, timestamps, buffering, and secure upload. Spatial reconstruction, behavior extraction, freshness, and WorldGraph generation run in QVET Cloud.

Illustrative autonomous delivery robot moving along an urban street.

Who uses it

For teams operating autonomous systems in real places.

Robotaxi and AV teams
Prioritize new-city coverage, compare fresh observations, and add RF-derived context to existing mapping stacks.
Delivery robotics
Monitor road-edge changes, stopping behavior, and route conditions across operating areas.
Mapping providers
Request targeted refreshes and consume versioned features with clear source history.

Solana contribution network

Pay for useful coverage, not undirected collection.

Customers fund requested cells in USDC. QVET verifies accepted observations, then the marketplace program releases payment once from escrow.

Sensor captures and WorldGraph products remain off-chain. Solana handles escrow, settlement, and replay protection.

  1. 01

    Customer defines target cells

  2. 02

    USDC budget enters campaign escrow

  3. 03

    Contributor covers requested cells

  4. 04

    Accepted reward settles once