Typed · Shared · Open

One bus for spatial data

SpatialDDS is an open protocol for real-time exchange of spatial world models. Robots, vehicles, AR clients, sensors, and digital twins publish and consume the same typed spatial data — poses, detections, anchors, zones, maps — on one shared bus.

Read the specification Try the web binding Connect a ROS 2 robot
A city waterfront where a delivery robot, an autonomous car, a drone, sensor poles, and people with phones are all connected by one shared stream of spatial data.

Typed spatial data

Poses, detections, anchors, zones, and maps as typed messages that carry uncertainty, attribution, and provenance — not just coordinates. Heavy content rides by reference; the bus stays light.

Discovery by location

Ask what covers here. Geohash-indexed discovery over DNS and well-known HTTP resolves a place to the services, anchors, and content that cover it — no registry, no gatekeeper.

Every system speaks it

DDS where machine-tempo QoS matters. Plain GET and WebSockets for browsers. Carried over ROS 2, MQTT, and MCAP recordings. The data model is one; the transports are many.

Proven in the open

Machine-checkable conformance suites run against public datasets, and public reference deployments you can visit — every claim in the spec is one someone can re-run.

Venue demo

The reference demo

A venue as a live model: robots, ducks, declared water and a browser, all on one bus. Shown at ISMAR 2026 (watch the clip).

Positioning service · spatialdds branch

OpenVPS

An open visual positioning service on the same bus. It announces its coverage and answers localization queries with computed poses; clients discover it, no configured endpoint.

Perception adapter

SceneScape sidecar

A stock Intel SceneScape deployment publishing its scene analytics as typed samples: tracks, detections, zones and a georeference.

14IDL profiles
216conformance checks
5public datasets validated
4protocol bridges
2live reference demos