“Where is unit 12 right now, what is its battery level, and what is it assigned to?”
Recent events plus latest-state projections, tuned for live maps, subscriptions, geospatial search, alerting and short-window replay.
Operator-owned data platform
The operator-owned data plane for physical operations
PhyCloud ingests, streams, correlates and retains vendor-neutral events from every robot, drone, vehicle and sensor you operate — giving you one trustworthy data layer that outlives any single vendor.
- 14:02:11.204 amr.a7 observation.pose zone B·4 · batt 68% · task ACTIVE Fresh
- 14:02:11.640 wms.orders assignment.created TASK-4471 assigned · 2 units Fresh
- 14:02:12.008 drone.114 observation.pose inspection leg 3 · alt 24 m Fresh
- 14:02:12.531 camera.n12 media.available stream ref · retention: operational Stale
- 14:02:13.117 amr.a7 telemetry.gap link loss 1.4 s · recorded Data gap
Your operations span many vendors. Your data doesn't.
Every system you deploy arrives with its own data model, retention policy and integration limits. The gap is not a lack of telemetry — it is the absence of a common, trustworthy event layer that survives a vendor change.
Combine systems that were never designed to work together
Robots, vehicles, fixed sensors and the software systems around them each ship their own schema, retention policy and search interface. PhyUDM gives them one contract.
Reconstruct an incident without opening six products
Retrieve the complete history around an incident and replay a synchronised timeline across every source, with data gaps shown rather than silently omitted.
Prove what happened, and who looked
Append-only events, hash-chain provenance and an immutable audit trail over every search, export, share and administrative action.
Change vendors without losing the record
Your operational history lives in your data model, in open formats, in storage you can own — not inside the platform of whichever hardware you bought last year.
Acquire, normalise, retain — as three separable concerns
Separating how data is acquired from how facts are represented and how events are used is what turns a dashboard into an integration architecture.
- 01
PhyTrace and connectors acquire
SDKs, managed vendor connectors, gateways and your own pipelines capture source data and normalise it — with buffering, retry and transport handled for you.
PhyTrace - 02
PhyUDM represents the facts
An open, Apache-2.0 event model covering drones, ground robots, vehicles, cameras, sensors and the software systems around them. Implement it without buying anything.
PhyUDM - 03
PhyCloud retains and serves
A hot store runs today's operation; an open lakehouse preserves your long-term operational memory. One API, one catalog, one governed platform.
PhyCloud
Events, media and actions are not the same problem
Treating them as one is how platforms end up storing video in event rows and letting a dashboard issue commands. PhyCloud keeps them on separate planes with different controls.
Event plane
Structured PhyUDM facts: identity, location, motion, health, mission and incident state, detections, annotations and provenance. Indexed for subscriptions, correlation and replay.
Media plane
Video, audio and high-volume sensor data stay in governed object storage or your existing VMS. PhyCloud holds a catalog of stream references, time bounds, permissions and hashes — not the bytes.
Action plane
Commands carry materially greater risk than observation, so they are isolated: explicit connector capabilities, step-up authentication, human confirmation and full command audit.
A hot store for right now. A lakehouse for everything since.
You should not have to know which tier holds an event. Query by source, incident, geography or time window and PhyCloud plans the retrieval across both.
“Across last year's missions, where did communication dropouts correlate with carrier coverage?”
Canonical history in open columnar files with signed manifests — for multi-year analysis, AI, compliance and bulk export. Managed by us, owned by you, or both.
Moving an event into a Parquet file never breaks the trust model: the original event hash stays stable and a signed manifest records exactly how immutable logical events were packaged.
One platform, many operations
Any organisation running physical systems from more than one vendor has the same underlying problem. The connector bundles differ by industry; the event contract, provenance model and storage tiers do not.
Talk to us about your operation
We are looking for design partners running physical systems from more than one vendor. Tell us how you reconstruct an incident today.