SPEKTRA

A modular estimation and analytics core for heterogeneous data sources

SPEKTRA is a modular estimation and analytics core that processes data from different sources within one structure. The predictive maintenance, anomaly detection and fleet monitoring applications are built on top of it.

Every unmanned flight leaves a substantial amount of data behind, yet most of it stays in the logs and never becomes an operational decision. SPEKTRA processes those records so they can feed decisions running from maintenance planning through to fleet management.

One core, three applications

All three modules use the same estimation core. Adding a use case does not mean rebuilding the infrastructure.

Predictive maintenance

Estimates the remaining life of components and raises a warning before failure. Maintenance is planned against real usage rather than a fixed calendar.

Anomaly detection

Flags unusual behaviour in flight records. Operators see which flight diverged and where, on a single screen.

Fleet monitoring

Shows the status of many vehicles together. Flight profiles are compared and performance is tracked over time.

Four decisions that shape the product

These four choices shape the architecture and the way the product is used.

  • Vendor independentWorks with records from different autopilot ecosystems, PX4 and ArduPilot among them. A data abstraction layer keeps source formats away from the core.
  • Estimates with confidenceEvery estimate arrives with a confidence interval. Operators can see how firm a warning is and set priorities accordingly.
  • Shared estimation coreClassical probabilistic filtering and learned models run under one architecture. Adding a scenario does not mean rebuilding the infrastructure.
  • Data sovereigntyAn on-premise installation is supported for organisations with strict confidentiality requirements, keeping data within their own infrastructure.

A layered structure

The flow from data source through to delivery is separated into layers, with the application modules sitting on a shared core.

Data sources

TelemetrySensor streamsFlight logsMission parameters

Data abstraction layer

PX4, ArduPilot and vendor-independent adapters

Estimation and fusion core

Probabilistic estimation methods and learned models

Application modules

Predictive maintenanceAnomaly detectionFleet monitoring

Delivery

Cloud serviceOn-premise installation

Early access

SPEKTRA is in early access. Get in touch to discuss pilot deployments and data collaboration.