Sensor Fusion: Stone Soup Integration

Overview#

A single GMTI radar track loses a vehicle the moment it enters a treeline. An acoustic sensor picks it up on the other side. Without fusion, the operator sees a gap; with Stone Soup, the track continues unbroken, automatically correlated across sensor types. The Stone Soup integration brings this open-source multi-sensor track fusion directly into the Argus platform. Stone Soup is the Defence Science and Technology Laboratory (Dstl) open-source framework for tracking and state estimation, providing a library of filters, data associators, and fusion algorithms with no vendor lock-in. Argus submits detections from multiple sensor types into Stone Soup fusion sessions and polls fused track outputs, producing a correlated multi-sensor picture that is more accurate and complete than any single sensor source alone.

Key Features#

Open-Source Fusion Framework - Stone Soup is fully open-source (MIT license), eliminating vendor lock-in for the sensor fusion pipeline. The framework is maintained by Dstl and the open-source community, with peer-reviewed algorithms and transparent implementations.
Multi-Sensor Track Fusion - Fuse tracks from heterogeneous sensor types including GMTI radar, acoustic sensor networks, video analytics trackers, and manually reported positions. The fusion engine correlates detections across sensors using configurable data association algorithms, maintaining fused tracks that incorporate measurements from all contributing sensors.
Detection Submission - Submit detections from any platform sensor to a Stone Soup fusion session via the GraphQL API. Each detection includes sensor type, position estimate, measurement uncertainty, timestamp, and source identifier. The fusion engine processes submitted detections against existing tracks, either updating existing tracks or initiating new ones.
Track Polling - Poll fused track states from active fusion sessions. Each fused track returns the current estimated position, velocity, uncertainty ellipse, contributing sensor list, track quality score, and last update time. Tracks can be displayed on the operational map with uncertainty visualisation.
Configurable Fusion Parameters - Configure fusion session parameters including process noise model, measurement noise model, data association gate size, track initiation threshold, and track deletion criteria. Different configurations can be applied for different operational scenarios (high-clutter, low-clutter, mixed sensor environments).
NATO Interoperable - Detection and track formats align with NATO standards, enabling fusion of data from coalition sensor systems that provide STANAG-compliant outputs.

Use Cases#

Multi-Sensor ISR - Fuse GMTI radar tracks with acoustic detections and video analytics to build a comprehensive ground picture with higher confidence than any single sensor.
Track Continuity - Maintain target tracks when a target moves between sensor coverage areas -- acoustic coverage picks up where radar coverage ends, maintaining continuous tracking.
Clutter Reduction - Use multi-sensor correlation to distinguish genuine targets from sensor clutter, reducing false positive rates through cross-sensor confirmation.
Coalition Sensor Integration - Fuse sensor data from multiple coalition partners' systems into a shared operational picture without requiring uniform sensor types.

Integration#

GMTI Radar Monitoring - GMTI tracks are submitted to Stone Soup for fusion with other sensor sources.
Acoustic Sensor Network - Acoustic triangulation fixes are submitted as detections for multi-sensor fusion.
Military Video Analytics - Video-derived tracks with geocoded positions are submitted for fusion.
Unified Operational Events - Track fusion updates generate events in the unified timeline.
Common Operational Picture - Fused tracks are displayed on the operational map with source provenance indicators.

GraphQL: sensorTracks, sensorTrack, stoneSoupFeedConfigs (queries); pollStoneSoup, saveStoneSoupConfig (mutations).

Open Standards#

STANAG 4607 Ed 3.0 (GMTI): The platform implements a full binary parser for STANAG 4607 Ground Moving Target Indicator packets, decoding dwell segments, target locations via the SA32 geodetic format, and SNR data, which are then submitted to Stone Soup fusion sessions as detections.
STANAG 4676 Ed 1 (Track/Entity Association): Track messages are parsed from and serialised back to both the JSON and XML variants of STANAG 4676, enabling coalition partners to exchange fused track data using NATO-standardised identity and environment classifications.
Cursor on Target (CoT) 2.0: TAK/CoT contacts are normalised from CoT fields (uid, type, hae, ce, le) into the fusion detection format and ingested as a named source type, preserving circular and linear error values as quality scores.
ITU-R M.1371 (AIS): Automatic Identification System maritime position reports are accepted as a named source type and routed through the unified fusion pipeline, with tracks placed in the MARITIME operational domain.
RTCA DO-260B (ADS-B): Automatic Dependent Surveillance-Broadcast aviation transponder updates, identified by ICAO24 address, are accepted as a named source type and fused in the AIR operational domain.
WGS-84: All sensor positions, target locations, and fused track coordinates are expressed in WGS-84 geodetic coordinates; great-circle distances derived from WGS-84 are used in GOSPA and OSPA track quality metric computations.
ISO 8601: Timestamps on detection payloads, track query filters, and all persisted track records are formatted as ISO 8601 UTC strings throughout the integration.

Last Reviewed: 2026-04-02 Last Updated: 2026-04-14