Skip to content
    Arc-RF · Signal Intelligence

    See the Spectrum.
    Build the Picture.

    Arc-RF turns spectrum observations into model-readable evidence. It classifies supported RF signatures, maps detections back to physical frequency coordinates, and can pair calibrated direction finding with TAK publication in a program-configured system.

    97.2% Controlled mAP@501.5 ms Model-OnlyDirection FindingTAK / CoT Ready
    Arc-RF direction-finding ground station with tactical display
    The Problem

    The Spectrum Is Crowded & Hostile.

    Relevant emissions compete with Wi-Fi, Bluetooth, telemetry, interference, and multipath. A useful RF picture must preserve what was observed, distinguish signatures the model has been trained to recognize, add bearing evidence where the hardware supports it, and show uncertainty instead of reducing every signal to an alarm.

    97.2%
    Controlled RF Benchmark
    mAP@50 across 37 labeled drone and controller classes
    1.5 ms
    Model-Only Inference
    Per-spectrogram benchmark; capture, STFT, and transport excluded
    3
    Physical RF Outputs
    Center frequency, occupied bandwidth, and burst duration
    CoT
    Operator Integration
    TAK publication when configured with the program network
    RF-YOLO pipeline / spectrum to location

    See the signal.
    Build the evidence.

    Arc-RF converts spectrum observations into visual representations that edge models can detect and classify. Compatible calibrated hardware can add direction-finding evidence; configured nodes can combine bearings into an estimate and explicit uncertainty for the operator picture.

    Arc-RF / operational flowOperational flow model
    1. 01

      Capture

      Recorded or configured spectrum observations

    2. 02

      Render

      Time-frequency representation

    3. 03

      Detect

      Edge model finds signal signatures

    4. 04

      Bear

      Calibrated direction finding adds geometry

    5. 05

      Publish

      Estimate, uncertainty, and confidence to TAK

    Interfaces / modes
    Single nodeCooperative nodesOrganic UASTAK / CoT
    Operational result
    RF energy / classified event with location evidence when configured

    Configuration note / Classification, speed, bearing, and geolocation performance vary with the receiver, model, RF environment, antenna geometry, and deployment.

    Capabilities

    RF Evidence. At the Tactical Edge.

    Spectrum as Vision

    Arc-RF converts recorded I/Q observations into spectrograms, detects learned signal signatures, and maps each detection back to center frequency, bandwidth, and duration. Classification depends on the trained model and the RF conditions it has been validated against.

    Phase Interferometry DF

    When paired with a coherent multi-channel receiver and calibrated antenna array, Arc-RF can add angle-of-arrival evidence to a classified signal. Bearing performance depends on calibration, array geometry, SNR, multipath, and installation.

    Cooperative Geolocation

    Program-configured nodes can exchange bearing evidence and combine intersecting lines of bearing into a location estimate and uncertainty region. Accuracy depends on node geometry, synchronization, transport, and the local RF environment.

    Organic UAS Baseline

    A compatible UAS can carry an additional receiver to improve geometry where fixed nodes are impractical. Flight authority, route selection, communications, and safety behavior remain part of the integrated program configuration.

    Field Power & PACE

    Available node configurations can pair battery and solar-support options with program-selected communications bearers. Endurance, duty cycle, mesh, cellular, and satellite availability vary by hardware and network plan.

    TAK / CoT Integration

    Arc-RF can publish standardized Cursor on Target events into ATAK or WinTAK when integrated with the mission network. Each event carries only the identity, bearing, location, and confidence fields available from the configured sensors.

    Deployment Configurations

    Scales from a Scout to a FOB.

    A

    Standalone Ground Station

    A single field node monitors a configured band, classifies supported signatures, and can publish a line of bearing into TAK. Coverage and confidence vary with antenna placement, receiver, model, and spectrum conditions.

    B

    Cooperative Node Set

    Multiple calibrated nodes combine bearing evidence into a location estimate with an explicit uncertainty region. The program selects the transport path; geometry, synchronization, terrain, and multipath determine resulting accuracy.

    C

    Node + Organic UAS

    A compatible UAS can extend the measurement baseline for mobile or infrastructure-limited missions. Result timing and accuracy depend on airspace, route authority, receiver integration, link availability, and observation geometry.

    The Pipeline

    From I/Q to RF Evidence.

    Arc-RF applies a repeatable evidence pipeline: convert I/Q observations into time-frequency views, detect supported signatures, preserve physical coordinates, and add direction-finding geometry when the selected receiver and antenna configuration provide it.

    01
    Ingest
    A configured receiver or recording supplies raw I/Q
    02
    Visualize
    STFT renders 640×640 spectrograms with metadata
    03
    Classify
    A validated model detects supported signal classes
    04
    Publish
    Configured DF adds bearing or location evidence to CoT

    Turn Invisible Activity Into Mission Evidence.

    Arc-RF combines spectrum classification, configuration-dependent direction finding, cooperative geolocation, and TAK integration in a field-oriented architecture. Get in touch to review the receiver, antenna, model, network, and power configuration for your program.

    Data rights / governance

    Sovereign by design.
    Open by choice.

    Arc builds on open-source foundations, including LlamaFarm and Atmosphere. Programs can deploy core workflows inside customer-controlled infrastructure, define data boundaries, and negotiate model, sensor-data, and mission-log rights in the acquisition terms. Open interfaces preserve an exit path without promising that every mission dependency is open source.

    LlamaFarm on GitHubDocumentation
    ■Disconnected Training Option■Open-Source Foundations■Program-Defined Data Rights■Modular Acquisition Path
    Program office / contact

    Put Arc
    on the mission.

    Defense inquiries

    [email protected]

    Partnerships & commercial

    [email protected]

    Powered by Rownd