Technology & IP landscape

Not a drone detector.
A fusion, tracking and evidence platform.

elvone's own material describes time-aligned detections, 3D tracks, calibrated classification, heterogeneous sensor coincidence, evidence replay and multi-site correlation across modular RF, thermal, optical, acoustic and radar sensing cubes. Read that way, the interesting research problem isn't another drone-detection model: it's measurement-to-track association and multi-sensor fusion.

This page is a working research and patent-landscape brief, not a finalized roadmap and not legal advice. Treat every paper and patent named below as a starting point for a proper literature review and a formal freedom-to-operate search, not as a settled conclusion.

01 · The stack

Seven research layers

The priority is layers three and four: measurement-to-track association and multi-sensor fusion, not layer five in isolation.

01

Sensor physics & propagation

The radar equation, RCS, micro-Doppler, acoustic and RF propagation, thermal radiometry, atmospheric attenuation, clutter.

02

Detection

CFAR, beamforming, direction-of-arrival estimation, STFT and wavelets, background subtraction, thermal and optical detection.

03

Measurement-to-track association

Deciding whether a radar return, an RF emission, a camera box and an acoustic bearing are the same physical object.

04

Multi-sensor fusion

Combining heterogeneous, asynchronous, differently-uncertain measurements into one track, not one model per sensor.

05

Drone classification

RF fingerprinting, radar micro-Doppler, EO/IR and acoustic signatures, and fusing them rather than trusting any one alone.

06

Threat & risk estimation

Bayesian inference, trajectory prediction, intent estimation, geofencing, Remote ID consistency checks.

07

Evidence, replay & simulation

Deterministic reconstruction, provenance, model and calibration lineage, sim-to-real validation.

02 · Reading list

Ten papers, in order

  1. 1. Recent advances in multisensor multitarget tracking using random finite sets. A survey of RFS approaches (PHD, CPHD, GLMB, LMB, PMBM) for multisensor tracking. The foundation for the fusion engine: whether a radar return, an RF emission and a camera box are the same object is a data-association problem before it is a machine-learning one.
  2. 2. Asynchronous multi-rate multi-sensor fusion based on random finite sets. Addresses sensors with genuinely different sampling rates and arrival times, which is exactly the situation across radar, RF, optical, thermal and acoustic cubes running at different rates.
  3. 3. Towards an open-source simulation platform for counter-UAS sensor fusion (EU JRC). A C-UAS platform built specifically around multi-modal sensor fusion, tracking and visualization: a useful reference architecture to react against.
  4. 4. A counter-drone visualisation platform incorporating sensor-data fusion (EU JRC). A practical implementation using a message broker and Dynamic Time Warping to align trajectories that were never synchronized in the first place.
  5. 5. Detection and classification of multirotor drones in radar sensor networks: a review. Radar cross-section, micro-Doppler and rotor signatures, and why distinguishing a drone from a bird on radar alone is hard.
  6. 6. Multi-sensor fusion for UAV classification based on feature maps of image and radar data. Fuses thermal, optical and radar features directly, rather than voting between three separate single-sensor classifiers.
  7. 7. Single and multiple drone detection and identification using RF-based deep learning. RF fingerprinting and multi-drone identification, and the datasets that approach depends on.
  8. 8. Acoustic source drone detection using a tetrahedral microphone array and deep neural networks. 3D acoustic localization from a directional array: acoustic range is poor alone, but a bearing that intersects a radar uncertainty ellipse is informative.
  9. 9. Drone detection network based on RGB-thermal imaging multimodal fusion. Feature-level alignment between RGB and thermal, rather than running two detectors and merging their outputs after the fact.
  10. 10. Toward resilient multi-modal drone detection in cluttered environments. A current system-level survey comparing EO/IR, radar, acoustic, LiDAR and RF on range, false-alarm rate, latency and size/weight/power, useful for a sensor-posture engine.

03 · Sensor physics, not just physics

A digital twin needs a sensor model, not a rigid-body model

The question isn't “how does a drone fall.” It's “given this drone, this weather, this terrain and this RF environment, what detection should each sensor actually produce.” That maps directly onto elvone's own stated rule: physics stated, not assumed.

Radar

  1. 01Target state
  2. 02RCS model
  3. 03Propagation
  4. 04Clutter
  5. 05SNR
  6. 06CFAR
  7. 07Range / azimuth / elevation / Doppler
  8. 08Measurement covariance

Thermal

  1. 01Drone
  2. 02Motor / battery / ESC thermal model
  3. 03Atmospheric attenuation
  4. 04Background temperature
  5. 05NETD
  6. 06Point-spread function
  7. 07Image
  8. 08Detector

Acoustic

  1. 01Rotor RPM
  2. 02Blade geometry
  3. 03Acoustic spectrum
  4. 04Atmospheric attenuation
  5. 05Terrain / building reflections
  6. 06Microphone array
  7. 07Direction of arrival
  8. 08Measurement covariance

RF

  1. 01Transmitter
  2. 02Modulation
  3. 03Frequency / channel
  4. 04Propagation
  5. 05Interference
  6. 06Receiver
  7. 07IQ / spectrum
  8. 08RF features

04 · Build order

Simulate before you build five physical cubes

The simulator should emit the same detection messages the real cubes will, so the fusion engine, the classifier and the evidence pipeline can all be built and tested before the hardware is finished.

Phase 1

A software-defined digital twin

  • Synthetic drone trajectories in a physics simulator
  • Radar, camera, thermal, acoustic and RF sensor models against the same scene
  • One common detection schema across all five
  • A fusion engine producing 3D tracks from simulated detections

Phase 2

Break it on purpose

  • Inject latency and clock drift
  • Drop measurements and inject false ones
  • Simulate sensor outages
  • Vary sampling rates, weather and clutter

Phase 3

Add classification

  • Remote ID
  • RF fingerprinting
  • Radar micro-Doppler
  • Thermal and acoustic classification

Phase 4

Build the actual differentiator

  • Physics-aware, asynchronous, uncertainty-aware fusion
  • Deterministic evidence replay across all of the above

05 · Prior art already on the board

Don't file “multiple sensors detect a drone”

Generic radar, RF, optical and acoustic sensor fusion for drone detection is already crowded. These six are the ones to check first, not an exhaustive search.

CN118409309BCombines radar, acoustic and optical feature extraction with multimodal, ML-based classification and target positioning. Granted in China, January 2026.
US9715009B1Ground radar, RF antenna, optical/IR sensor, a sensor-fusion processor, 3D drone position and multi-sensor threat assessment.
US10670696B2An integrated detection, classification and interdiction architecture across radar, EO/IR and RF, with sensor fusion for locating, classifying and tracking.
CA3076695CDistributed sensor arrays and microcells, inter-sensor communication, and fusion at a central system: relevant to a distributed sensing-cube architecture.
WO2022172217A1Multiple sensor inputs, corroboration and fusion to establish a single drone "ground truth" for interception.
US20220069923A1Passive RF monitoring and extraction of physical-layer characteristics: bandwidth, power, center frequency, modulation, duty cycle. RF fingerprinting itself is heavily populated.

06 · Where the whitespace might be

Fourteen concepts, three buckets

CrowdedExisting prior artWorth investigating
Generic radar + RF + EO drone detectionCrowded
Generic multimodal drone classificationCrowded
RF drone fingerprintingCrowded
Radar micro-Doppler classificationCrowded
Distributed sensing cubesExisting prior art
Cross-site track identity & correlationExisting prior art, worth a closer look
Latency-aware heterogeneous fusionWorth investigating
Physics-aware probabilistic fusionWorth investigating
Dynamic heterogeneous coincidence thresholdsWorth investigating
Detection-only, low-bandwidth sensor fabricWorth investigating
Deterministic multi-sensor evidence replayWorth investigating
Simulation-to-real-flight calibration loopWorth investigating
Automatic sensor placement / posture optimizationWorth investigating
Sensor-health-aware fusionWorth investigating

07 · The thesis

“A physics-grounded, asynchronous, uncertainty-aware fusion and evidence engine for heterogeneous C-UAS sensors, coupled to a calibrated sensor digital twin.”

That covers radar, RF, EO/IR, acoustic, tracking, simulation, AI and evidence as one coherent program, while steering around the most obviously crowded patent territory.

← All documentsSee elvone Core