Problem

Hard to see, hard to stop, no time

Small drones and balloons fly low and slow over cities, where existing air defense has blind spots. Better detection alone is not enough.

Goal

Decide on a low-risk response in time

Support operators in real time, and place radars and jammers in the right spots before an incident, so there is time to decide.

Threat

Three barriers at once

01Low signature

Low altitude, low speed, small radar cross-section. Wide-area radar struggles, and city clutter causes false alarms.

02Collateral damage

Shooting something down over a crowded city can cause serious secondary damage.

03Golden time

Judging a target perfectly in a few seconds is physically impossible.

Threats by radar cross-section and altitude

Based on published RCS measurements (Drones 7(1):39, Sensors 19(22)).

System 1 · Reactive

Counter-UAS: detect, identify, respond

Radar, RF and camera data go through one pipeline that scores each threat and suggests soft or hard response options. The simulator and real sensor input use the same interface.

01Fused detection

Radar and RF combined in space and time

02Classification

Drone, balloon or bird

03Threat assessment

Score from 0 to 100 with AHP weights

04Response support

Soft or hard options ranked with TOPSIS, including collateral damage

Real-time operating picture with alerts, multi-track tracking, an engagement approval queue and LLM queries.

bunkr OPERATE mode: live drone and bird tracks over Gwanghwamun Square, a critical threat alert and the event log — OPERATE mode: live tracks, engagement approval queue, threat level and event log.

Insight

Real-time decisions only work if the radars and jammers are already in good positions. With the same decision engine, placement decides how much time you have.

T+0 · intrusion startsT+60s · reaches target

Random placementDetected late: 8 s to decide

Optimal placementDetected early: 45 s to decide

Values use test data.

System 2 · Proactive

Deployment optimization

Before an incident, which assets should go where to stop the most threats? The system starts from public data: critical facilities, population by time of day, terrain, roads, airspace and building heights for line of sight.

Each area gets a weight from protected zones, foot traffic and nearby facilities. Rooftop candidate sites are scored by what each device can cover, and the system picks sites that maximize weighted coverage across the full chain:

Radar: detectRF scanner: detect and identifyCamera: identifyJammer: soft-kill

Choosing sites under an installation limit is NP-hard, so it uses a greedy approximation that guarantees at least 63% (1 − 1/e) of the optimum, and suggests placement options.

bunkr DEPLOY mode: optimal radar, scanner and jammer placement on a 3D map of Gwanghwamun Square with a coverage panel — DEPLOY mode: optimal placement on a 3D map with coverage KPIs.

System

One defense cycle

Together, the two systems turn after-the-fact response into a full cycle: plan the placement before an incident, and support decisions in real time during one.

Stakeholders

Three groups, one screen

Procurement

"Is it worth buying?"

Optimal placement, ROI and budget justification.

Command

"Can I approve this?"

Threat awareness and the risk of misjudging an engagement.

Operators

"Is it easy to use?"

Readable screens and protection against wrong inputs.

Software engineer who designs

Software engineer
who