Autonomous · GPS-denied ·
Vertical Takeoff · Multi-Sensor Detection.
Real-hardware validated. Intercept demonstrated.
Built for threats others have no answer for.
DESIGNED FROM THE THREAT BACKWARDS.
Engineering Thesis
Existing counter-UAS architectures were designed for different threats. Three assumptions that no longer hold for Class I: that threats are RF-controllable, that platform size grants sensor range, that cost parity is irrelevant. Intercepting an FPV drone with a five-figure guided missile doesn't solve the problem. It shifts it.
Re-Definition
What does it actually take to reliably stop Class I FPV drones at scale? A platform that operates physically on the same plane in close range. Sensing without depth estimation. Guidance without black-box NN. An architecture follows.
FOUR ARCHITECTURAL DECISIONS.
Existing interceptor systems force drone hardware into missile-like geometries — to make classical guided-missile algorithms run on them. We take the inverse path: our algorithm is built from the ground up around the X-frame geometry that makes offensive FPV drones so dangerous today. We harness their physical strengths — instantaneous thrust vectoring, hover capability, sub-second response against evasive targets — instead of engineering them away.
Quad Architecture
Higher agility and higher lateral force than a fixed-wing. A quad platform meets an FPV threat in close range on the same physical plane. Lower unit cost than comparable fixed-wing platforms of the same class. One person, one system, one takeoff.
Minimalist & GPS-Denied by Design
One fixed-mounted camera and an IMU is enough. No GPS, no LiDAR, no stereo cameras, no gimbal. Onboard state estimation directly from the camera image — operates under electronic warfare and in GPS-denied environments without fallback paths to RF infrastructure. Structurally immune to RF jamming and therefore one of the few kinetic responses also to fiber-optic-controlled FPVs.
Vertical Takeoff
No ramp, no launch infrastructure. One-person-deployable, designed for mobile infantry and convoy operations. Where existing counter-UAS systems require stationary deployment.
Deterministic Guidance
Vision-based guidance with explainable, deterministic trajectories. Edge-AI exclusively for detection. The engagement decision runs through classical control engineering, not through black-box NN. Reduced sensor architecture, reduced failure modes.
WHY QUAD, NOT FIXED-WING.
| Criterion | Fixed-Wing | Quad (HERAKLES) |
|---|---|---|
| Engagement Geometry | Long-Range, Loiter | Close-Range, Pursuit |
| Lateral Agility | limited | high |
| Take-off Infrastructure | Ramp required | Vertical, infantry-deployable |
| Unit Cost | higher | lower |
| Threat-Class Match | Class II/III | Class I |
Class II/III fixed-wing interceptors address a different threat profile. For Class I FPV threats in the close-range infantry context, the quad architecture is structurally superior.
INTERCEPTOR ANATOMY.
THREE-PHASE DEFENSE.
Multi-sensor detection across three parallel paths:
- Acoustic detection in proximity range — triggers camera-swivel toward source
- Visual detection with wide-angle tracking after camera alignment
- Manual operator trigger: button press launches search mode, interceptor ascends and scans 360°
Operator-Authorization gates takeoff. Post-launch the system operates autonomously in the terminal phase. Disengage-Override available to the operator at all times.
Vision-based guidance with onboard state estimation. Kinetic collision as effector mechanism, no warhead against personnel or infrastructure.
VALIDATION STATUS.
In-house simulation environment with full flight dynamics, calibrated sensor-noise models and parameterised threat profiles. Batch-evaluation across thousands of engagement scenarios against evasive, accelerating and sensor-noisy targets. Identical code path between simulation and flight hardware closes the classic sim-to-real gap. Multi-seed testing with pre-registered pass/fail criteria.
Standardised bench routines for every hardware iteration: powertrain characterisation, sensor calibration, failsafe verification and authorization logic under realistic load profiles. Automated regression tests safeguard function and safety before any field deployment. Closed-loop engagement stable on bench under real-time constraints.
Controlled flight campaigns on secured test ranges. Step-wise escalation from platform verification through tracking engagements to a fully-autonomous intercept of an evasive FPV threat under operator authorization. Every real-flight feeds sim and bench back with empirical data. Target class, engagement geometry and test conditions disclosed with video.
TRL 6 — system prototype demonstrated in relevant environment.
ROADMAP.
[ Q1 2026 ]
Sim validation of the engagement pipeline completed — batch evaluation across broad engagement scenarios passed.
[ Q2 2026 ]
First Intercept ✓ — first fully-autonomous intercept of an FPV threat in real hardware. Hardware build and bench validation completed.
[ Q3 2026 ]
Multi-target engagement · Detection-range extension · Robust engagement against aggressive evasion maneuvers
[ Q4 2026 ]
Field validation in realistic environment · Platform iteration v2 · Hardening against weather and environmental variability
Real-Hardware-Validated. Pre-Seed Round opens Q3 2026.
Engineering briefings available under NDA.