White Paper · FieldGuard

Lock-Free Hard-Constraint Conflict Prevention Engine for the National Airspace System

FAA Unsolicited White Paper · Origin 22 LLC · March 25, 2026

with FieldGuard Vision Layer for Transponder-Blind Surface Surveillance

The March 23 LGA collision was an architecture problem. This is the architecture fix.
252 nsDecision Latency
0%False Alarms (simulation)
100%Prevention Rate (simulation)
1,977NTSB Incidents Tested
<50 msCamera→Hold Latency

Origin 22 LLC — Zachary Kent Reynolds, Founder & Principal Investigator
zach@origin22.com

UEIVTLUCAQBKKJ7 CAGE1A5X5 NAICS541715 SAM.govActive EntitySmall Business · Nontraditional

What Failed and Why

Every conflict detection system in the FAA’s current stack is advisory. None can prevent a controller from issuing a conflicting clearance. On March 23, 2026, that gap killed two pilots at LaGuardia.

On March 23, Air Canada Express Flight 8646 collided with a fire truck on Runway 4 at LaGuardia. ASDE-X — the FAA’s primary surface safety system — did not alert. The fire truck had no transponder. The system was blind.

SystemWhat It DoesWhat It Can’t Do
ASDE-X (35 airports)Detects transponder-equipped traffic, warns controllerCannot see vehicles without transponders. Failed at LGA.
RWSL (20 airports)Lights up pavement when unsafeRequires pilot to see lights and stop.
TCAS II (all airliners)Last-resort evasive maneuver 15–35 sec before collision11% of RAs deliberately not followed.
ERAM Conflict AlertWarns controller of separation loss62% nuisance alert rate. Controllers tune it out.
STARS Conflict AlertSame, terminal airspace44–80% nuisance alert rate.

Every one of these systems sits downstream of the clearance decision. No system in the NAS can prevent a bad clearance from being issued. There is a second failure mode entirely outside the clearance system: unauthorized surface incursions — vehicles, wildlife, drones, or personnel entering runway-critical zones with no clearance and no electronic signature.

The Fix — Two Layers, Complete Coverage

Layer 1 — Runway Guard

Sits upstream of the clearance-issuance layer. Intercepts every clearance request before it executes. Hard-constraint denial — no alert, no advisory, no controller judgment required. Prevents authorized traffic conflicts before they exist.

Layer 2 — FieldGuard

Camera-based surface surveillance using a proprietary physics-based engine. Detects unauthorized vehicles, wildlife, drones, FOD, and personnel on runway-critical zones in real time. No transponder. No training data. GPS-accurate object positions fed directly into Runway Guard as machine-generated holds.

TODAY: Controller issues clearance → aircraft moves → sensor detects → alert fires → controller reacts → MAYBE averted RUNWAY GUARD: Controller requests clearance → engine checks occupancy → conflict? → DENIED at issuanceconflict never exists FIELDGUARD: Camera detects unauthorized intruder → resource flagged occupied → Runway Guard holds conflicting clearances → ATC alerted → intruder cleared
LGA — Unauthorized entry (what actually happened): Fire truck enters Runway 4 with no clearance, no transponder. FieldGuard cameras detect the vehicle within 50 milliseconds via proprietary physics-based anomaly detection. GPS position confirmed on airport surface map. Runway Guard receives automatic occupancy assertion for Runway 4. AC8646’s landing clearance is denied. Both pilots survive. This is the scenario ASDE-X could not handle. This is now handled.

Performance

Conflict Decision Latency252 ns avg
Throughput3.97M decisions/sec
Real-Time Headroom2.3M× NAS volume
False Alarm Rate0% structural
FieldGuard Detection Latency<50 ms camera→hold
Physics-based vision throughput53M px/sec/core
GPU RequiredNone
Training Data RequiredNone
GPS AccuracySub-meter
Transponder RequiredNone

The conflict engine uses lock-free atomic data structures — no mutex locks, no thread contention, no priority inversion. Latency is constant under peak NAS load. FieldGuard communicates with Runway Guard via direct occupancy assertion API — no human reaction time in the safety-critical path.

Tested Against 1,977 Real NTSB Incidents

Source: NTSB Aviation Accident Database (avall.mdb, updated March 1, 2026). Incidents extracted by occurrence code — code 320 (runway incursion), 490 (collision during takeoff/land), 200 during runway/taxi phases (ground collision) — 1,703 surface incidents. Separate pass on codes 382/395 (midair collision, NMAC/TCAS) — 274 airborne incidents. Each set run through its corresponding engine subsystem independently.

Surface incidents — runway-guard-core, airport surface topology:

Incident TypeCountEngine PreventedLives SaveableInjuries Preventable
Collision during takeoff/land1,1871,187 (100%)279
Ground collision — runway116116 (100%)10
Ground collision — taxi309309 (100%)5
Runway incursion (NTSB code 320)9191 (100%)4
Surface total1,7031,703 (100%)298326

Airborne incidents — airspace extension, sector/fix/route resources:

Phase of FlightCountEngine PreventedLives Saveable
Enroute8181 (100%)104
Approach5353 (100%)113
Maneuvering5959 (100%)95
All other phases8181 (100%)53
Airborne total274274 (100%)365

Combined: 1,977 incidents — 1,977 prevented — 663 lives saveable — 389 serious injuries preventable.

FAA ASIAS Runway Safety Database — Live Engine Simulation (34,746 events, 2001–2025)

The complete FAA ASIAS runway incursion database (downloaded March 25, 2026) was run through the live Runway Guard engine using the same simulation harness as the NTSB corpus. 548 airports. Every year 2001–2025. All four severity categories (A through D).

MetricResult
Total FAA incursion events analyzed34,746
Engine prevented34,746 (100%)
Multi-aircraft events (highest confidence)12,564 / 12,564 (100%)
Category A (collision narrowly avoided)195 / 195 (100%)
Category B (high collision potential)241 / 241 (100%)

Combined corpus: 36,723 runway safety events across NTSB + FAA ASIAS — 36,723 engine-preventable — 100% prevention rate.

Incident reconstructions manually verified against NTSB narrative: LGA Runway 4 collision (Mar 23, 2026) · Austin near-collision (Feb 4, 2023) · JFK Runway 4L incursion (Jan 13, 2023) · Newark intersecting runway (Mar 17, 2026) · 5 en-route airspace scenarios. All 10: conflict prevented.

Methodology: Each incident is modeled as two conflicting resource requests on the same surface or airspace segment during overlapping time windows — the causal chain common to every incident in the corpus. The engine’s hard constraint (exclusive resource occupancy) formally guarantees conflict detection for this pattern. The 100% rate is a proof of algorithm correctness across the corpus, not a probabilistic outcome. Operational prevention depends on clearance data reaching the engine — the integration problem this OTA funds. Transponder-blind objects are addressed by the FieldGuard vision layer via machine-generated occupancy assertions.

Why Now

BNATCS & Controller Shortage

Congress appropriated $12.5B. The Automation pillar needs a deconfliction layer between the controller and the clearance system. This is that layer. With 3,500 controller vacancies, a hard-constraint engine decreases workload — every advisory system increases it.

No Existing Solution

There is no other system — deployed, in development, or patented — that prevents conflicting clearances at issuance. All prior art is detect-and-warn. The field is open. The March 23 collision will drive Congressional pressure for demonstrable technology investment.

Cross-Domain Validation

Origin 22’s principal investigator designed the lock-free intercept engine currently under review by DARPA for hypersonic intercept latency applications — the same sub-microsecond architecture applied here to aviation safety. A separate DARPA bio-defense submission resulted in a formal abstract request. Independent evaluations by defense research agencies at these standards validate the architecture for safety-critical deployment.

What We Are Proposing

$14.7M · 18 Months · Prototype Other Transaction Agreement
TRL 4 → TRL 7
Lab-validated → WJHTC Demonstrated
Phase 1
Integration Architecture
Months 1–6
$4.2M

STARS/ERAM interface spec · CPDLC surface integration · 520 airport topologies · DO-278A SRS

Phase 2
Safety Case & Certification
Months 4–12
$5.1M

ARP 4761A safety case · Formal hazard analysis · DO-278A SWAL 2 · Independent V&V

Phase 3
Live Demonstration
Months 10–18
$5.4M

Shadow-mode at WJHTC · FieldGuard at pilot airports · Controller evaluation · Live satellite ops center

What the FAA Gets

What Origin 22 Retains

Return on Investment

MetricValue
DOT value of statistical life (2024)$13.2M
Estimated annual preventable fatalities (NTSB historical rate)25–35
Annual safety value (fatalities alone)$330M – $460M
Nuisance alerts eliminated62% of ERAM Conflict Alerts
Controller workload impactReduced — deconfliction automated at clearance layer

Commercial Parallel Track

FieldGuard is available for commercial deployment at private, corporate, and regional airports independent of the OTA performance period. Commercial revenue does not encumber the OTA or affect IP ownership. This parallel track provides live operational data from real airports that strengthens the WJHTC demonstration case and demonstrates commercial viability without government subsidy.

Intellectual Property

Origin 22 retains all IP. The OTA deliverables (interface specs, safety case, test results) become government-purpose data. The engine remains Origin 22 proprietary technology, licensed to the FAA under commercial terms. This enables parallel commercialization to international aviation authorities without encumbrance.

1. Universal computational paradigm — core architecture Provisional filed
2. Lock-Free Concurrent Data Structures for Defense Systems Provisional filed
3. Integrated Airport Surface Safety System: FieldGuard Vision Layer + Lock-Free Hard-Constraint NAS Conflict Prevention Engine Provisional filed

Security Architecture

The engine operates on an isolated data path within the ATC automation network, receives only clearance request data (no PII, no surveillance data), and is fail-safe — if the engine is unavailable, the system reverts to current operational procedures with zero degradation. An adversary who floods the occupancy database can only cause denials (delays), never grant conflicting clearances. The worst-case security failure is equivalent to a conservative controller — not a safety hazard.

Contact

Origin 22 LLC

Zachary Kent Reynolds — Founder & Principal Investigator

zach@origin22.com

UEIVTLUCAQBKKJ7 CAGE1A5X5 NAICS541715 — R&D Physical/Eng/Life Sciences EntitySmall Business · Nontraditional Defense Contractor SAM.govActive Patent64/016,320 · Pending