RELIEF NAVIGATION SYSTEM / RNS

Route intelligence for convoys, when the map is no longer true.

RNS reads paired pre-event and post-event satellite imagery, detects road network damage, classifies segments by vehicle load class, and delivers a short ranked set of candidate routes for a human planner to verify and authorize. Built for humanitarian logistics and civil-protection operations.

Request a briefing How the system works

STANAG 2174 / MLC REFERENCEDHUMAN-AUTHORIZED ROUTINGPROOF-OF-CONCEPT STAGE

01 / THE PROBLEM

After a disaster, the map is wrong. Finding out how wrong is slow.

In the first days after an earthquake, flood, or conflict event, road condition is the decisive unknown for relief movement. Imagery exists. Trained analysts exist. What is scarce is route knowledge: which specific segments, out of everything visible, actually decide whether a loaded convoy gets through.

Current practice is manual map reconnaissance. Analysts trace damage by hand, planners hold route options in documents and spreadsheets, and verification requests compete with everything else the response is trying to do. Hours pass while the binding constraint on the entire supply chain sits unresolved.

The bottleneck is not imagery availability. It is the conversion of imagery into a verified, vehicle-appropriate routing decision that a responsible officer can sign.

MANUAL RECON

Analyst hours per network

Damage tracing is done by hand against archived imagery, segment by segment.

SEARCH SPACE

Dozens of candidate segments

Only a small fraction determine convoy feasibility. Finding them is the hard part.

ACCOUNTABILITY

Decisions need a named owner

A route must be authorized by a person, with evidence, under doctrine.

02 / THE SYSTEM

From imagery to a short ranked set of authorized routes.

The pipeline converts paired satellite imagery into a routing decision a planner can sign: detect damage, class each segment by vehicle load, generate constrained candidates, and verify the critical few before dispatch.

Sensory

Change detection from paired satellite imagery.

Pre-event and post-event imagery of the same road network is compared to surface probable damage: debris fields, washouts, bridge failure, blocked carriageways. The output is a damaged-network hypothesis, not a finished answer.

PRE/POST IMAGERYCHANGE DETECTIONDAMAGE HYPOTHESIS

Triage

Load-class classification of road segments.

Each segment is classified by the vehicle load it can plausibly carry, referenced to NATO STANAG 2174 Military Load Classification. A truck convoy and a light vehicle column are different routing problems, and the system treats them that way.

STANAG 2174MLC CLASSINGSEGMENT TYPING

Conduction

Constrained route generation.

From a wide field of possible road segments, the system generates a short ranked set of candidate routes under the operator's constraints: vehicle class, payload, turnaround requirements, known hazards. Twelve candidates collapse to the critical few that decide whether the convoy gets through.

CONSTRAINT SOLVINGRANKED CANDIDATESBOTTLENECK ISOLATION

Reflex

Verification with full provenance tracking.

Candidate routes are dispatched for ground or drone verification before authorization. Every confirmation or correction is logged against the segment record, so the next assessment starts from verified knowledge rather than imagery alone.

GROUND / UAV VERIFYPROVENANCE LOGFEEDBACK LOOP

The collapse of the search problem. A typical event area presents twelve or more plausible road segments per axis. RNS reduces that field to the critical few that determine convoy feasibility, ranked with constraints and confidence states attached.

12+ SEGMENTS → 3 RANKED ROUTES

03 / WHY RNS IS DIFFERENT

Designed for accountability, built around provenance.

Provenance and confidence states

Every segment carries an explicit state: AI-assessed and unverified, or ground-confirmed with source and timestamp. Planners always know which parts of the route are inferred and which are verified. This separation is the safety-critical core of the system, not a reporting feature.

Human authorization at the final step

RNS ranks candidate routes. It does not dispatch. The final route authorization remains with a named human operator who is legally and doctrinally accountable for the decision, with the system's evidence trail attached.

Deliberate scope limits

RNS is not a crowdsourced tool, not an autonomous router, and not a public data platform. Inputs are tasked imagery and verified field reports. Outputs go to the responsible planner, and nowhere else.

CONFIDENCE-STATE MODEL

AI-ASSESSED / UNVERIFIED

Derived from imagery change detection. Carries a model confidence score and the imagery pair it was assessed from. Usable for planning. Never sufficient for authorization alone.

GROUND-CONFIRMED

Confirmed by field team or UAV tasking. Carries the verification source, method, and timestamp. This is the only state on which a route may be authorized.

CONCEPT DEMO

See a synthetic walkthrough of the four-stage pipeline on a sample mudslide scenario. Open the demo.

04 / WHO IT IS FOR

For humanitarian relief organizations in needs assessment.

At proof-of-concept stage, RNS is scoped to one user community and one phase: relief organizations that must decide whether routes are passable while needs assessment is still underway. These are the kinds of organizations we are shaping the system around. They are not current customers, and no endorsement is implied.

International humanitarian NGOsField teams assessing access and route feasibility during the first hours of response.
UN humanitarian agenciesOrganizations that coordinate multi-agency needs assessment and logistics planning.
National disaster response authoritiesGovernment bodies that own civil-protection needs assessment and initial relief movement.
Relief assessment teamsMobile units that verify damage and access conditions on the ground or by drone.

05 / STAGE AND STATUS

Proof of concept. Looking for the right validation partner.

Myelix is pre-revenue and the RNS pipeline is at proof-of-concept stage. We make no claims of deployed operations, existing contracts, or measured field performance. What we have is a working architecture, a defined doctrine for human authorization, and a clear validation plan.

WHAT WE ARE ASKING FOR

A design partner. A logistics or geospatial team willing to shape requirements against real doctrine and operational constraints.
A past event for validation. Access to a documented historical response, with its imagery archive and ground truth, so we can test the pipeline against decisions that were actually made.
Honest evaluation. We would rather fail a rigorous pilot than pass an easy demo.