// AND-00 · NEICRONE ANDROMEDA PROJECT · CONSOLE LIVE

Andromeda

Geospatial suitability engine for physical AI deployment. Andromeda scores real parcels against the hazard, terrain, access, and capacity constraints that decide whether an embodied-AI site can actually be built, crewed, and operated.

Open the Console Read the Suitability Model

// DEFAULT WEIGHT VECTOR · BALANCED APPRAISAL

25%
Flood safety: 20%
Buildable slope: 20%
Transit access: 15%
Landslide safety: 15%
Zoning capacity: 5%
Parcel size

// Problem Frame

0108

Physical AI lives or dies by location.

Sandboxes, field trials, and logistics integration all inherit their site. The wrong parcel is not a setback; it is a write-off.

// SITE SELECTED BY INTUITIONUNPRICED

A wrong site is paid for twice.

Once when the ground fails to support the programme, and again when the evidence it produced turns out to be unusable. Slope, flood exposure, buildable envelope, and access are not discovered by walking a parcel — they are measurable before anyone stands on it.

// WHAT IT COSTS

Instrumentation sunk into a parcel that floods one season in four
Safety case reopened because the ground moves under a fixed rig
Data quality capped by a footprint too short to hold a full trajectory
Regulatory friction discovered after capital is committed, not before

// SITE SELECTED BY MODELSCORED

Systematic parcel appraisal.

Andromeda scores real cadastral parcels against the constraints that matter for embodied AI operations, under a weighting that is stated rather than assumed. The output is a ranked, defensible shortlist — and the vector that produced it, so the reasoning survives the meeting.

// WHAT IT DELIVERS

Every parcel scored against the same six constraints, city-wide
Hazard treated as an exclusion, not a footnote on a brochure
Weighting made explicit, published, and adjustable per deployment intent
Shortlists that carry their weight vector with them into review

// Capability

0208

What Andromeda does.

Six capabilities over one scored surface. Everything below runs against real cadastral geometry, not a sample.

AND-01

Multi-criteria parcel scoring

Every parcel in an ingested city carries a 0-100 suitability score under the active weight vector. Move a weight and the whole surface re-scores.

AND-02

Hazard-aware filtering

Flood severity, landslide susceptibility, and slope are first-class exclusions, not annotations. Hazard-flagged parcels can be dropped from the surface entirely.

AND-03

Operational access scoring

Walking catchment to transit, permitted floor area ratio, and developable footprint — the three constraints that decide whether a site can be crewed and built.

AND-04

Scenario weight presets

Seven shipping profiles spanning planner, investor, and policymaker intent. A city may publish its own on top without displacing the built-ins.

AND-05

Shortlist generation

High-scoring candidates ranked, compared side by side, and exported as a report with the weight vector that produced them attached.

AND-06

Timeline-ready geometry

Selected sites carry their identifiers forward, so field telemetry and evidence land against the parcel that was scored.

// Suitability Model

0308

Six criteria, one vector.

The default profile. Every score on the map is this vector applied to a parcel — nothing is hidden behind it.

// FACTOR// WEIGHT// SCORING LOGIC// WHY IT IS WEIGHTED

Flood safety
25%
Distance from flood severity
Standing water ends an instrumented run and destroys edge hardware.
Buildable slope
20%
Flatter parcels score higher
Grade limits hardstanding, rig placement, and low-speed manoeuvre envelopes.
Transit access
20%
Walking distance to a stop
Crew rotation and partner attendance decide how often a site is actually used.
Landslide safety
15%
Low susceptibility scores higher
Unstable ground is a permanent constraint, not a seasonal one.
Zoning capacity
15%
Permitted floor area ratio
Permitted envelope decides whether enclosure and hangarage are legal at all.
Parcel size
5%
Larger developable footprint
Contiguous footprint sets the longest trajectory a sandbox can hold.

// NORMALIZATION

Weights are relative, not percentages. Scoring divides by their sum, so a vector need not total 100 — raising one factor lowers every other factor's share without anyone rebalancing the rest by hand. The percentages above are that derived share.

1.00RAW SUM

// VECTOR, AS ENCODED IN THE TILE URL

sl.20_fl.25_ls.15_tr.20_zf.15_ar.05

Every distinct slider position is its own cacheable URL, so the map swaps tile sources on a weight change rather than re-querying the same geometry.

// Layers

0408

What the map can show.

Toggled per session. Data and hazard layers require an ingested city; terrain works anywhere on Earth.

Cadastral and service geometry, served from the ingested city.

Parcels (suitability)LIVE
The scored surface. Colour is the score under the active weights.
Zoning districtsLIVE
Permitted use and floor area ratio, as ingested from the planning authority.
Transit stopsLIVE
Stop geometry behind the walking-catchment criterion.

Exposure surfaces. Either a scoring input or a hard exclusion.

Flood zonesLIVE
Severity-banded extents. Drives the flood-safety criterion.
Landslide susceptibilityLIVE
Graded instability. Low susceptibility scores higher.

Elevation-derived, and the only group that works with no cadastral ingest at all.

HillshadeLIVE
Relief shading from the open terrain tiles.
3D terrainLIVE
Terrarium-encoded elevation, extruded under the scored surface.

Scoped, not yet ingested. Each becomes a criterion as well as a layer.

InfrastructurePLANNED
Grid capacity, water, and haulage-grade road access.
OwnershipPLANNED
Title and tenure, so a shortlist can be filtered to what is actually acquirable.
Environmental constraintPLANNED
Protected designations and noise envelopes that gate outdoor operation.

// Scenarios

0508

Different goals, different weights.

Seven profiles ship with the engine. A city may publish its own on top; the built-ins are never displaced.

Balanced appraisal

Even footing across capacity, access and hazard

Industrial & logistics

Flat ground and large contiguous footprints; transit matters less

Hazard resilience

Screens out exposure first — flood, landslide and unstable ground

Transit-oriented development

Walkable catchment dominates; suits corridor and station-area studies

Investor — development yield

Buildable capacity and access lead; hazard is a discount, not a veto

Residential expansion

Habitable slope and services, with real weight on flood safety

Terrain screening only

For places with no cadastral data — scores only what elevation can tell us

// BAND ORDER01Flood safety02Buildable slope03Transit access04Landslide safety05Zoning capacity06Parcel size

// DEPLOYMENT INTENT · STARTING PROFILE06 MAPPED

Physical AI sandboxSBX-02Balanced appraisal
Default vector. Hazard exclusion on — a sandbox cannot carry a flood discount.
Field Ops staging areaFLD-01Industrial & logistics
Slope and footprint lead; transit retained for crew rotation.
Logistics yard integrationLOG-06Industrial & logistics
Same vector, minimum-area floor raised to yard scale.
High-safety / regulated siteHIL-05Hazard resilience
Exposure screened before capacity is considered at all.
High-throughput / access priorityLOG-06Transit-oriented development
Walking catchment dominates; suits corridor and station-area studies.
Unsurveyed regionFLD-01Terrain screening only
No cadastral ingest required — scores what elevation alone can support.

// Status & Access

0608

What is actually running.

Stated plainly, including what is switched off. A suitability engine that overstates its coverage is worse than no engine.

Authentication
NONE IN THIS BUILDOPEN
The console published here carries no identity provider. Nothing is signed in, so nothing is attributed: every write path is closed and the surface is read-only. The standalone engine resolves per-city membership — investor, planner, admin — from an organization session; that build is provisioned per engagement.
Authorization boundary
DATA LAYER, NOT EDGEGATED
Route matching at the edge decides what is reachable; it does not decide what is readable. City access is resolved against membership at the query, so a public route still returns nothing for a city a caller cannot read.
Anonymous access
PER-CITY ALLOW-LISTGATED
Ingested cities are readable only when explicitly published. There is deliberately no allow-all value — a city cannot become public as a side effect of a missing environment variable, and an unlisted slug is refused rather than served.
Map availability
OPEN · DEMO + TERRAINOPEN
The console runs on this site. Two surfaces need no provisioning: a demo workspace on synthetic parcels, and terrain screening anywhere on Earth. Cadastral scoring against a real city is the part that requires an ingest and a publication decision.
Scenario saving
CLOSED WITHOUT A WORKSPACEGATED
Weight profiles and shortlists persist against an organization. With no session there is nothing to persist against, so the console keeps scenario state in the session only — the weights survive a pan, not a reload.
Score language
0-100, WEIGHT-QUALIFIEDOPEN
A parcel reads as a plain figure — score 81 — and that figure is only meaningful under a stated vector. Every export carries the weights that produced it, so a score can never travel without its own terms.
Coverage
PER INGESTED CITYGATED
Cadastral, zoning, and hazard scoring exist where a city has been ingested. Outside that extent the engine falls back to the terrain-only model, which needs no local data at all.

// CUSTOM SCENARIO REQUEST

Bring us the constraint, not the parcel.

Partners specify the deployment intent — enclosure, hazard tolerance, footprint, crew access — and Andromeda returns a ranked shortlist for a named city, with the weight vector and the excluded parcels attached. Custom weight profiles are published against the city and persist for the engagement.

Open the Console Request a Shortlist

// Integration

0708

Where the shortlist goes.

Andromeda is the spatial foundation the rest of the platform stands on. It selects the ground; the other lines instrument it.

// AND-00 · SOURCE

Scored candidate sites

A ranked shortlist, each parcel carrying its score, its weight vector, and the hazard exclusions applied to produce it.

SBX-02

Physical AI Sandboxes

Candidate parcels enter sandbox siting already screened for hazard exposure and buildable envelope, so enclosure design starts against real ground rather than an assumption about it.

FLD-01

Neicrone Field Ops

Staging areas are chosen on crew access and surface state, and the shortlist doubles as the site brief the run sheet is written against.

LOG-06

Active Logistics Deployments

Yard and lane integration planning inherits the same scored surface, so a deployment footprint can be tested against zoning capacity before a partner is committed to it.

// RETURN PATH · SPATIAL TIMELINE

Telemetry lands on the parcel that was scored.

Sites keep their identifiers, so runs, incidents, and evidence written by Field Ops attach to the same record the appraisal produced. Over an engagement the score stops being a prediction and starts being something the field can answer back to.

// Technical Note

0808

The pipeline, without adjectives.

For the reader assessing whether Andromeda can be pointed at their own cadastre.

// SCORING

POSTGIS

Criteria are normalised and combined in SQL against the parcel geometry, so a weight change is a query, not a rebuild. Hazard intersection and transit catchment are spatial joins, computed where the geometry already lives.

// DELIVERY

MVT

Scored parcels are served as Mapbox Vector Tiles cut by ST_AsMVT, at 4096 units of extent with a 64-unit buffer so polygons straddling a seam are not clipped into gaps.

// BASE MAP

OPENSTREETMAP + CARTO

Keyless public style endpoints — positron, dark matter, and voyager — overridable to a self-hosted or commercial style without touching the application.

// TERRAIN

TERRARIUM · MAPZEN / AWS

Open elevation tiles, near-global and keyless. They drive hillshade and 3D extrusion, and they are what makes terrain-only screening possible where no cadastre exists.

z10

TILE FLOOR

Below this a single tile would scan half a city. Requests under it are refused rather than served slowly.

z14

DETAIL THRESHOLD

Under it a tile can hold tens of thousands of parcels no one can tell apart. Those tiles aggregate to a 48-cell grid and read as a suitability heatmap instead.