Capabilities, with the stage of each one

Queueing, contention, blocking, starvation and failure as consequences of the model rather than parameters of it. The default method for production lines, warehouses and terminal processes.

Manufacturing Operations Twin

Populations of individuals with their own attributes, decisions and routing. Used where the behaviour of the population is what decides the outcome, as with passengers, operators and vehicle fleets.

Airport Digital Twin

Subsystems sharing state so a fault propagates through the physics rather than along a script. Demonstrated across orbit, power, thermal, attitude and communications.

Satellite Mission Digital Twin

Runs are seeded and inputs versioned, so a run reproduces exactly. Results are reported as distributions across replications, because one run of a stochastic model is a sample rather than an answer.

Method

Repeated replication under sampled inputs to characterise the spread of an outcome rather than its central case. Delivered as part of an engagement; it is a way of running the models above, not a separate engine.

Reconstructed connectomes executed as running spiking networks, with circuits that can be driven, recorded and ablated. Research infrastructure rather than an industrial method.

NeuroSim

Objects, relationships, states, events, constraints and actions, in the terms the operation uses. Built with the people who run it, because a model no domain expert has checked is a model of an assumption.

The operational model

Fitted on one period of operational history and tested against another it never saw, with the deviation reported rather than summarised.

Validate

Plants and terminals are generated at runtime from layout and equipment data, so a second facility is a second dataset rather than a second build.

Airport Digital Twin

Alternatives run from an identical starting state and compared on measured outcomes. Demonstrated by instantiating complete additional plants; not yet a platform feature.

Scenario comparison

Files, exports and database extracts of demand, cycle times, downtime and movement, used to calibrate and then to validate.

Live read paths that keep a deployed twin current. Delivered as integration engineering per site; no shipped connector exists, and this entry will say otherwise only when one does.

Integrations

Unity 6 with the universal render pipeline, generating the environment at runtime from layout data rather than from an authored scene.

Public systems

Burst-compiled jobs over struct-of-arrays storage with spatial-grid neighbour avoidance, GPU instancing and three distance levels of detail. The crowd costs a handful of draw calls rather than ten thousand scene objects.

Airport Digital Twin

WebGL builds with an explicit performance budget per target, trimmed deliberately rather than degrading unpredictably.

Run the systems

TypeScript and React surfaces over model state, with control panels generated from a parameter registry so an interface cannot drift from what it controls.

Urual: a job state machine, a worker runtime, a plugin system, storage and artifacts, and a Postgres schema behind all of it. Driven entirely over HTTP.

Urual

The engine owns simulation state and every client is a viewer. Enforced across a process boundary in the mission twin, where the client computes no physics and says so when the engine stops.

Satellite Mission Digital Twin

A Go SDK that is a translation of the platform's HTTP API and nothing more: no simulation logic, no plugin knowledge, no state beyond a session.

The engine running entirely on one machine, with no checkout, no licensing and no outbound network call. This is the deployment model for operational data that may not leave a site.

Urual

Docker and Docker Compose for reproducible engine environments, in use across the platform and the Python simulation engines.

Tests written against what the model computes rather than against the interface. The mission twin's engine carries 198 of them.

Satellite Mission Digital Twin

The binding constraint derived analytically from stage capacity and corroborated independently by queue depth. This is analysis over model state, not a learned model, and calling it AI would be a category error.

Manufacturing Operations Twin

Protective behaviour that emerges from state crossing a threshold rather than from a timeline: load shedding, downlink inhibit and safe-mode entry in the mission twin. Deterministic by design.

Satellite Mission Digital Twin

Systematic exploration across a defined parameter space, distributed over the execution engine. Depends on the platform reaching the point where large run sets can be scheduled and compared.

Search over the space of operational actions using the simulation as the objective function. Under investigation. Nothing is deployed and no result is claimed.

Research

An engineer asking a question in the operation's own terms and getting an answer resolved against the model and the runs, with the assumptions attached. Under investigation, and it executes nothing.

Research

Alsadaany operates no trained model in any delivered system today. Terra's product direction depends on vision and sensor fusion, and that work is ahead of the company rather than behind it. It is listed here so its absence is explicit rather than inferred from silence.