Simulation
Runs of a computational model that imitate a system under stated conditions.
Bundle → Layer → Finding → Questions Filled
3 bundles · 3 layers · 5 findings · 10 questions
Setup What was simulated and how.
Model and inputs
The model version, parameters and initial conditions.
Model version
The model and software version used.
- Which model and which software version ran the simulation?
- Which assumptions and simplifications does that model make?
Inputs and scenario
Parameters, initial conditions and the scenario applied.
- Which parameters, initial conditions and scenario were used?
- Where do the input data come from, and how current are they?
Execution How the run was performed.
Run record
Time step, random seeds and computing environment.
Reproducibility
Seeds, time step and environment needed to reproduce the run.
- Which random seeds, time step and computing environment were used?
- Can the run be reproduced with the recorded setup?
Credibility Whether results can be trusted for the intended use.
Verification and validation
Evidence that the model is solved correctly and fits reality.
Validation evidence
Comparison of results with reference data.
- Against which reference data was the model validated, and with what error?
- Is the intended use within the validated range?
Uncertainty
Stated uncertainty and sensitivity of results.
- What uncertainty ranges are stated for the results?
- Which inputs are the results most sensitive to?
Classifiers Filled
- Family
- World Models
- Category
- Information and virtual systems
- Entry kind
- event
- Navigation path
- NAV.INF.VRT.SIM
- Domain
- INF.VRT.SIM
- Industry
- Cross-industry
- Tags
- simulationinf.vrt.sim
What it is Filled
A simulation is the execution of a computational model over time to imitate the behaviour of a real or hypothetical system under stated conditions, producing results that are used for analysis, training, design or decisions. The mathematical model itself, a reusable scenario definition and a live digital twin synchronized with a real asset are separate subjects.
In scope
- Run identity, intended use, model and scenario bindings, input snapshot and configuration.
- Recorded execution state, environment, coupling, restart lineage and simulated coordinate semantics.
- Output lineage, repeatability evidence, verification and validation references, uncertainty and scoped review.
- Run-specific access, retention, custody and controlled release metadata.
Out of scope
- Development or lifecycle of the mathematical model, reusable scenario, dataset, asset or digital twin.
- Solver implementation, scheduler orchestration, real-world actuation, autonomous certification or substantive decision authority.
- A universal physics, training, social or regulatory profile; operational details of dangerous applications.
Why it exists Filled
Runs of a computational model that imitate a system under stated conditions.
Distinguishing features Filled
- An execution of a model, so the same model can yield many simulations with different inputs.
- Its results are only as credible as its verification, validation and stated uncertainty.
- Not synchronized with a live asset, which separates it from a digital twin.
- Distinct from a scenario, which is the reusable definition of conditions a simulation may run.
What robots and AI may and may not do Filled
Must not
- Present simulated results as observed measurements.
- Use a model outside its validated range for safety decisions without saying so.
- Tune inputs to reach a predetermined answer without disclosing it.
- Discard runs that contradict a preferred outcome.
Only with a human decision
- Accepting simulation results as the basis for safety, certification or policy decisions.
- Approving a model as validated for a given use.
May
- Run simulations with recorded model versions, inputs and seeds.
- Report results together with their assumptions and uncertainty.
- Compare results across scenarios and flag sensitivity.
- Check whether a requested use lies within the validated range.
Moral aspects Filled
- Simulations inform decisions on safety, public policy and investment, and hidden assumptions can mislead.
- Overconfidence in precise-looking results can shift responsibility away from human judgement.
- Simulations of people or populations can encode bias from their inputs.
Who is affected
- People affected by decisions based on results
- Engineers and analysts running simulations
- Regulators and certifying bodies
Owners Filled
Steward
The analyst or engineering team that runs the simulation and answers for its setup and reporting.
Roles
- Simulation steward
- Maintains run identity, purpose, namespace and custody assignments.
- Simulation operator
- Records actual setup and runtime evidence without asserting unearned validation.
- Credibility reviewer
- Assesses bounded use, discrepancies and uncertainty under delegated authority.
- Data custodian
- Enforces access, license, retention and approved payload disposition.
- Result consumer
- Preserves origin labels and use limits when interpreting or deriving results.
Master systems
- Simulation data management system
- Model repository
- Experiment tracking system
Links to other meta-models Filled
references
- WM-KNW-019 - Candidate binding to the mathematical/computational model master; pin exact version in an operational profile. A model is semantically necessary but this particular registry binding is not yet required.
- WM-VRT-012 - Optional reusable scenario definition; keep the actual applied condition snapshot here and scenario lifecycle outside.
- WM-VRT-001 - Optional requesting or contextual twin; no asset synchronization ownership or required live connection is inferred.
- wm-vrt-012-simulation-scenario - A simulation runs one scenario definition.
aligned
- FMI 3.0.2 - Optional component-interface alignment subject to supported capabilities and tested mappings, not universal conformance.
- SED-ML Level 1 Version 4 - Optional experiment-description projection; it describes setup and tasks rather than proving actual execution.
- PROV-O - Map an actual run to Activity and its snapshots/results to Entities; local event taxonomy is not an OWL equivalence assertion.
neighbor
- WM-KNW-019 Mathematical / Computational Model - A model version is used by the occurrence; its equations, development lifecycle and reusable validation baseline remain externally mastered.
- WM-VRT-012 Simulation Scenario - Reference a reusable scenario when one exists and preserve the actual applied conditions. A separately registered scenario is optional; a run need not execute exactly one scenario.
- WM-VRT-001 Digital Twin - A run may be invoked by or use data from a twin. This record does not own asset synchronization. Lack of synchronization is not a defining prohibition on simulations.
- Experiment, ensemble and dataset masters - An experiment can group many occurrences; outputs and observation datasets retain separate identities and origin labels. This dossier references them without adopting their lifecycles.
- wm-vrt-001-digital-twin - A digital twin is synchronized with a live asset, while a simulation need not be.
requires
- wm-knw-019-mathematical-computational-model - A simulation executes a computational model.
What else AI and robots need to interact with it Filled
Identity and identifiers required Filled
- Identified by a run identifier in the simulation data management or experiment tracking system.
- Linked to the model version, scenario and input dataset identifiers it used.
Direct properties not applicable Not applicable
Not applicable
A simulation is a computational run; the quantities it reports are model outputs, not physical measurements of the simulation.
Recognition optional Filled
- Recognized by a model, inputs, a run configuration and results labelled as simulated.
- Often confused with the model, with measured data or with a digital twin.
Capabilities and actions required Filled
- Can explore conditions that are dangerous, costly or impossible to test in reality.
- Can be repeated with varied inputs for sensitivity and uncertainty analysis.
- Can train people and systems in virtual conditions.
Hazards and failure modes required Filled
- Wrong decisions from unvalidated models or unstated assumptions.
- Results taken out of the validated range.
- Irreproducible results when setup is not recorded.
Standards and interfaces required Filled
- Functional Mock-up Interface (FMI) for model exchange and co-simulation.
- IEEE 1516 High Level Architecture for distributed simulation.
- ASME V&V standards for verification and validation.
Context of use required Filled
- Used in engineering, climate and earth science, health economics, defence training and planning.
- Certification bodies may accept simulation evidence only under validated processes.
Sources Filled
- Functional Mock-up Interface Specification - Modelica Association
- Simulation Experiment Description Markup Language: Level 1 Version 4 - SED-ML Editorial Board
- NASA-STD-7009B: Standard for Models and Simulations - National Aeronautics and Space Administration
- A Summary of Industrial Verification, Validation, and Uncertainty Quantification Procedures in Computational Fluid Dynamics - National Institute of Standards and Technology
- PROV-O: The PROV Ontology - World Wide Web Consortium
- RFC 3339: Date and Time on the Internet: Timestamps - Internet Engineering Task Force
- Functional Mock-up Interface (FMI) standard, Modelica Association
- IEEE 1516 Standard for Modeling and Simulation High Level Architecture, IEEE
- ASME V&V 10 Standard for Verification and Validation in Computational Solid Mechanics, ASME
Open questions
- Restore independent external review and complete live source/version and claim verification before canonical promotion.
- Develop and test nested schemas with steady-state, failed, cancelled, restarted, stochastic and partially observed runs; include denied access, withheld inputs and lawful payload disposal.
- Review broader simulation profiles and any required HLA or ASME normative alignment before adopting their conformance claims or runtime requirements.
- Executable nested instance schemas, cross-field requirements, runtime adapters and conformance fixtures are not supplied.
- Full profiles for discrete-event, agent-based, social, training, real-time and hardware-coupled simulations require further evidence and domain review.
- HLA federation semantics and ASME conformance were supplement leads only; neither is admitted as a reviewed normative alignment.
- Independent external review and direct live HTTP/version verification remain open.
Machine files
Provenance
world-models research · reviewable-draft
Built from: models/wm-vrt-003-simulation/spec.yaml, ver-cy/world-models/card-supplements/wm-vrt-003-simulation.json