Controlled-environment agriculture research space with hydroponic crops

Research that improves controlled CEA project decisions.

Crop Urbanis conducts applied research alongside product development to make project information more structured, traceable, comparable and usable across the lifecycle of vertical farms and hydroponic greenhouses.

Discuss a research collaboration Explore the Platform

The research question

Research principles

The objective is not to automate professional responsibility. It is to make project information and decisions more reliable, source-aware and reviewable.

Traceability before fluency

Determinism for critical calculations

Human review for consequential decisions

Evaluation before expansion

What Crop Urbanis may publish

Crop Urbanis may publish methods, data schemas, evaluation protocols, anonymised findings, limitations, model comparisons, architecture notes and non-confidential research updates.

Client datasets

Supplier-confidential information

Private-beta organisation identities

Proprietary implementation details

Model-routing controls

Commercially sensitive outputs remain private unless publication is separately authorised.

Active research tracks

Applied research feeds directly into the Crop Urbanis Platform.

CEA project data model

How should fragmented project information be represented so requirements, assumptions, alternatives and decisions remain traceable?

Research class: Applied product R&D
Method: Project entities, evidence and source schema, requirements and constraints, ownership and review states, decision history, versioning and output provenance.
Public boundary: Operational product research in the internal MVP; organisation data remains separately governed.
Public evidence: Module and workflow status are described publicly; project records remain private unless separately publishable.
Product decision: Supports the Project Brief, Requirements, Decision Control and Handover modules.
Next validation: Selected organisational workflow evaluation and review of data-model coverage.

Supplier-document intelligence

How can heterogeneous supplier documents and equipment specifications become a comparable project record?

Research class: Applied product R&D
Method: Document classification, field extraction, terminology normalisation, unit mapping, source-location capture, missing-field detection, contradiction flags and comparison criteria.
Public boundary: Supplier rights, confidentiality and publication conditions are defined for each scope.
Public evidence: Product workflow is described publicly without exposing supplier-confidential material.
Product decision: Supports Supplier and Equipment Intelligence.
Next validation: Permissioned document-intelligence evaluation with a defined source and review scope.

Scenario reliability

How can AI-assisted project workflows preserve deterministic control of physical, production and financial calculations?

Research class: Applied product R&D
Method: Formula ownership, assumption versioning, sensitivity inputs, validation rules, scenario comparison, human approval and error or exception paths.
Public boundary: The Platform does not replace qualified engineering, agronomic or commercial judgement.
Public evidence: Deterministic workflow and review boundaries are public; project assumptions remain scope-specific.
Product decision: Supports the Scenario and Economics Engine.
Next validation: Review of defined scenario sets, formula ownership and accepted exception handling.

Agronomic knowledge retrieval

How can project teams retrieve relevant agronomic information without losing source context or professional boundaries?

Research class: Applied product R&D
Method: Scoped knowledge bases, retrieval evaluation, source-aware outputs, contradiction handling, multilingual terminology, expert correction and project-specific context.
Public boundary: AI supports retrieval and controlled drafting; qualified professionals review consequential use.
Public evidence: Methods and boundaries can be described without exposing private knowledge bases or project data.
Product decision: Supports AI-assisted evidence retrieval, controlled drafting and expert review.
Next validation: Retrieval benchmarking, source-context checks and qualified-review evaluation.

Human-reviewed AI

How should the system divide work between models, deterministic services and accountable professionals?

Research class: Applied product R&D
Method: Task classification, AI confidence and exception handling, deterministic checks, review states, correction capture, escalation, output-acceptance criteria and audit trail.
Public boundary: Qualified professionals retain responsibility for agronomy, engineering, safety, regulation and project acceptance.
Public evidence: The role division and review boundary are described publicly; individual review records remain private.
Product decision: Governs the Platform’s overall reliability model.
Next validation: Human-in-the-loop study and review of acceptance, correction and escalation paths.

Long-horizon exploration

Image, sensor and operating evidence is a later validation track. It does not define the current Platform or private-beta scope unless separately promoted to an active research track.

Roadmap

Image, sensor and operating evidence

Roadmap / scoped research only

Potential work includes crop-state imagery, environmental logs, anomaly detection, operating issue classification, reviewed KPI feedback and protocol-version comparison.

Crop-state imageryEnvironmental logsReviewed operating feedback

Research class: Later validation track
Method: A measurable evaluation scope, permitted data and qualified review must be defined before expansion.
Public evidence: No production deployment is implied.
Limitation: This track remains roadmap or scoped research unless separately operational.
Product decision: Keep distinct from the current core workflow until evaluated.
Next validation: Define data rights, evaluation criteria, review roles and publication conditions.

Discuss a research collaboration

Selected research organisations or technical partners may be considered for a scoped private-beta or research collaboration.

Research inside private beta

Access, data, publication, intellectual property and confidentiality are agreed separately; public contact does not create automatic participation or access.

Work with Crop Urbanis on applied CEA product research.

Discuss a research collaboration