Skip to main content

Applied research

Research for spatial agent systems.

We investigate what spatial agents need, then apply what we learn to the systems we build.

Spatial grounding

Agents need a working sense of space.

Spatial grounding connects an agent to what an organisation's places, data, rules, and software state actually mean.

The system must preserve that meaning as it chooses tools, runs programs, creates outputs, and checks the work. That is the spatial requirement we bring to the standard agent stack.

Current programme

Questions guiding the work.

These are active research questions, not claims that the problems have already been solved.

01Spatial grounding

How can an agent understand what the spatial data actually represents?

Coordinates are not enough. The system must preserve place, identity, time, source, uncertainty, and the state of the software doing the work.

02Models and orchestration

Which combination of models, programs, and tools works for the job?

We compare complete configurations because the model name alone does not determine how a workflow behaves.

03Trusted operation

How should the system respond when an output is wrong or uncertain?

A useful system must recover, ask for review, or stop when the work cannot be trusted.

Insights

Read the practical work.

Research notes and engineering guides for people building spatial agent systems.

Spatial agent systems

What geospatial agents need beyond the standard stack

Spatial grounding, context, tools, durable state, and checking.

Read this subject
Workflow engineering

Building and evaluating real spatial workflows

Practical work on orchestration, software use, migration, and operational limits.

Read this subject
Spatial infrastructure

Data, compute, and geospatial systems in the cloud

Technical guides for the environments where spatial work runs.

Read this subject

Axis Spatial Arena

Compare complete agent systems.

Arena shows which models, tools, prompts, runtimes, and checks were used, what happened, what failed, and what the result still cannot prove.

Spatial tasks connected to outputs, checks, and reviewable results.

Start here

Working on a spatial agent system?

Bring us the workflow, system, or research question you need to make concrete.