CONTEXT(CLONE)
Sign in

Roll up drone inspection footage across an asset network

For: Asset integrity manager for overhead lines, pipelines or telecom towers

Pattern: Hierarchical decompositionNeeds physical AI: plannedDesigned for 50 to 3,000 agents
Get early accessSign in to use

The pain today

Drone flights return footage for every span and tower. Engineers watch a fraction of it, and a defect on an unwatched structure is found when it fails.

The ask

“I attached the drone mission logs, the footage index by tower and span, and our defect catalogue. For each structure, tell me what visible defects the footage shows against the catalogue, with frame references, and which structures had no usable footage.”

Plain words, as you would say it to a colleague. Edit it to fit your case before you send it.

What you attach or connect

  • Drone mission logs
  • Footage indexed by structure
  • Defect catalogue with examples
  • Asset register

The unit of work

One worker task per one structure's footage segment.

Why a swarm fits

The network breaks into circuits, sections and structures. Each structure's footage is assessed alone against the catalogue, and each section is checked for coverage before it rolls up.

Not for

Defects that need measurement, thermal or internal inspection. Visible-light footage cannot show them.

The decision tree

6 typed decisions, each with an action for every answer

At fixed moments in a run, the engine puts one narrow question to a decision model. The decision model never writes text: it answers yes or no with a probability, picks from listed options, or gives a score, about a small slice of the material. The engine then does exactly what this tree says, which is what makes the run auditable. The thresholds are the template's design values, not measured results.

  1. Planner, while planning

    Split or keep serialYes or no, with a probability

    While planning

    Do the mission logs show this circuit's footage divided into sections that can be assessed without each other?

    Sees only: One circuit's mission log summary and footage index

    Why: Planned: splits follow how the footage was actually flown.

    • Yes: 0.60 or higherthenSplit the work
    • Unsure: 0.40 up to 0.60thenKeep serial
    • No: below 0.40thenKeep serial

  2. Split or keep serialYes or no, with a probability

    While planning

    Is this section's footage indexed by structure precisely enough that each tower or span can be its own task?

    Sees only: One section's footage index entries

    Why: Planned: a loose index would send workers footage of the wrong structure.

    • Yes: 0.60 or higherthenSplit the work
    • Unsure: 0.40 up to 0.60thenKeep serial
    • No: below 0.40thenKeep serial

  3. Scope checkA choice among options

    Before work starts on a unit

    Do the mission log's position and time entries agree with the footage index that this segment shows the named structure?

    Sees only: One footage index entry and the matching mission log lines

    Why: Planned: a structure with no confirmed footage is reported uninspected, never clear.

    • Log and index agreethenAccept
    • They disagreethenMark unresolved
    • No usable footage recordedthenSkip this unit

  4. After workers, the judge checks

    Evidence checkYes or no, with a probability

    After a worker answers

    Does the written description of the flagged frames meet the defect catalogue entry's own criteria for that defect type?

    Sees only: Planned design: one catalogue entry's text and the description of the flagged frames

    Why: Planned: a defect is reported against the catalogue's wording, not a model's impression.

    • Yes: 0.85 or higherthenAccept
    • Unsure: 0.50 up to 0.85thenEscalate to a strong model
    • No: below 0.50thenReject and retry

  5. Reconciler, while merging

    Conflict checkYes or no, with a probability

    While reconciling

    Do the two independent descriptions of the same frames name different catalogue defects, or does only one name a defect?

    Sees only: Planned design: two descriptions of one set of frames

    Why: Planned: frames the two families disagree on go to the inspector as disagreements.

    • Yes: 0.70 or higherthenMark unresolved
    • Unsure: 0.30 up to 0.70thenMark unresolved
    • No: below 0.30thenAccept

  6. Accountable person, before anything is settled

    Person decidesYes or no, with a probability

    Before anything is reported as settled

    Does this structure have a flagged defect, a disagreement between descriptions, or no confirmed footage?

    Sees only: One structure's findings with frame references

    Why: Planned: a qualified inspector confirms every defect and sets repair priority.

    Accountable: A qualified inspector confirms every flagged defect and sets the repair priority.

    • Yes: 0.20 or higherthenAsk a person
    • Unsure: 0.05 up to 0.20thenAsk a person
    • No: below 0.05thenAccept

The fleet: who does what

Model tiers by role, not brands: you choose the models. Strong reasoning models plan and reconcile, small fast models do the wide work, and the judge is a decision model from a different family, so it does not share the workers' blind spots.

  1. Planner

    A strong reasoning model builds the tree from network to structure and matches mission logs to footage segments.

    Decisions here:1. Split or keep serial2. Split or keep serial3. Scope check

  2. Workers

    A perception model for video and images, planned, one task per structure, reporting catalogue defects with frame references.

    Designed for 50 to 3,000 agents, one worker task per one structure's footage segment. Each worker receives only its own unit.

  3. Judge, from a different model family

    A second perception model from a different family, planned, rechecks each flagged frame against the catalogue entry.

    Decisions here:4. Evidence check

  4. Reconciler

    A strong reasoning model rolls up per section and lists structures with missing or unusable footage.

    Decisions here:5. Conflict check

  5. Accountable person

    A qualified inspector confirms every flagged defect and sets the repair priority.

    Decisions here:6. Person decides

Checked before anything is accepted

  • Every defect cites frame references and the catalogue entry it matches
  • Mission logs are used to confirm the footage shows the named structure
  • Structures with no usable footage are reported as uninspected, never as clear
  • Each section's coverage is verified before it joins the network view

What comes back

  • Per structure: catalogue defects seen, with frames
  • Structures not inspected and why
  • Section and circuit roll-ups of inspected versus uninspected
  • Frames the two perception models disagree on

What to measure

  • Share of flagged defects a qualified inspector confirms
  • Defects found in the field on structures reported clear
  • Share of captured footage that gets assessed
  • Cost per structure assessed

Names of measures only. No result is claimed for this template.

Templates open in the workspace chat with the ask filled in. Nothing runs until you send it.

Get early accessSign in to use

More in Physical AI and field operations

Check site photos against the construction programme

For: Construction project manager or owner's representative tracking progress across sites

Progress is reported as a percentage in a spreadsheet.

Pattern: Map, verify, reduceNeeds physical AI: planned5 decisionsDesigned for 30 to 800 agents

Review plant walkdown video from four specialist angles

For: Site HSE manager or operations manager at a process or manufacturing plant

Walkdowns are filmed but reviewed by one person with one checklist.

Pattern: Specialist panelNeeds physical AI: planned5 decisionsDesigned for 8 to 300 agents