Doctrine · Solver
The weights are data
How many plans a pass authors, and which profile calculates each. Tuning a site should not need a compiler.
The slate
The slate is what this page sets: how many candidate plans a pass authors, and which profile calculates each. It is capped at three, on purpose. More candidates is not more insight, and a fourth option is one more thing to read under time pressure.
Two planning values live here as well:
| Setting | What it does |
|---|---|
| Replan every | The cadence. Zero means the planner runs on demand only. |
| Interceptor cruise | Set it above what the airframe actually does and every intercept point is one it cannot reach. |
Planning profiles
A profile is a named, versioned set of weights.
| Knob | What it decides |
|---|---|
| Objective | What the plan is trying to do: protect, minimise leakers, or balance. |
| Weights | Protection, economy, standoff and coverage, relative to each other. |
| Leakage weight | How much an expected leaker costs the plan. The first number most people reach for. |
| Effectors per target | How many defenders may be committed to one threat. |
| Commit below | The time to impact at which a defender is actually committed. |
Three profiles ship. PROTECT maximises coverage of protected assets and is the default. ATTRIT maximises kills and spends the magazine to do it.
Shipped profiles copy, they do not edit
A shipped profile can be copied and the copy tuned freely, but the original cannot be changed. An operator who tunes PROTECT itself and then cannot get back to stock has lost the reference point the other profiles are read against.
A copy records what it started from, because this began as PROTECT is the first thing anyone asks about a tuned profile.
What the planner weighs
Every pass re-solves the engagement and ranks the results. It does not only ask can we hit everything.
- Predicted leakers. How many threats are expected to get through.
- Cost. So a costly interceptor is not spent on a cheap drone.
- Coverage. How much of what you are protecting stays defended.
The result is that there is always a feasible, ranked, cost-aware option in front of you. The human decision is approve or adjust, never work it out from scratch under fire.
A plan is not authority
A profile decides how a plan is calculated. Whether that plan may run is decided somewhere else entirely, by the posture in force and the gate on each effect it proposes. Tuning a profile can never widen what the plan is allowed to do.
Where to go next
- Decision support for the gates a plan still has to clear.
- Constants for the values underneath.