Operational Disruption Delta
How much more does the best route cost after disruption evidence changes the network?
Δcost = Cdisrupted - Cclean
Δrelative = Δcost / Cclean
Δrelative = Δcost / Cclean
Rangeunbounded
UnitEUR / %
Executable course · paper profile v11 jul17
From disruption paper to verified aviation decision.
Learn how one system turns disruption evidence into warped route costs, keeps an exact classical answer as operational truth, compares a paper-profile quantum lane, and uses Depthmetric Aviation to block and repair invalid workflow candidates.
Operational authorityExact classical route
Quantum statusPaper-profile simulation
Evidence statusCurated demo feed
AMS · FRA · DXB · SINCONTROLLED WORKFLOW
Intentcargo mission
Evidence6 disruptions
Kernelwarped graph
Classicalexact truth
Quantumcomparison
Depthmetric Aviationverify + heal
Rule: agents coordinate; deterministic code computes, verifies, and decides.
01 · LANGUAGE
Two Deltas, two different questions.
The disruption paper and the self-healing control plane both use Delta. Combining them into one number would be a category error.
Depthmetric Aviation Control Delta
How far does a candidate violate weighted invariants for this workflow?
Intent → Candidate → P(I) → O(C)
Δcontrol ∈ [0,1] → decision
Δcontrol ∈ [0,1] → decision
| low risk | ε = 0.20 |
| medium risk | ε = 0.10 |
| high risk | ε = 0.03 |
| critical risk | ε = 0.00 |
Blocker rule: a missing airport or duplicated visit blocks the candidate even when the weighted scalar Delta looks small.
02 · PAPER MATH
Warp the graph before asking a solver.
The deterministic kernel samples each directed great-circle route at 20 points, measures disruption influence, and changes edge costs. Solvers never receive raw prose.
Directed graph
G = (V,E). Airport-to-airport costs may differ by direction, so cost(i,j) need not equal cost(j,i).
20-point exposure
Every route is sampled at exactly 20 equidistant great-circle positions using R = 6371 km.
Risk warping
Intersecting routes receive directional severity. Non-intersecting routes receive exponential distance decay.
Base route cost
Disruption severity
Midpoint distance
Route relation
Full closure
Great-circle distance
d = 2R asin(√(sin²(Δφ/2) + cos φ1 cos φ2 sin²(Δλ/2)))Intersecting influence
influence = directionalExposure × severityWarped directed cost
C'(i,j) = C(i,j) × (1 + Σ influencek)Base costEUR 1,000
Influence0.600
Warped costEUR 1,600
This calculator isolates one disruption and one edge for learning. The engine aggregates all typed disruptions across every selected directed edge.
03 · AGENTIC FLOW
The agent does not own mathematical truth.
Select each stage to see where probabilistic interpretation stops and deterministic control begins.
agent
Interpret operational intent
Input
Output
Invariant
Failure action
04 · SOLVERS
Classical truth, quantum evidence.
The hybrid architecture is not a race where the quantum lane automatically wins. Both lanes receive a comparable problem, and Depthmetric Aviation verifies their candidates under one contract.
| Dimension | Classical lane | Quantum lane |
|---|---|---|
| Role | Operational reference | Research comparison |
| Current implementation | Exact enumeration of required hub orders | Warm-start QAOA simulation |
| Default availability | Every run | Optional and policy-gated |
| Paper parameters | Same warped directed matrix | p=2, 4096 shots, CVaR 0.2, COBYLA max 15 |
| Eligibility | Tiny exact demo instances | At most 4 selected airports under default profile |
| Failure behavior | Ask human if no valid route | Retain verified classical result |
| Production claim | Executed locally | Simulation, not a real QPU |
H = Hroute + 10 Hconstraint + 2 Hrisk
λc = 10 strongly penalizes invalid tours. λr = 2 adds node risk. For mandatory-visit feasible tours, route cost determines their ranking.
05 · LIVE CONTROL LOOP
Watch one invalid quantum candidate heal.
This calls the same API as the aviation workbench. Noise intentionally removes a required hub; Depthmetric Aviation must block it, apply one classical warm-start fallback, and verify again.
Worked mission · pharmaceutical cold chain
Curated feed · exact classical execution · paper-profile quantum simulation
Predict the sequence, then run it.
block → heal once → allow
06 · CONTROL BOUNDARY
Who may do what?
Agentic does not mean autonomous authority. The design assigns interpretation to agents and truth-preserving operations to code.
Agent-owned
- Interpret cargo and operational objective
- Select approved evidence connectors
- Normalize source material into typed candidates
- Propose an eligible solver tool
- Explain verified observations
Code-owned
- Validate evidence and immutable profile
- Compute geometry, exposure, and cost warping
- Enforce solver eligibility and budgets
- Verify route feasibility and comparability
- Compute both Deltas and decide continuation
Agent must never
- Invent severity or decay coefficients
- Change Hamiltonian weights silently
- Declare its own route verified
- Treat raw QPU output as operational truth
- Retry until a preferred result appears
Every selected airport appearsMissing a required hub is a blocker.
No airport is duplicatedA tour must be a unique airport sequence.
Comparable configurationClassical and quantum lanes use the same warped problem.
Bounded healingAt most one automatic workflow repair in the initial product.
07 · KNOWLEDGE CHECK
Can you defend the architecture?
Choose an answer. The feedback explains the engineering reason, not just the label.
08 · CONTINUE
Move from understanding to operation.
Use the focused labs for deeper interaction with each half of the system.