Depot · Kansas City, MO
OPTIMIZATION, IN THE OPEN
Every route starts with a choice.
15 cities. 4 vehicles. A real quantum run, made visible.
Recorded experimentLoading evidence…
Against the proven model optimum
64 sequential trajectories
Projected straight-line kilometers
Reused across vehicle decisions
The delivery network CONTIGUOUS US
Measured customer assignments across the United States
ACTUAL RUN DATA
Illustrative city centers
Lines represent straight-line travel
Attempt 1 of 64Replay the recorded search
01 / 64
A better solution, one sample at a time
Best feasible clustering objective · lower is better
QuantumRandom
Invalid attempts never enter the incumbent. Both methods use the same capacity checks.
THE TAKEAWAY
A sample is a proposal.
The quantum register proposes customer subsets. Capacity checks accept or reject them; the next vehicle works with the customers still unassigned.
Exact CVRP reference—
Final route gap—
What this page does — and what the run means
Real data, replayed locally
Clustering attempts, rejection reasons, measured routing histograms, and intermediate atom excitation probabilities come from the Python emulator. Playback changes the displayed snapshot; it does not run a new solver in your browser.
Two different spaces
US cities are the delivery locations. The tiny atom register is a separate layout in micrometers that represents route-edge decisions. Its geometry is fitted to a QUBO and is not a miniature map of the United States.
An honest hybrid
Routing for up to three customers is marked classical/trivial. Four-customer routes use measured edge-encoding samples with emulator-only DMM controls. Larger clusters use a labeled classical size fallback. Full exact references are evaluation only.
Distance & uncertainty
Distances are straight-line equirectangular projections at 38° latitude, not driving directions. Shared finite shot banks and a small instance are not evidence of quantum advantage. Occupations are probabilities, not a definite vehicle route during evolution.