Inspiration

Pasqal challenged us to see whether a neutral-atom quantum computer could tackle vehicle routing, the problem parcel carriers solve every night, when every extra variable costs an atom.

What it does

It assigns delivery customers to capacity-limited vans by encoding the clustering as a slack-free QUBO on a Rydberg atom register, whose measured bitstrings become delivery plans.

How we built it

We derived a one-hot QUBO sized to the smallest feasible fleet, designed a provably valid two-rail "ladder" atom layout, emulated it in Pulser, and checked every reference against brute force and CPLEX.

Challenges we ran into

Atom couplings can't be zero or negative, so faithfully embedding the problem's interactions was hard, and on larger instances our registers lost most of the objective.

Accomplishments that we're proud of

With just 8 atoms per instance and a pulse we never tuned, our registers hit the optimal clustering about 40× more often than random sampling, with up to 96% of shots feasible.

What we learned

A good-looking embedding on paper doesn't guarantee good results on the device, so we learned to judge every design choice by what the emulated shots actually produce.

What's next for Capacity-Feasible CVRP Clustering on Neutral Atoms

We'd keep registers small and use them as a quantum subproblem solver inside a light classical search, so the approach scales to larger delivery fleets.