Pasqal Introductory Track

Inspiration

We wanted to understand whether neutral-atom quantum computing could help solve vehicle-routing problems—and what prevents promising mathematical models from working on an atomic device.

What we built

We built a pipeline that generates verified routing instances, encodes customer-to-vehicle assignments as QUBOs, maps them to atomic layouts with BLaDE, and evaluates simulated solutions against exact classical references. Our reduced assignment encoding cuts the number of atoms by roughly 44–71% compared with the assignment-and-capacity-slack baseline.

What we learned

A compact encoding and a good geometric fit are valuable, but they do not guarantee good solutions. Controlled simulations showed substantially higher feasibility with ideal interactions than with interactions produced by atom positions. Pulse-tuning gains also struggled to transfer to those layouts.

Challenges

Our main challenge was preserving the original problem through every translation. Soft capacity penalties can change which solution is preferred, while geometric constraints distort the intended interactions. By separating these effects and checking feasibility, optimality and routing quality independently, we identified where our approach succeeds—and where more faithful encoding is still needed.