This paper studies a different way to turn industrial scheduling and routing problems into quantum-friendly math. Instead of the usual quadratic form (QUBO, quadratic unconstrained binary optimization), the authors write problems as higher‑order unconstrained binary optimization (HUBO). HUBO can express more complex rules directly. That can make the quantum representation more compact and reduce the number of binary variables — and therefore the number of qubits — needed to represent a problem.

The team formulates three representative industrial cases as HUBO models: a shared‑transportation problem (QUEST, quantum‑enhanced shared transportation), the capacitated vehicle routing problem (CVRP), and a production or assembly‑line scheduling problem. They compare these HUBO encodings with equivalent QUBO encodings. In the two routing cases they report that HUBO can reduce qubit scaling from linear to logarithmic for a fixed problem, which means fewer qubits are needed as the instance grows. The production scheduling case keeps the higher‑order structure but is too large for direct quantum simulation, so they pair it with a hybrid quantum–classical rolling‑horizon workflow.

The paper also explores how these HUBO models would run on quantum hardware. HUBO needs higher‑order interaction terms, which map to deeper quantum circuits and many two‑qubit gates. Deeper circuits are harder to run on today’s noisy quantum processors because of limited gate fidelity and short coherence times. To study algorithmic behavior, the authors validated their formulations with classical solvers such as simulated annealing, and they benchmarked small routing instances using a quantum‑inspired method called bias‑field digitized counterdiabatic quantum optimization (BF‑DCQO) in classical simulation.

BF‑DCQO is a digitized counterdiabatic protocol that adds corrective terms to speed up adiabatic evolution and reduce the number of digital steps (Trotter steps) required. The “bias‑field” part is an iterative feedback loop: after each run the qubits are measured, the measured averages are turned into longitudinal bias fields, and those biases guide the next run toward lower‑energy (better) solutions. The counterdiabatic correction is built approximately with a nested‑commutator expansion so it stays practical to implement in simulation or on early hardware.