A 25-Qubit Test Puts Quantum Routing on the Logistics Map
Quentir Defense Monitor
Evidence-based insights for quantum defense and security. Published by Quentir Systems LLC · August 6, 2026.
Vehicle routing becomes harder when trucks gain a reason to travel together. A dispatcher must still choose who goes where and when, while also identifying road segments where vehicles can safely form a close convoy. Each useful pairing adds another relationship to the calculation. A new preprint turns that coupled problem into a small quantum-hardware test and, more importantly for a defense buyer, shows where the quantum piece fits inside a much larger logistics system.
Talha Azfar and Ruimin Ke's primary paper combines roadside video analysis, traffic classification, route optimization, and microscopic vehicle simulation. In a 24-hour simulation of Troy, New York, their cooperative routing model reduced modeled fleet tractive-energy demand by 18.5 percent against a non-cooperative baseline. A 25-active-qubit subproblem also ran on IBM hardware. The strongest reported shallow-circuit setting sampled feasible answers 38.6 percent of the time and the exact classical optimum 14.2 percent of the time.
Those numbers describe two different achievements. The energy reduction came from the simulated fleet architecture. The hardware run tested whether a small routing subproblem could survive real quantum noise. Keeping those layers separate is essential. The paper offers a concrete systems experiment, while its quantum component remains far from selecting routes for a working convoy.
The route becomes a quadratic problem
The familiar vehicle-routing problem assigns vehicles to stops while respecting constraints such as capacity and timing. Platooning introduces a reward when two vehicles share the same road segment at compatible times. That reward depends on a pair of decisions, so the authors express the problem as quadratic unconstrained binary optimization, or QUBO. Routes, timing, and platooning relationships become binary variables and quadratic terms in one cost function.
That form is a natural input for the Quantum Approximate Optimization Algorithm. IBM's QAOA documentation describes a hybrid loop: a quantum circuit samples candidate solutions, a classical optimizer adjusts circuit parameters, and the process repeats. The paper modifies the circuit into a linear chain to reduce expensive two-qubit operations. On the reported 25-qubit instances, this cut CNOT depth by 66.7 percent relative to the authors' dense QAOA circuit.
The surrounding pipeline is at least as important. Roadside units estimate vehicle motion from video and decide whether traffic on a segment is stable enough to reward close coordination. The optimizer then works on localized partitions rather than the entire road network. Selected routes return to the traffic model for evaluation. This is edge-assisted quantum optimization as a component, with classical perception and decomposition doing much of the system-level work.
The traffic result also needs the right reading. SUMO is a microscopic simulator: its official model description says it represents individual vehicles and car-following behavior. That makes it useful for testing interactions across a road network. It does not reproduce damaged roads, intermittent communications, deliberate interference, military vehicle mixtures, or the consequences of a late delivery. The 18.5 percent figure belongs to the specified Troy simulation and assumptions.
Quantum pillar: computing (optimization and logistics). Use posture: dual-use. Technology readiness: TRL 3 of 9. The optimization ran on real quantum hardware at small scale, while the fleet behavior and energy result came from synthetic traffic.
Why a logistics office should pay attention
The defense relevance begins with throughput and exposure. A force must move food and fuel. It must also deliver ammunition and repair parts alongside medical supplies, even when routes and schedules are under pressure. The Department of Defense has called contested logistics a problem in which an adversary tries to disrupt or defeat friendly sustainment. Better dispatching could help a force recompute assignments as road availability and demand change alongside vehicle status.
Platooning makes that optimization question tangible. The U.S. Army's public account of its leader-follower tactical vehicle work describes one crewed vehicle controlling several unmanned followers to improve force protection and sustainment throughput. The Army system and the research paper are separate developments. Together they show why a program office may care about routing multiple coordinated vehicles instead of solving isolated shortest paths.
The capability is dual-use. Commercial fleets could use coordinated routing to save energy and increase road efficiency. Military logisticians could seek the same basic benefits while adding mission priorities, protected mobility, uncertain infrastructure, and communications constraints. The published optimization does not encode an adversary, concealment, threat avoidance, or rules for operating armed vehicles. Its defense value is therefore an architectural clue: localized sensing can update a combinatorial dispatch problem, and a specialized solver can sit behind that decision layer.
There is also a defensive advantage in reducing the number of people required inside a convoy, as the Army's leader-follower rationale makes clear. Routing software could support that objective by keeping vehicle groups synchronized and reallocating loads. Yet a smaller crew does not make a route resilient by itself. A buyer would need to measure how the optimizer behaves when sensor feeds are stale, links disappear, map data is wrong, vehicles drop out, and several constraints change at once.
The offensive reading is indirect. Faster logistics can increase operational tempo for any force that owns the capability. The paper supplies no evidence about military employment or about performance against active disruption. Its result supports the dual-use label because the same routing machinery can serve civilian transport and defense sustainment, with sharply different assurance demands around it.
What the hardware result establishes
The quantum run matters because it moves one piece beyond a software-only proposal. The authors mapped localized routing instances to 25 active qubits, executed shallow circuits on a physical processor, and reported both circuit depth and answer quality. At depth parameter two, a sampled answer was feasible in 38.6 percent of shots. The exact optimum appeared in 14.2 percent. Deeper settings took longer without improving the observed solution quality.
That is evidence of small-scale hardware execution, not evidence of quantum advantage. The paper compares its linear-chain circuit with a denser QAOA construction and checks sampled results against an exact classical optimum for the small instances. It does not demonstrate that the quantum route is faster, cheaper, more accurate, or more scalable than a strong classical logistics solver. In a procurement context, the classical baseline must remain in the test harness.
The architecture also carries a decomposition cost. The road network is split into communities so that subproblems fit the available hardware. Local solutions can miss a better global combination. The authors identify both a partitioning gap and hardware noise. They also flag limited qubit connectivity for further work. Those constraints matter because real dispatch problems grow with vehicle count and time windows. Cargo rules interact with route alternatives to add more complexity.
A useful next experiment would preserve the paper's clean separation of layers while replacing benign assumptions one at a time. Recorded traffic from mixed military vehicles would test the perception and energy models. Communications outages would test whether edge nodes can continue safely. Route closures and changing delivery priorities would test replanning. A classical solver should face the same instances as the quantum workflow and a hybrid fallback, under identical deadlines and compute budgets.
The buyer's evidence ladder
For a program office, this paper is a reason to watch the interface between quantum routing and autonomous logistics, rather than a reason to specify a quantum processor. The immediate reusable asset is the problem formulation: pairwise platooning rewards enter directly as quadratic terms, and localized traffic estimates decide when those rewards apply. That model could be benchmarked on classical hardware today while quantum methods mature.
Three thresholds stand between this result and acquisition relevance. First is scale: a useful system must handle operational time windows across the vehicles and destinations in a real distribution plan. Second is realism: energy and safety must hold alongside delivery performance. Tests must use representative roads and payloads across varied weather with disrupted data. Third is assurance: the dispatcher must return valid decisions on time and reveal when confidence is low. It must also fall back safely when any compute service is unavailable.
The published work clears an early rung because real quantum hardware produced solutions to small instances. It remains at TRL 3 because the closed-loop fleet result used synthetic traffic and the paper reports no operational vehicle trial. The defense-specific opportunity is resilient optimization for sustainment. The present evidence supports continued comparative testing, with mission value measured in delivery completion and energy use alongside replanning time and dependable fallback behavior.
Sources
Primary source: Talha Azfar and Ruimin Ke; other material from IBM Quantum, Eclipse SUMO, the U.S. Department of Defense, and the U.S. Army.