Case study

Improving Army logistics with quantum computing

Client

The challenge

At scale, optimization problems can’t be solved efficiently or accurately using classical methods. Quantum solutions have the potential to help, but hardware noise gets in the way of insights that customers are seeking.

Impact

12X

improvement in the likelihood of finding an optimal solution with Fire Opal over the default hardware execution

The outcome

Fire Opal improved the performance of quantum computers to a level that the early results could finally give the Army confidence that quantum route optimization could be a feasible way to improve convoy logistics, allowing them to build a roadmap toward implementing the solution at scale.

Research

–

"Infantry wins battles, logistics wins wars.”‍

Those are the famous words of General John J. Pershing, commander of the American Expeditionary Forces in Europe during World War I.

Heeding those sentiments, the Australian Army is strategically investing in technological innovation to find better solutions to the complex logistics challenges they face in managing the efficient and safe deployment of personnel and equipment on the battlefield. For a difficult class of problems in an area called “optimization”, quantum computing is on the roadmap for exploration.

With the help of our quantum infrastructure software, they’ve now been able to test and validate a quantum computing solution on real hardware that promises to outperform their existing methods.

Logistics challenges demand better solutions

Problems such as vehicle routing and deployment scheduling are fundamentally challenging mathematical problems, which can require extreme computational resources.

These are important examples of a class of problems known as “optimization problems.” Put simply, these involve finding the best schedule or path among many slightly different choices, in a way that maximizes some quantity, like speed or cost savings.

While it may sound surprising, determining how and when to deploy 120 convoys across just a few route options can take a month of computing time. Compared to a typical urban transit network or logistics company managing thousands of vehicles, it's easy to envision that this is a daunting problem at scale.

The complexity compounds in real-world scenarios where decisions must be made in real-time, and the objectives and risks are constantly changing. Due to the impractical time and compute resources required to solve such problems optimally (i.e. to obtain the right answer), most solutions typically require the use of a heuristic algorithm. These are approximate solutions that rely on many assumptions; the quality of the output is never perfect and only as good as the assumptions, which increases risk exposure and generally leaves a lot of room to do better with alternate strategies.

In search of better and faster solutions to their logistics challenges, the Australian Army turned to Q-CTRL with the challenge of validating their thesis that quantum computing has the potential to help.

Army presented a real-world problem: optimizing the departure schedules and routes of 5,000 convoys as part of Exercise Talisman Sabre—the largest combined training activity between the Australian Defence Force and the United States military.

The Army doesn’t just want to get their convoys to the destination—they need to minimize the total deployment duration while maintaining a precise ordering of convoy arrivals. This needs to happen in spite of the fact that the convoys vary in size and speed; some are light and fast, others large and slow. Those simple asks, called “constraints,” constitute a major complication that tends to break existing solutions in the world of mathematical optimization.

This problem is broadly indicative of a wide range of logistics challenges that the Army faces on a daily basis.

While current classical solutions can meet existing needs, they are still less than optimal, and the Army foresees that quantum computing can provide a strategic edge. They are investing in quantum today to gain foresight into the potential advantages and to understand the timeline and technological advancements required to achieve key milestones.

Figure 1. The routing options of the Army’s convoy deployment problem. Dynamic factors, such as fluctuating route congestion, make the problem more intractable.

Turning to Q-CTRL to achieve practical outcomes

The Australian Army had been exploring next-generation quantum computing solutions to these logistics problems but needed convincing evidence that quantum provided a path forward. They recognized that today’s machines are so prone to noise and error that hardware failures overwhelm the results end-users seek. Plus, limited hardware size put a major constraint on the types of problems they could solve.

The Army turned to Q-CTRL to deliver a promising solution to their convoy scheduling problem and to help them understand the path and timeline to achieving useful solutions via quantum technology that perform better than current classical methods.

Q-CTRL’s products made it easy to test prototype quantum optimization solutions and achieve meaningful results despite today’s medium-scale and error-prone hardware.

Making quantum computing useful for end-users through quantum infrastructure software is Q-CTRL’s specialty. The team focused on coming up with the best utilization of the hardware available to them to tackle the Army’s challenging constrained-optimization problem, abstracting this for simplicity of implementation, and pushing the quantum hardware to its performance limits.

Q-CTRL’s optimization solution addressed all vehicles by combining the strength of both classical and quantum computers—this is commonly referred to as a hybrid quantum-classical algorithm. The 5,000 vehicles were divided into convoys composed of groups of ten. This approach allowed them to do more with the small-scale hardware available today. But they pushed further, implementing a new hardware-optimized method for the quantum circuit that uses even fewer resources than traditionally required. These strategies allowed the Q-CTRL team to “stretch” the limited quantum resources available today.

Then, the team focused on improving execution. First, they tuned up the classical part of the algorithm so it ran most efficiently. While this seems straightforward, it’s a challenging machine learning task to implement this part of the algorithm such that it runs quickly and can consistently deliver useful outcomes (for the experts, this is the selection and tuning of the classical loop).

Next, they dealt with the real challenge—hardware errors. While quantum computers struggle with the fundamental challenge of calculation-corrupting errors, our Fire Opal product can get maximum performance out of quantum hardware by preventing and reducing errors during circuit execution. That’s how Fire Opal has demonstrated such extraordinary results on real quantum computers, improving solution quality and accuracy by a factor of thousands. Better quantum hardware performance enables larger problems to be solved more accurately and faster, which makes quantum solutions more competitive against classical methods.

By developing hardware-optimized methods for both quantum and classical execution, Q-CTRL made it possible to solve complex, high-value problems on today's noisy, medium-scale devices.

Making quantum competitive with classical solutions using quantum infrastructure software

Q-CTRL delivered a solution to optimize the entire 5,000-vehicle convoy and convincingly demonstrated the promise of quantum computing to provide advantages over existing classical methods.

The ultimate solution that the team delivered not only ran successfully on real quantum hardware but also delivered superior results to a benchmark classical heuristic solver. The new solution implemented through Fire Opal reduced the total deployment duration by more than two hours compared to the benchmark solver. This ten-percent improvement meant that all the convoys were able to reach their destination before midnight, as opposed to staying on the road past one o’clock in the morning.

Figure 2. The final deployment schedule generated by Q-CTRL’s hybrid quantum-classical solution. The Q-CTRL solution routed the convoys such that the deployment order was preserved, and the last convoy arrived at the destination 19 hours and 17 minutes after the first convoy departed. The Q-CTRL solution deployment time was roughly 2 hours shorter than the benchmark classical heuristic solver. Note that the solution time in this case is the duration of the optimized deployment schedule, rather than the computational runtime of the algorithm.

In achieving these improvements, every part of the solution mattered, but error suppression delivered by Fire Opal was essential to improving the quantum portion of the hybrid algorithm; adding error suppression effectively expanded the amount of value that could be demonstrated on hardware.

In tests on IBM hardware, Fire Opal improved the execution on the device enough to increase the probability of finding an optimal solution by 12X. Beyond improving solution quality, reducing errors also doubled the number of convoys that could be optimized simultaneously and improved the time-to-solution by six times compared to execution without Fire Opal.

Figure 3. Fire Opal prevents and reduces errors on quantum hardware in order to deliver better outputs. In our routing optimization work, error suppression increases the likelihood of finding a correct solution by 12X compared with default hardware execution. Fewer errors also enable bigger problems (more convoys) to be solved on the same hardware and reduce the amount of time required to reach a solution (less averaging is required to achieve a useful result). This solution was executed on 16-qubit IBM Guadalupe, which posed the primary limitation on the number of convoys that could be accommodated. Today’s larger machines—like the 127-qubit IBM devices supported by Fire Opal—can scale this demonstration even further.

This particular problem was run at a scale where “brute force” calculation could still be practically executed (quantum advantage was not achieved), but the results make clear that quantum solutions can be of higher quality than existing benchmark tools. Of course, it’s also possible in principle to deliver improved classical software—but after decades of demand, classical approaches haven't been able to scale to the needs of these optimization problems.

The results of this exercise provided the Army with convincing evidence of the potential advantage of quantum computing and of the value that Q-CTRL’s quantum infrastructure software can bring in meaningfully shortening the time horizon to quantum advantage for practical use cases.

Moreover, Q-CTRL was able to provide the Army with insight into the technological advances required for them to leverage quantum solutions in the future. As larger quantum computers become available, the Army is becoming strategically equipped to leverage the quantum solutions they are validating today, which promise to yield significant efficiency improvements in the future.

Optimally routing 120 convoys can take more than a month of classical computation. The Australian Army is evaluating the potential of quantum computing to provide improvements; however, it’s been difficult to validate the feasibility of a quantum solution due to hardware noise. With Fire Opal, an algorithmic enhancement software, we are able to achieve results on quantum computers that build confidence in our quantum roadmap. LTCOL Marcus Doherty, SO1 Quantum Technologies, Australian Army

Leveraging hardware-aware application solvers

In this case, the team exploring quantum applications had deep expertise in the field, but in general, this shouldn’t be assumed. In fact, the opposite must be true in order for quantum computing to become broadly adopted. The capabilities delivered by Fire Opal must be usable by anyone in the Army tasked with logistics management (rather than quantum computing).

Fire Opal’s abstracted application solvers make it easy for any organization to build and validate logistics solutions on quantum hardware. The Quantum Approximate Optimization Algorithm (QAOA) Solver within Fire Opal enabled the Army to define their logistics problems in simple modeling terms and then solve them using hybrid quantum-classical computation—all without ever seeing a quantum circuit.

Everything, from the problem setup to execution, is done behind the scenes and tuned for execution on real quantum hardware. Just one command will deliver performance near the maximum limits set by nature.

And it really works. Across a range of benchmark tests, Fire Opal’s QAOA Solver leveraged automated error suppression and well-configured algorithmic modules to reduce compute costs by up to 2500x by reaching the correct answer in fewer iterations.

Figure 4. Fire Opal's hardware-aware QAOA solver accepts input in simple modeling formats and performs the entire quantum-classical problem setup and execution so that it can be used without requiring any quantum expertise. The combination of error suppression and noise-aware problem setup achieves superior performance when executing on real hardware compared to other out-of-the-box QAOA solutions.

Preparing for future deployments

Quantum computers hold significant promise for optimizing logistics, reducing risk, and enhancing combat capabilities. The road to reaping these benefits involves early use case design and solution validation, as quantum computing is a rapidly evolving field. As quantum capabilities progress, the Australian Army's investment in developing solutions, with the support of Q-CTRL’s infrastructure software, positions them to harness the power of quantum computing for logistics as soon as possible.

And because logistics wins wars, early adoption of quantum computing can deliver a strategic edge now and in the future.

Learn more about how the Army is leveraging quantum computing to transform operations with the help of Q-CTRL in our webinar.