Quantum-inspired tech for Logistics: next-gen optimisation

Plan better routes faster with the latest technology.

InfinityQ‘s quantum-inspired technology uses the power of quantum computing while avoiding the difficult scaling problems. It solves logistics problems faster than any software available today. And it does a better job, too.

It’s called TitanQ, and it outperformed Gurobi, a traditional optimisation product marketed with the tagline ‘The World’s Fastest Solver,’ by a long way.

In a test run optimising routes between 10.000 locations, Gurobi took more than five hours to produce its best solution. InfinityQ’s TitanQ, in contrast, needed just one minute. One. Minute. That’s an incredible time-saving.

Re-running a plan or checking an alternative scenario during strategy planning is suddenly no big deal. And time isn’t the only improvement.

TitanQ’s routes were 1.5% shorter. That may not sound like much, but it adds up. Say your trucks cover 1.000 miles a day. In that case, the shorter routes remove 5.475 miles from roads annually. That’s more than five days of driving, saving money – and the planet.

Sounds great, right? So why aren’t we using quantum-inspired technology already? And how does it work? It turns out the secret sauce is something called an Ising machine.

The ising machine

The term ‘Ising machine’ initially sparked whimsical images of beautifully decorated cakes in my mind. I still see them – even after discovering the technology is spelled with an ‘s’, not a ‘c’. But I digress.

An Ising machine is a dedicated piece of hardware needed to implement an Ising model. The Ising model itself isn’t new, Ernst Ising and Wilhelm Lenz laid the theoretical foundations in the 1920s, proposing efficient mathematics to solve hard problems. Yet practical implementation wasn’t possible — until the oscillator-based Ising machine, which is new.

Academics created the first proof-of-concept only three years ago. InfinityQ, excited about the technology and its commercial applications, built their own machine and moved it into an AWS data centre four months ago. TitanQ runs on this machine and has a generous edge over traditional solvers because the innovative hardware lets it use the efficient Ising model maths. Fig 1 illustrates the difference with conventional solvers.

Say we need to find the best routes to deliver parcels to customers. The traditional solver works like a funnel: all possible routing solutions go in at the top, and the solver assesses each in turn. The software saves the best option, indicated by the blue line, and presents it to the user at the end of the run.

The Ising model, on the other hand, works like a sieve. All possible solutions still go in at the top, but the solver assesses them all simultaneously. That’s why it’s so fast.

And because it assesses all options simultaneously, TitanQ always finds the best solution. Traditional solvers make compromises: limited computing power means a trade-off between the best solution and the time it takes to produce the best solution. Put simply, we can’t wait a week for perfect routes if we’re delivering parcels tomorrow.

Traditional solvers can use different methods: assessing each possible solution in turn is one, but certainly not the fastest. A greedy algorithm is a popular alternative: the solver takes a starting position and determines only the best next step, then moves into the new position and finds the best next step again, repeating the process until there are no more steps.

The result is not the best solution, but it’s not a bad approximation. Think of it like walking to a restaurant in a city you don’t know well. You get there okay, but you probably didn’t take the most efficient route. There’s space for improvement.

TitanQ doesn’t leave that space. That’s why its routes were 1.5% shorter.

It also doesn’t run out of memory and crash, as Gurobi did in a second test optimising routes between 100.000 locations. Mind you, Gurobi is a good product, and we shouldn’t be unfair. The problems are just the limitations of conventional solvers because it’s the best we can do with conventional hardware and software. It’s why operations researchers spend much time and effort specifying problems precisely and parsimoniously.

That process becomes easier and faster with the new quantum-inspired technology. Perhaps the sauce is not so secret. Yet building a solver like TitanQ takes rare talent and skill. Talent and skill InfinityQ has – and you can use.

Who’s it for

TitanQ can be a game-changer for high-volume logistics: getting that 1.5% improvement translates into savings and lower CO2. The solver also accelerates the network review process, making it easier to optimise for seasonality and capture those incremental efficiencies.

Another great application is for companies working with fixed service areas. Gas boiler engineers, white goods delivery and installation, and heavy/bulky delivery drivers often work in fixed areas because daily optimal planning is too hard. That causes inefficiencies.

Think of two neighbouring areas, each with an engineer operating at half capacity. Now imagine the efficiency gains when we optimise for the day: suddenly we can make do with just one engineer.

Smaller improvements, like better allocations near area boundaries, also help the bottom line. TitanQ is fast enough to plan daily and get those benefits.

Then there’s a case I’m curious about: superfast delivery platforms. The number of players has settled, and the pressure to become profitable is rising. One way to reduce cost is to have riders deliver to multiple addresses in one ride. Now, planning a few addresses isn’t necessarily complex, but speed is essential. Can quantum-inspired tech deliver that edge?

Who do you think will lead the way? Drop me a note if you have questions or suggestions.

---

For more about new developments in tech and AI, click here.