Updated 4 pm ET, 27 Sep 2018
Ford Motor Company has quietly signed a US $100,000 contract with NASA’s Quantum Artificial Intelligence Laboratory (QuAIL) to use the space agency’s quantum computer in its autonomous car research, according to a Space Act Agreement obtained by IEEE Spectrum.
The contract, which was signed in July by Ford’s chief technology officer, Ken Washington, will kick off a year-long effort to use QuAIL’s D-Wave 2000Q quantum annealer to address optimization problems of interest to the motor company.
Quantum annealers are aimed at solving a range of optimization and machine-learning problems, in theory very much faster than traditional digital computers. Quantum computers encode information in qubits, enabling massively parallel computation relying on purely quantum effects. Quantum annealing uses quantum tunneling and interference to deliver the most efficient solution—the global minimum—to a problem.
Joydip Ghosh, Ford’s technical specialist for quantum computing research, told Spectrum that the company would initially be working on a generalization of the classic traveling salesman problem—how to plot the most efficient route around a territory consisting of multiple cities. “Route management for fleet vehicles is a problem that we face in a real-world scenario,” he says, referring to Ford’s Chariot microtransit service. “If you try to solve this problem with a computer that we have today, there are so many options that you can easily run out of time. We think that quantum computing could be an alternative computing platform.”
“One of the things we’re hearing from our customers as we’re deploying some early fleets in cities [is that] they’re not being deployed optimally,” adds Washington. “That’s a real problem we need to have an answer to. Ultimately, we’ll bring autonomous vehicles and ride services to those cities in a smart way that actually makes the experience in the cities better.”
The agreement calls for the company to provide NASA scientists with two or three optimization cases to map into Quadratic Unconstrained Binary Optimization (QUBO), the form of input accepted by its $15 million D-Wave annealer. NASA will then provide feedback, train a Ford researcher in the use of its computer, and provide regular access to it.
Ford is not the first auto company to consider quantum computing. In 2017, Volkswagen used a quantum annealer to optimize routes for 10,000 taxis in notoriously traffic-clogged Beijing. The researchers concluded that quantum annealing would work for time-critical tasks like traffic optimization. (Quantum annealers deliver results in a matter of milliseconds).
Principal scientist Florian Neukart says that Volkswagen is now using quantum computing to improve reinforcement learning techniques for software agents to learn about interacting with their environment, for example in automated parking. “The goal is to show that we can augment artificial intelligence techniques with quantum computers,” he tells Spectrum. “We came up with a formulation allowing us to evaluate multiple configurations of a neural network in one annealing cycle.”
Volkswagen even believes that quantum computing could help simulate molecules to develop new batteries, which remain the biggest cost driver for today’s electric vehicles. “These are not new battery materials yet, but our intention is to show that quantum computers are useful for this field of applications,” he says.
Ford is not quite as far along in its quantum journey, which began in 2016 with Washington’s Research and Advanced Engineering team. “Quantum computing was there on our radar screen, and we made a commitment to start getting smart about it, and the best way to get smart is to bring in some talent,” says Washington. Ford hired Ghosh from the University of Wisconsin-Madison earlier this year, and signed its NASA contract in July.
“We thought partnering with NASA was a way to quickly come [up] to speed with knowing how to frame a problem in the quantum space, that did not require us to make a substantial capital investment,” says Washington. “For us, it’s not about having the hardware available, it’s about how to solve a problem.”
Daniel Lidar is the director of Center for Quantum Information Science and Technology at the University of Southern California. “There’s no question that quantum annealers can solve traveling salesman problems,” he tells Spectrum. “That’s been known for quite a while now. The question is whether they can do so better than you can do on traditional technology. It’s a real race between continuously improving classical technology and likewise improving quantum technology. Companies are pretty savvy about the fact that you can’t expect, even within the next couple of years, to be able to extract a quantum advantage from these machines.”
Although Ford will be using the annealer for autonomous vehicles research, its quantum computing effort is not part of Ford Autonomous Vehicles LLC (FAV), a new business that Ford formed in July to encompass most of its self-driving research, engineering, and operations, including its ownership stake in Argo.AI. The company intends to invest $4 billion into FAV over the next five years.
“Quantum is too far out to roll into that business yet,” says Washington. “For us, quantum computing is one of many things we’re doing to imagine and prepare for what might be around the corner, so that we can disrupt ourselves as opposed having others disrupt us.”
Washington would not say whether Ford would continue the NASA contract beyond its one-year term, but says that the company is now in quantum computing “for the long haul.”
Editor’s note: Following publication, and without disputing the accuracy of the interview or the NASA contract, Ford clarified that its NASA quantum computing project is not specific to mobility or autonomous vehicles. The first application to be addressed is likely to be route optimization for diesel commercial truck fleets.