Self-Driving Cars: A Complex Photonic Supply Chain

We tend to think of autonomous vehicles largely in terms of their technical challenges. But in a talk at the Frontiers in Optics meeting in Washington, D.C., Ira Moskowitz, the director of advanced-manufacturing programs with the Innovation Institute at the Massachusetts Technology Collaborative, USA, focused on another, less-well-known side of the challenge.

In addition to its formidable technical requirements, explained Moskowtiz, the self-driving-car revolution will create new and highly complex supply chains and new educational and workforce needs. Together, he suggested, those developments will present “huge challenges.” And that’s especially true for the photonic integrated circuits (PICs) that will drive some of the new vehicles’ core functionalities.

Lidar PICs: Autonomous-vehicle linchpin

Moskowitz zeroed in on one particular PIC component that will constitute a linchpin of practical self-driving cars: lidar. PIC-based lidars are still under intensive prototyping and development, and not quite ready for prime time. But when perfected, they’ll be orders of magnitude smaller and cheaper than the clunky mechanical lidars familiar from early autonomous-vehicle experiments such as Google’s self-driving car. Getting to that level of integration and cost savings, Moskowitz maintained, will be essential for making autonomous vehicles practical.

“Until we have these kinds of integrated photonic devices,” he said, “we really can’t enable this autonomous-vehicle revolution.” Focusing on lidar also makes sense, Moskowitz added, because automobiles today, which bristle with cameras and sensors, already include most of the other photonic components that, in one way or another, will end up in autonomous vehicles. Practical, integrated lidar is the missing link.

Byzantine supply chain

So what does it take to build a sustainable manufacturing ecosystem for lidar PICs? Quite a bit, it turns out. For starters, Moskowitz sketched out a byzantine supply chain involving multiple overlapping stages stretching from design through wafer fabrication, testing, probing, assembly and packaging into subsystems and finished products for use by the autonomous-vehicle industry—which, he said will likely prove to be “the most challenging customer base you could ever imagine having to ship into.”

“This is a very complicated supply chain,” said Moskowtiz, “for some of the most complicated devices I’ve ever seen.” Muddying the waters still further, many of that complex chain’s moving parts—both technical elements such as design tools and intricate processes for aligning lasers on the chip, and business logic such as subcontracting relationships and factory support services—aren’t yet in place, and are being developed in some sense on the fly. “It’s a very complex situation,” he said.

Zero defect tolerance

Moskowitz also called attention to “the talent side,” and how the drive for autonomous vehicles is creating new requirements for workforce education and training. Automobile manufacturers, he observed, are ramping up to hire literally thousands of employees to support autonomous-vehicle development. “The problem is finding the people to go into those jobs,” Moskowtiz said. And that quandary could prove one of the biggest obstacles to sustainable PIC manufacturing for self-driving cars.

The reason, Moskowitz explained, are the new skill sets that serving this unusual industry segment will require. For safety reasons, auto companies purchasing lidar PICs will demand literally perfect devices, with zero defect tolerance—a mandate that will in turn mean learning new approaches to defect modification and more sophisticated analytics than the current workforce may be accustomed to. The car industry also expects 100 percent on-time delivery and ongoing, steady component-cost reductions, year in and year out.

All of this, Moskowitz says, will boost the overall education and specialization requirements for autonomous-vehicle workers across the board, from technicians (who may find that two-year degrees are no longer sufficient) to Ph.D.s. And the autonomous-vehicle revolution will also require “an entire army of I.T. and cybersecurity folks.”

“Robots on four wheels”

Ultimately, said Moskowitz, these complications—coupled with the need for photonic components to interact with other subsystems and trends such as artificial intelligence, smart cities and the Internet of Things—will call for a new production ecosystem. Such an ecosystem, he suggested, needs to look at autonomous vehicles as “robots on four wheels,” and to expand to domains well beyond the PIC fab.

He suggested that public agencies such as his organization, the Massachusetts Technology Collaborative, and public-private partnerships such as AIM Photonics could help the process of building such an ecosystem, given the inherently cross-cutting technology domain inhabited by autonomous vehicles. “It’s not simply photonics,” said Moskowitz. “It’s also software, automation, robotics, smart cities, mass transit and regulation.”