Software-defined vehicle building blocks are crucial to generative AI

DETROIT — Cars could someday look out for their drivers, warning of road hazards, suggesting available dinner reservations and playing favorite songs — think safety net, personal assistant and entertainer all in one.

The auto industry is on the brink of technology innovation based on generative artificial intelligence, Sripriya Raghunathan, vice president of automotive at Qualcomm Technologies, said at the SAE World Congress event here Wednesday.

"Automotive and technology are rapidly merging, bringing a transformation that industry has not seen over a century," she said.

For a car to predict a driver's coffee order or warn of an incoming blizzard, the automotive development process must prioritize in-vehicle software earlier and plan for virtual updates over the life of the product, Raghunathan said.

Generative AI is rapidly changing, she said. "Every day, there's a new model coming," while vehicles that can't receive updates will become obsolete.

Software-defined technology will underpin AI in vehicles. According to Raghunathan, some of the building blocks that will enable a generative AI future for the industry are:

Decoupling hardware and software: Upgrades are limited once a vehicle leaves the factory today because hardware and software are coupled, she said. Most vehicles are connected, which allows for over-the-air updates, but only for certain technology, such as the infotainment system.

"Consumers are not able to enjoy the benefits of [many] new features that are available in the market after the purchase of the vehicle, so the vehicle soon becomes obsolete," Raghunathan said.

Uncovering problems late in the development cycle is expensive, she added. An earlier focus on software allows automakers and suppliers to resolve issues sooner and to continue enhancing the vehicle over time.

Unifying in-vehicle architecture: Automakers should optimize in-vehicle architecture to make technology processes more seamless, she said. Vehicles need unified architecture to efficiently run advanced driver-assistance systems, generative AI features and digital display applications.

A vehicle should be divided into four zones — digitalization and personalization, driver monitoring and safety, automated driving and maps, and cybersecurity and networking — each of which communicates with its engine control module. A central computer processor oversees the zones and notifies the functions in the car, she said, "enabling greater efficiency, scalability and flexibility."

Establishing industry standards: Suppliers and automakers should work together to establish standards and tools for broad software development, Raghunathan said. The industry has many opportunities to collaborate, especially as automakers integrate third-party programs, such as payment systems, into the vehicle.

Combining auto and tech industry talent: As software changes the auto industry, engineers must learn new ways of software development while also working through legacy challenges. Automakers and suppliers alike should add Silicon Valley-level software developers to their existing workforce.