Interview: Volvo’s engineering lead discusses tech stacks

Volvo’s superset tech stack

The superset tech stack is a modular engineering platform that the company says will be used to build all future vehicles. It allows Volvo Cars to make safer cars, more efficiently, and to improve them over time through software.

“Effectively, the superset tech stack is a true superset strategy,” says Bakkenes. This means Volvo can build small “B” class cars like the EX30 right up to premium “F” class cars like the EX90 in a scalable way.

Discussing the benefits, Bakkenes says: “Very often, companies have one architecture for smaller cars and one for mid-sized cars. They have different physical layouts in terms of engines or electric motors. But with a superset technology stack, we effectively have a one-track implementation for our software and electronics stack, which is a lot more efficient for us as a company.”

“As we build more cars based on the superset tech stack instead of bespoke platforms, it will contribute to our margins,” he adds. As the superset tech stack underpins all Volvo Cars’ upcoming vehicles, it converges all engineering efforts into one single direction, rather than working on specific, siloed, car projects.

Working with Nvidia on AI

The Volvo EX90 (pictured above) is the company’s first car to be software-defined, built on the superset tech stack. It is based on a centralised core compute architecture using the Nvidia Drive Orin system-on-chip architecture.

Nvidia Orin orchestrates everything in the car, from powering the deep learning capabilities underpinning Volvo’s artificial intelligence (AI)-based active safety and driving assistance systems, to helping introduce safe autonomous driving in the future and delivering best-in-class customer experience.

Instead of large steps between different generations of cars, Volvo Cars can continuously enhance the performance of each vehicle in its line-up simultaneously, with the same superset tech stack. It also enables closed-loop development, which allows Volvo Cars to reduce the time it takes to roll out new software features.

For instance, with its EX90 model, Volvo has begun collecting data from fleets of cars to improve the behaviour of the vehicle over time. “We can understand if something is wrong by asking a fleet of cars to collect data from, for example, a certain intersection or under certain driving conditions. We can use the raw data we get back to retrain our data models,” says Bakkenes.

Real-time data for product development

In effect, the closed-loop system, built on the superset tech stack, uses real-time data to improve product development. The data processing is powered by what Bakkenes says is one of northern Europe’s largest AI supercomputers.

Through its fully owned software company – Zenseact – Volvo has been using the Nvidia DGX-based AI supercomputing platform to help develop safe autonomous driving. The company plans to use Nvidia DGX systems, optimised for large workloads, for training current and future AI models, based on uploading data from the fleet of Volvo cars being driven on public roads.

To support its strategy to make driving safer, Bakkenes says Volvo equips all its cars with 5G connectivity.

“With a superset technology stack, we effectively have a one-track implementation for our software and electronics stack, which is a lot more efficient”

Alwin Bakkenes, Volvo Cars

He says the cars make use of AI in a hybrid setup, which uses in-vehicle processing capabilities at the edge, along with the 5G-connected cloud-based processing. 

Through the partnership with Nvidia, he says Volvo’s goal is to equip every car with “potent silicon”,  on which complex AI models can be deployed that make the cars the safest it has ever built. 

Volvo has also standardised on the Google Automotive Android-based operating system for its cars. “What Google gives us is a very good starting position for a stable infotainment system and provides us a way to integrate in a larger ecosystem,” says Bakkenes.

This means Volvo not only uses Google Automotive as a complete application framework for developing its own in-car applications, but it also enables Volvo to work with third-party developers. For instance, in Sweden, Volvo has partnered with parking app providers to integrate the app into the infotainment system. “It makes the experience of owning and operating a Volvo car easier,” he adds. 

Google Automotive also provides Google Maps and Google Assistant for voice control, where the driver simply says “hey Google” and asks Google Maps to navigate somewhere.

Modern cars put a lot of emphasis on the infotainment system. Certainly, car reviews on the web often make a point of highlighting the shortcomings of these in-car systems. Bakkenes says that from a customer satisfaction, loyalty and preference perspective, a car’s infotainment system plays a very important role.

“It’s not just about the media player or the radio, but about how the car interacts with the driver. How you implement that digital relationship while the car is being driven is extremely important,” he says.

Over time, the infotainment system is used to add more functionality to the car, which, for Bakkenes, means Volvo Cars can improve the customer experience by updating vehicles that have already shipped based on the data it collects from the fleet of cars on the road.