-
Mia Khalifa’s Response to Dating Rumor Goes Viral - 9 mins ago
-
These tenants want to own their buildings. L.A. is collecting millions to help - 14 mins ago
-
International Criminal Court Issues Arrest Warrants for Netanyahu and Gallant - 31 mins ago
-
Snorkeler Channels Steve Irwin in Dream Encounter With Ocean Creatures - 43 mins ago
-
Do California teachers have the right to slam Trump? Yes. And no - 55 mins ago
-
Trump’s Tariffs - about 1 hour ago
-
Musk and Ramaswamy Unveil Plan to Overhaul Government - about 1 hour ago
-
Wildfire smoke is polluting California, New York and beyond. Protect yourself - 2 hours ago
-
Major Atmospheric River To Hit California With ‘Life-Threatening’ Floods - 2 hours ago
-
Israel Strikes Near Beirut as U.S. Envoy Pushes Ahead on Cease-Fire Talks - 2 hours ago
Volvo Relying on ‘Relentless’ Data and Engineering to Make Cars Safe
Ninety years ago, the Volvo brand was born with a firm mission statement in a handbook. It said that “an automobile is driven by people,” and that “the fundamental principle for all construction work is and must therefore be, safety.”
In 1959, a Volvo engineer perfected the idea for a modern three-point safety belt. Instead of keeping the innovation to themselves, Volvo gave away the plans. To date, that simple design has been credited with saving at least a million lives.
That same drive toward safer cars propelled the company to standardize the headrest in 1970. Today, SAE International, a global association of more than 128,000 engineers and related technical experts that develop consensus standards that set the bar for the automotive industry, has guidelines for headrest best practices and governments regulate their application because of their effectiveness in preventing and lessening injuries related to collisions.
Today, Volvo’s vehicle safety comes not just from materials like fabric, metal and plastic, but from millions of lines of code, dozens of different sensors and more than one type of artificial intelligence. All that is housed in the Volvo Cars Superset technology stack.
The company’s future electric models will use the same base systems, modules, software and hardware. Like Legos, Volvo’s Superset building blocks can be configured in many different ways. Depending on vehicle needs, different elements can be incorporated or removed. This is different than how many automakers create vehicles today.
“We have variants for things like air suspension and then we actually pick the pieces across the entire technology stack. We make them compatible with themselves. We’re doing the same when it comes to advanced driver assistance system sensors and sensor configurations,” Alwin Bakkenes, head of software engineering at Volvo Cars told Newsweek.
“And because we built this as a bunch of Lego bricks, we actually combine this and configure this to product specific variants, which means that we’re actually not following the traditional platform strategy, but we have this big toolbox which we continuously evolve.”
The company says this approach allows all its engineering effort to be channeled into one single direction. That will lead to improved quality, increased speed-to-market and vehicles that improve over time using a closed-loop development process, it says. For drivers, that means their vehicles get safer and more capable in a repeating cycle.
The new Superset was first featured on the Volvo EX90 electric SUV. Work on it will benefit the upcoming EX60 and the rest of the electric lineup.
Bakkenes made the analogy of an iPhone and the iOS operating system. It works the same on your computer, tablet and phone, and is updated regularly. The Volvo system will work the same way in different vehicles.
All Volvos now have massive amounts of sensors so they can “see” the world around the vehicle in ways never seen before. Volvo has its own perception algorithms and is basing everything on core computing so it can run these advanced AI models.
“We invested in one of Europe’s largest AI supercomputers and data centers to collect data from our fleets and build our models on top of that, and then can continuously improve the performance of these systems that we’re deploying. This is how we go from innovating on for things such as safety belts back in 1957 and now we’re innovating based on data,” said Bakkenes.
Volvo notes that it never takes data from unwilling customers as those transfers are all an opt-in service.
The company says the software stack is compatible with several generations of physical hardware. Newer displays may be snappier with faster computer chips, but it can deploy the same functionality to all vehicles.
More like a video game system, the Superset should improve over time as the company and other developers learn how to use it to its full potential.
“We talk about relentless progress. We want to be relentless about optimizing and improving over time. The whole key to this is data. The more we understand the performance of the product, the better we can make it. We continuously analyze metrics to make sure that we understand how we can make them better. And I think we’re confident that we’re going to make them even better over time,” said Bakkenes.
That includes Level 3 hands-free, eyes off advanced driver assistance technology, which is in the works for this new Superset. The company says it has a sensor set that enables full self-driving in highway-like scenarios, unsupervised, safely.
“We have redundant computing in the EX90. We have redundant braking, redundant steering, and all these kinds of systems in place. We will be starting actual in market verification in California, and we intend to roll out and activate on existing cars full self-driving capabilities,” he said.
Volvo works with Nvidia on this updateable platform, using AI to improve both safety and convenience. It takes information from all the perception sensors be it radar, lidar, camera or other, to understand the best way to navigate the vehicle through its surroundings. All the processing happens on board. When an accident or even a near miss happens, if the user allows it, Volvo will look at the data later to make updates.
“What we’ve done in the past is analyze that manually, and then somebody would code an updated piece of software, and then you send it back. What we’re able to do now is to say things such as, ‘well, actually this is a very interesting type of scenario, which is a bit tricky.’ So whenever we’re in a scenario like this, send us that data,” Bakkenes said.
“Then we can use that to directly train the algorithm. And that’s the power of this new Superset tech stack.”
Source link