In 1959, Volvo gave us the seat belt—here’s what its safety team is building now

GOTHENBURG, Sweden—Several years ago, Volvo Cars announced its “Vision 2020” plan. The goal was as simple as it was bold: by the year 2020, no one should be killed or seriously injured in a new Volvo.

As we get closer to that date, Volvo has started to get more specific about how, exactly, it plans to get there. First, there was the announcement earlier this month that, starting with model year 2021, all new Volvos will be restricted to 112mph (180km/h). And last week, the company invited journalists from around the world to visit its headquarters in Gothenburg, Sweden, to learn about its other plans for this initiative.

As you might have expected during such a tour, Volvo was at pains to let us know exactly how many safety innovations the company has been responsible for in the years since it introduced the three-point safety belt as standard in 1959. Side-impact protection structures, side airbags, whiplash-preventing seats, blind-spot monitoring, and plenty more active and passive safety systems throughout the years have helped the company earn its current reputation for safety.

“Looking into the data, we’ve done a lot with passive and active safety, but to get to zero, you have to tackle human issues,” says Håkan Samuelsson, president and CEO of Volvo Cars. By this, he means the trio of distracted driving, intoxicated driving, and inappropriate speed. “Now we’re coming into a situation where we have a technical capability to do something about this. We can let the car intervene if the driver is behaving badly. For example, driving outside a school: is it really individual freedom to drive past it at 250km/h? Do we have the right to intervene, or do we have an obligation? We want to enter into a dialogue, we don’t have an answer, and we don’t want to be big brother, but Volvo can lead the discussion on safety.”

333 comments

The speed-limiter announcement certainly sparked a discussion—333 comments and counting just on our short article alone. Although Samuelsson has said it’s worth doing if it saves even a few lives, he’s open about the fact that the speed limit was really the company’s way of throwing down a marker. “We are quite sure we need to send a signal of some kind to start this dialogue. But I think indirectly with that, you attract certain customers and probably discourage other customers, and I think we did this rather deliberately. We want to create a stronger and safer Volvo community. Now we’re opening up the discussions with insurance companies to see what benefits this would bring to Volvo customers. We want to attract people who want to drive safely. That would discourage the boy racers,” he said.

However, don’t expect future Volvos to have restricted acceleration to go with their new, lower top speed. “If you were to have some kind of emotional value of a powerful car, you feel that with acceleration, but you very seldom feel that with speed. And you could argue that, in some cases, it means safety—if you could complete an overtaking maneuver quickly for example. So we have no plans to limit acceleration, and with electric cars it will go in the other direction. There will be tremendous acceleration with electric cars, but they will never be very fast, because you empty the batteries too quickly.”

Volvo typically conducts 60-70 crash tests before approving a new model. While the XC90 has been on sale since 2015, it’s about to go through its mid-life refresh, hence the bashed-up orange ones you see illustrating this article.

Jonathan Gitlin

Volvo typically conducts 60-70 crash tests before approving a new model. While the XC90 has been on sale since 2015, it’s about to go through its mid-life refresh, hence the bashed-up orange ones you see illustrating this article. Jonathan Gitlin

This is the immovable object. The MY2020 XC90 hit it at 50mph, with a 40 percent offset.

Jonathan Gitlin

This is the immovable object. The MY2020 XC90 hit it at 50mph, with a 40 percent offset. Jonathan Gitlin

The aftermath.

Jonathan Gitlin

The aftermath. Jonathan Gitlin

This is the immovable object. The MY2020 XC90 hit it at 50mph, with a 40 percent offset. Jonathan Gitlin

The aftermath. Jonathan Gitlin

That’s lipstick on the passenger frontal airbag.

Jonathan Gitlin

The XC90’s trunk was packed with telemetry and recording devices. In addition to the 60-70 physical crash tests, Volvo will also conduct around 20,000 simulated tests when developing a new vehicle.

Jonathan Gitlin

Volvo’s crash-test center can send cars down two long tunnels. This is one of them. The moody red and blue lighting was done for our benefit—in normal use, it’s just gray down there.

Jonathan Gitlin

Volvo built this moose test dummy to simulate the effect of hitting a large quadruped at speed. Milliseconds before impact, the moose dummy is released from the support structure so it’s not attached to something before the crash takes place.

Jonathan Gitlin

Driver monitoring

Volvo thinks the solution to the problem of distracted driving and intoxicated driving will be the introduction of a robust driver-monitoring system, due to arrive with the second generation of Volvo’s Scalable Product Architecture in the early 2020s. (Scalable Product Architecture is the toolbox from which it makes the S90/V90/XC90 and S60/V60/XC60 vehicles, and the first SPA2 car will be a new XC90, probably by 2023.) The company is still deep in the research and development phase, so it’s premature to discuss the specs of the system at this point. But we do know it will involve gaze-tracking cameras, like the DMS that supports Cadillac’s Super Cruise feature.

That doesn’t mean a Volvo that pulls over to the side of the road with its hazard lights the first time you look away from the road for more than 15 seconds. Rather, the car will monitor its driver’s behavior, both for attentiveness via the cameras but also from their control inputs via the steering wheel and pedals. “We’re using multiple sensors, looking at every bit of information we can get from the car,” said Armin Kesedzic of Volvo’s active safety department. “How the steering wheel is being handled, how the pedals are being used. Gaze tracking is not our sole focus; we’re looking at all the eye behavior we can see. We’re also looking at facial expressions, body tilting; as much information as we can get.”

And in extreme situations, where the car detects a high risk of a collision or the loss of control, it will intervene. “The human driver is really very good, but we are still humans, so it varies over time,” explained Malin Ekholm, who runs Volvo’s Safety Center. “When you’re really at your best, I’m happy to let you drive. But when you’re not, I would like to support you to make you better than you are in that particular situation. And how we do that, that’s where we have to have this conversation.”

So the system would only activate under extreme circumstances. Examples include the driver taking their eyes off the road (or closing them) for a lengthy period, a similarly lengthy period of no steering input, extreme weaving across the road, or excessively slow reaction times.

Credit: Volvo

It also doesn’t mean Volvo (or anyone else) will be using the cameras to spy on you. In fact, the cameras won’t even be able to save raw video. “To us, a camera that delivers video is a useless sensor,” explained Mikael Ljung Aust, a researcher and driver behavior specialist with the company. “We don’t want the raw data back. We want your head position in numbers we can use. We want your gaze vector in numbers.”

At the past couple of CESes, I’ve seen a number of in-development DMSes, some of which go beyond simple gaze tracking to impute one’s emotional state from things like skin conductivity and respiration rate. While Ljung Aust and his colleagues have been using these tools as part of their research, he was skeptical that we’d see them show up in a production system any time soon.

“Typically, physiological measurements are hard to apply in vehicles, at least if you’re into critical stuff,” he told me. “We have an excellent driver-alert system that looks at your driving and picks up drowsiness, but anything faster than that, physiological measures are too slow to respond. There’s a delay that doesn’t quite keep up with what’s happening on the road, so you’re late picking it up if something goes wrong. If you’re looking for the more real-time stuff like you would do with an EEG or a helmet-and-magnet setup, the signal you’re trying to tease out is one-tenth of the background radiation.”

(Plus, as we both joked, it would be too hard to make sure the driver shaved their head and then covered it in conducting gel before running out on a quick errand.)

I also asked Volvo about the problem of ethnic or cultural bias in AI-trained camera systems. After all, there have been plenty of instances of a system working great when trained on white male faces, then stumbling when applied to women or people of color. “When we roll it out, we’ll make sure all the markets where we roll it out have this verified for all ethnicities,” Kesedzic told me. “It is a big question, and you need to do a lot of verifying, but we’re prepared to do that.”