The slow and steady march towards full automotive autonomy has, from the very beginning, been closely related to the advancements we’ve seen in the world of artificial intelligence. However, few companies working to develop autonomy are happy to let the machines do all the learning. Most have some way of getting in there and massaging the data, annotating or tagging or otherwise categorizing objects or scenarios to enable more control over the resulting digital driver that will, some day, take you from A to B.

Nissan and its partner Wayve have taken a different approach with what will be the next-gen ProPilot driver assistance system. Nissan’s engineers didn’t spend time teaching the car what a pedestrian is or how they move, what a cyclist is or how they behave, or indeed how to deal with a double-parked car on the side of the road.

They simply took a ready and willing digital brain and showed it how to drive. They fed it raw footage of good drivers driving in good ways and basically let the network figure it out. According to Tetsuya Iijima, Executive Chief Engineer of ADAS at Nissan, who led me on a scenic trip around Shimbashi and Ginza, the cars simply picked up the habits of those certified drivers. I have to say, the result was quite impressive.

The system, which Iijima called a brain, runs on a single Qualcomm SoC, but he was adamant that they’re not using Qualcomm’s dedicated automated driving software stack. The software I experienced comes to Nissan via Wayve, a British AI startup that’s helping Nissan make a big leap forward in the autonomy game. It’s potentially a great business pairing for Nissan, a brand that, in the U.S. at least, isn’t exactly known for pushing the envelope technologically. The company is indeed taking a somewhat conservative approach here, too, but we’ll get more into that in a moment.

Nissan doesn’t have a formal name for how it will market this system, just calling it a next-generation ProPilot, but as you can tell by the pictures, it relies on some extra hardware. The hat that Ariya is wearing isn’t just for style. Some extra sensors are hiding up there. The full suite includes 11 cameras and five radars looking all around the vehicle, plus one lidar sensor looking straight ahead.

Interestingly, though, there’s no complex sensor fusion happening here. The car relies entirely on the cameras at lower speeds, which meant my trip through Tokyo didn’t touch the lidar sensor at all. That’s only called in for higher speeds on the highway, where Iijima said cameras can’t see far enough to make the vehicle safe at speed.

And what about the radar? Those are for redundancy, but they’re not really tied into the autonomy feature at all. The Ariya still has its basic automatic emergency braking, which provides something of a safety net for this AI driver, much as it would for a human one.

Iijima slipped into that autonomy trope of saying the system is already safer than a human driver, but in the roughly 45-minute tour of Tokyo I received, I didn’t spot anything that contradicts that. He didn’t have to intervene once. The car handled everything from illegally parked delivery vans to indecisive pedestrians making a last-minute dives for the crosswalk.

Turns across busy lanes were handled without issue, even when the late-day sun was shining right into the car’s main lens. Most impressive was how smooth the car was. There was no jerkiness to the steering, no abrupt applications of gas or brake in reacting to those flocks of on-foot commuters.

That was helped by the system’s overall relaxed way of driving. The car accelerated slowly and took its time progressing through intersections. Chances are, you or I would have been more aggressive, even in this most polite of driving countries. But honestly, when the car is handling the duties behind the wheel, I never find myself in a rush. I certainly wasn’t on this day, anyway: My destination was the starting point of our Ginza loop.

Overall, this Nissan and Wayve mashup didn’t do anything that Tesla’s “Full Self Driving” can do, or the new MB.DRIVE Assist Pro that Mercedes-Benz is set to debut on the CLA this summer (which I wrote about here). But, it did it all more smoothly, and in about as bustling an urban environment as you can get.

So it’s all very impressive, but how is Nissan being conservative about it? The company is taking its time with the rollout. It won’t hit Japan until sometime in the fiscal year 2027. Iijima said it will come to the United States, but not until some time after that.

The big question is whether and where it will be hands-off in the U.S. Iijima said the technology is more than capable, even to the point of operating at Level 4, eyes-off mode. How Nissan restricts the system in the U.S., then, is largely a question of liability. Given the litigious nature of American drivers, it’s probably a safe bet that we’ll still be required to keep our hands on the wheel and our eyes on the road.

Iijima said that this feature will likely be an options package, and while he wouldn’t speculate on pricing, it’ll follow Nissan’s value-packed brand image. “We don’t want to sell at a very high price tag,” he said, and it actually could be cheaper than you might think.

“Currently, each sensor has its own brain,” Iijima said, resulting in a series of processors scattered around the vehicle. But with the new Qualcomm processor, there’s enough to process all that data centrally. “[It] is so intelligent and capable that the raw signal can be used. The hardware is reduced,” he said.

Iijima said that, once the initial R&D cost has cleared Nissan’s books, this could actually be cheaper than ProPilot 2.0 on cars today. That’s because current systems rely on a mesh of processors and sensors scattered about the car. This next-generation system simplifies all that onto a single chip.

Beyond that, Nissan won’t take the approach of just throwing the sensors on every car and then charging a fee to enable it. You’ll only get it if you pay for it.

It’s all sounding very promising, but it’s also sounding a long time away. In the accelerating world of driver assistance, 18 months is a lifetime. I’m tempted to say the autonomy landscape will look radically different in a year and a half, but then all this stuff is taking much longer to come to market than anybody thought, so it could just turn out that Nissan’s timed this perfectly.

Thanks for reading Around the Next Bend! This post is public so feel free to share it.