Tesla FSD has recently received permission to be activated in the Netherlands. This makes it the first country with FSD in Europe. My friend Dick Helders of XpengCam, who sometimes comments here on CleanTechnica, was one of the first to drive it in the country. He has a unique perspective as one of the few people who has driven VLA 2.0 in China and FSD in Europe.
The test comes in two parts, with the first part navigating some semi-urban spaces around the town of Arnhem, Netherlands, and the second part driving on some of the more suburban roads. I recommend watching both parts.
Part 1. Dubbed in English using automatic narration.
Part 2. English speaking and far more entertaining. More issues arise in this video.
In the Netherlands, there are narrow roads without a center line but with dotted lines on both sides. When two cars pass, they tend to need to go on to the right of the dotted lines to provide adequate clearance. While the Tesla did not drive completely down the center of the road, it tended to stay to the left of the dotted lines, forcing the car in the other lane to almost have to go off the road. In the one instance where there was a bus that was too wide to go over far enough, the Tesla did move over somewhat to avoid an impact during essentially a game of chicken. But it understandably made the driver uneasy and likely made other drivers who had to pull over farther a bit agitated.
The Tesla had problems with speed limits in the test, often going at the wrong speed. The on/off nature of FSD does not allow the co-driving seen in VLA 2.0, which lets the driver influence the intelligent driving system without disengaging it. FSD also parked in illegal spaces on the drive.
FSD also missed changing stoplights several times. The Netherlands has a density and complexity of stoplights that is higher than most parts of the US. That seemed to confuse the system, not knowing which light to look at.
FSD also seemed to pick the wrong lane, merging across multiple lanes at the last second to make a turn. That might not be an issue in parts of the US or some smaller towns, but it could become a problem in a large city with dense traffic, where merging becomes more challenging.
In the end of part two, FSD “crashed” but it was the software crashing, rather than the vehicle physically crashing. It is unclear why the system crashed, stopped working and locked him out. The car wasn’t in extreme weather or road conditions at the time. However, it did seem to give adequate advanced notice to take over and did not seem to put the driver in jeopardy of an accident. Does anyone with extensive FSD experience know what happened?
Overall, just looking at the steering wheel, it wasn’t the smooth, continuous movements seen in VLA 2.0. The Tesla often had to correct its trajectory. Some of the issues were things that felt unsafe, even if they were technically legal and didn’t put the car and its passengers in imminent danger. In some cases, FSD had challenges with like right of way. However, issues like that in the Netherlands would not be an issue in a different country with different written and unwritten rules.
However, this gets into one of the challenges of FSD in comparison to VLA 2.0. Tesla sends data to a central location where people label it, whereas VLA 2.0 actively learns from drivers. All decision making happens onboard with far more onboard effective processing power. In addition to learning through extensive simulations in its World model, the XPENG system has the ability to learn the unwritten rules of the road much more quickly.
FSD relies on a more centralized system. This is fine for the US, where the rules of the road are largely the same and driving is a bit less complicated throughout most of the country. However, both the written rules and unwritten rules can change significantly across Europe. As such, not only does Tesla run into data privacy concerns with its centralized data collection that will make it more difficult to meet UN and Chinese intelligent driving regulations, but the more centralized system inherently has a more difficult time adapting to local driving circumstances.
Overall, the review wasn’t completely critical, pointing out both positives and negatives. Advanced ADAS systems do have potential to add convenience and reduce accidents in Europe, including FSD. Even if FSD isn’t ready to hand over complete control.
However, the reviewer said that FSD was not as advanced or capable as VLA 2.0. Having ridden in a L03 prototype in Munich, VLA 2.0 already seems to be more advanced in Europe. And the XPENG system was a beta test in Germany, months before official release, while FSD is already offered to the public in the Netherlands. Knowing how fast VLA 2.0 learns, it will undoubtedly be far more advanced in Europe by the time it is released.
Of course, many will say Dick is biased as an XPENG owner and part of an XPENG user group. However, knowing him personally, he is also direct and does not hold back criticism, even of XPENG. Just straightforward by nature (and a lot of fun in person). But he is one of the first people to drive FSD in the Netherlands. And one of the very few who also has direct experience with VLA 2.0.
Many assumed that once the government approved FSD in the Netherlands that the rest of Europe would soon follow. However, with FSD facing regulatory hurdles and a level of secrecy in Europe around performance, EU-wide approval seems increasingly uncertain in the near term.
But what do you think? Will Tesla FSD improve performance and get approval across Europe soon? Or will other companies step up and take the lead?