If you listen to the companies deploying self-driving vehicle technology, the date for full deployment and L5 capability (full self-driving, no need for driver intervention at all) is just a few years away. Elon Musk felt comfortable enough with the idea to sell it as an option on the Tesla Model 3, though the company stopped offering the capability as a $5000 add-on this week. Still, full self-driving support is often treated as an eventuality that’s going to happen in a vaguely imminent way.
The reality, as Bloomberg explores, is a bit different. Modern self-driving vehicles have an Achilles heel — precipitation. And unless you live in the Atacama desert, the Sahara, Antarctica, or an equivalent location, you’re probably familiar with it. Rain and snow aren’t unusual across most of America, and self-driving cars apparently have major problems with both.
The technical causes are somewhat different, but the net effect is the same: Rain and snow change how the self-driving vehicle “sees” the street. Lidar cameras reflect off both, while cameras aren’t useful in fog or heavy snow. GPS lacks the precision it would need to serve as a standalone navigation service (even assuming it could identify active obstacles, which, well, it can’t).
There are companies like WaveSense, which have a ground-scanning radar system that maps what’s beneath a road and uses this information to detect the position of a car moving at speed. But burying ground-sensing radar stations deployed across America would cost an enormous amount of money. Besides, they aren’t really part of the infrastructure of self-driving vehicles envisioned by the average American when they hear the word. The idea of the self-driving car, at least as it’s been sold thus far, is that the vehicle contains all the hardware it needs to make decisions about its own location and optimal driving parameters, as opposed to a vehicle tied to external interfaces.
We’re not intrinsically opposed to the concept, to be clear, particularly not if that’s what’s required to make self-driving cars with Level 5 autonomy practically work. But it’s an area where the barriers to wider adoption aren’t always widely disclosed, and it suggests that there may be a substantial pause point between the deployment of driver-assist technologies, like those already found in many vehicles, and the full self-driving capabilities many Americans look forward to.
But it’s not crazy to contemplate these ideas, given that some analysts believe self-driving cars could make vehicle ownership itself a thing of the past. Left unsaid is how much life would change for those who have to continue to own their own vehicles if vehicle ownership became non-standard, and it was harder to find mechanics, parts, and components that weren’t under license to function as beholden repair centers for specific companies. Again, this would never be a problem in more densely populated areas where you’ll always have a population of shade-tree mechanics and companies that cater to them. But it could become an issue in rural areas, where there aren’t as many companies, period. Snow, rain, mud, and road washouts are all larger problems in these areas as well, and the deployment of buried ground-penetrating radar systems (or any system required to enable full self-driving) would be quite expensive relative to the tax base of the areas such a system would pass through.
New Report on Self-Driving Cars Ranks Tesla Dead Last
Tesla has played up its self-driving credentials for years, but a new industry report ranks the company dead last.
Waymo Buys ‘Thousands’ More Self-Driving Vehicles From Fiat Chrysler
When all the receipts are tallied, Waymo has surely just dropped tens of millions to expand its fleet.
Uber, Lyft Want to Ban Personal Use of Self-Driving Cars in Urban Areas
A new roadmap for self-driving car development claims that vehicle fleets should only be operated by companies in urban areas.
Perrone Robotics: We’ll Make Cars, Vacuums, Mining Trucks Self-Driving
No-longer-startup PRI sees scalable autonomous platforms. Even Raspberry Pi is enough for simple autonomy. Vertebrate Lab learnings may change how we think about autonomous devices.