I find it strange that Google's self-driving car project appears to be focused on small passenger vehicles, because the technology appears to me to have much more near-term potential in the long-haul trucking market. Self-driving passenger cars, after all, offer buyers the relatively minor convenience of not having to pay attention to the road while riding between urban locations. They also likely improve safety slightly--city driving not being all that dangerous to begin with--at some cost in speed (due to a less aggressive driving style).
Self-driving semis, on the other hand, offer substantial cost savings by eliminating the driver altogether. They likely also improve safety significantly--especially at night, when infrared sensors would surely do much better than sleep-deprived humans--as well as saving considerable time by eliminating rest and meal stops. (For really long trips, truck stops would no doubt be happy to offer filling service for driverless trucks at a tiny fraction of the cost of a full-time driver.)
One can easily imagine companies converting their entire fleet of trucks to driverless models, eliminating their driving staff entirely, and replacing their team of dispatchers with a few driverless truck programmers. Putting aside the technological hurdles--which exist in both markets--it's hard to believe that the commercial trucking market wouldn't be far easier to sell on this concept than even the geekiest first-adopter consumers.
Subscribe to:
Post Comments (Atom)
3 comments:
Excellent point!
There was a recent spate of articles about the technology underlying the self-driving car project. One of their dependencies is having exceptionally detailed mapping -- far beyond the schematic maps that they use in Google Maps. They've got this in Mountain View but not nationwide. (You could imagine they could do this in bulk from their StreetView project.) A second dependency is learning the rules of the road, which is (again, according to these articles) the subject of their ongoing work in Mountain View. Detecting bicyclists, cars, and all the other things that happen in urban driving, and responding appropriately, is a work in progress, although they're getting there fast.
Before you could ponder a product without a driver behind the wheel, you'd also have to have the legal and insurance framework in place, and you'd want to shake out the technology in an environment where mistakes are going to be less costly. That's probably where the slow 25mph buggies come in. No matter what, they're operating slow enough that occupants will likely be able to walk away from an accident.
Once all that is shaken out, then you're ready to ponder the real world. So long as they're still using those laser rangefinders, supposedly $75k/ea, the primary applications are going to be places where you're replacing a commercial driver who, at least for the most part, never needs to get out and do anything else. Taxis are probably easier than long-haul trucking because you can keep a handful of humans around to go rescue the driverless taxis when they're stuck (flat tire, etc.), but you know somebody like Walmart will go for it with trucks as well, for all the reasons you've cited.
Dan, it sounds like you're way more familiar with this technology than I am, so perhaps you know the answer to a couple of questions raised by your thoughtful comment:
1) Would it really be easier to map out an entire urban area in the required detail than, say, the interstate highway routes between a few large trucking terminals (which I assume are typically just off a highway)?
2) Wouldn't the legal implications of Google possibly being liable for death or injury to a passenger of a self-driving car be messier than the standard third-party liability for collisions with other vehicles that's already familiar from everybody's auto insurance? (I believe a head-on collision--more likely in an urban setting than on a highway--with both vehicles traveling at 25mph stands a good chance of causing serious injury to any occupants.)
Post a Comment