3 ms·

Generalized self driving cars seems like such a hard problem, laced with a huge amount of hard to handle edge cases and a huge amount of hard to enumerate possi

by briatx 8y ago

Generalized self driving cars seems like such a hard problem, laced with a huge amount of hard to handle edge cases and a huge amount of hard to enumerate possible scenarios. I think it will still happen, but they need to simplify the problem. The way to simplify it is to create automated driving lanes, similar to the current HOV lanes, that have standardized markers that are easy for computers to read, and then eventually extend it to automated driving routes. That will take some amount of buy-in by the government to install, but I have faith in our well paid lobbyists.

danenania 8y ago
Afaik, highways in non-severe weather are already pretty much solved, so I’m not sure that special lanes would make much difference, except perhaps during snow storms. The edge cases seem to be mostly around city driving, rural back roads, and bad weather.
- blattimwind 8y ago
  Just like teaching humans to drive. Yes, sure, you have to show them how to do it on the highway/BAB, driving safely at 130+ km/h, but it's really not hard (over long distances things you can't really teach make it difficult). City driving on the other hand takes time to master; here easily 80 % of the driver education is done in cities (0-50 km/h). Practically all the edge cases, all the difficult situations and complex maneuvers are done in cities. Outside cities driving is very simple and safe (if you stick to the limits).
- briatx 8y ago
  > highways in non-severe weather are already pretty much solved And yet all the automated driving driver fatalities have been on the highway, with the Mountain View crash in particular having been caused by weathered / confusing road markings. I wouldn't consider that solved at all. But to the larger point, self driving lanes would only be a first step. Eventually it would develop into a standard set of markings and signage for automated driving friendly roads so that you could safely have self driving cars in things like snow storms. There's no need to limit it to highways only, although there I think there is some benefit in the near term to segregating automated driving cars from human driving ones.

by briatx 8y ago

danenania 8y ago

Afaik, highways in non-severe weather are already pretty much solved, so I’m not sure that special lanes would make much difference, except perhaps during snow storms. The edge cases seem to be mostly around city driving, rural back roads, and bad weather.

blattimwind 8y ago
Just like teaching humans to drive. Yes, sure, you have to show them how to do it on the highway/BAB, driving safely at 130+ km/h, but it's really not hard (over long distances things you can't really teach make it difficult). City driving on the other hand takes time to master; here easily 80 % of the driver education is done in cities (0-50 km/h). Practically all the edge cases, all the difficult situations and complex maneuvers are done in cities. Outside cities driving is very simple and safe (if you stick to the limits).
briatx 8y ago
> highways in non-severe weather are already pretty much solved And yet all the automated driving driver fatalities have been on the highway, with the Mountain View crash in particular having been caused by weathered / confusing road markings. I wouldn't consider that solved at all. But to the larger point, self driving lanes would only be a first step. Eventually it would develop into a standard set of markings and signage for automated driving friendly roads so that you could safely have self driving cars in things like snow storms. There's no need to limit it to highways only, although there I think there is some benefit in the near term to segregating automated driving cars from human driving ones.