Ask HN: Why isn't there more effort to build highway-bound self-driving cars?

[flerovium]

Why isn't there more effort to build highway-bound self-driving cars?

1. Most commuting is on highways. A highway is a controlled environment with very few edge cases--less than existing L4 areas like Arizona and Pittsburgh suburbs.

2. Highway driving captures 90% of the value of a self-driving car, certainly for commuters. Any long trip, where it is valuable to disengage, spends most of its time on the highway.

3. An autonomous highway in a single city would save its commuters an absurd quantity of time and attention.

4. Most new cars can navigate from onramp to offramp. But L4 providers haven't taken the step to let drivers disengage, despite supporting L4 driving in limited non-highway environments.

5. One might claim that it isn't as valuable for ridesharing, but it's is hard to believe that there isn't enough value for commuters, or there isn't enough funding.

6. To handle edge cases, it is always legal to slow, even stop on a highway, and there is always a shoulder to pull over.[1]

7. For rare edge cases, a remote driver can serve as a backstop, as in Waymo's L4 driving.

8. Imagine how much better highway traffic would be if one could disengage and read, work, or think.

I understand that it involves government collaboration, but companies have created significant regulatory change when necessary. For some highways, it requires no government collaboration.

[1] There are rare exceptions, but the geofence could extend only to highways with a shoulder.

[zamadatix]

I'd pick high speed rail before this. Not only do I think it's better for the use case but I don't think anywhere near 90% of the value of self driving cars is on the highway. If anything this would probably increase car traffic delays as cities would become even more inundated with cars once you got off the highway.

[flerovium]

That might be even better, but a single company can make self-driving cars without reforming the American suburban model.

Practically, even in cities with excellent commuter rail (e.g. Boston), car commuters still outnumber rail.

[muzani]

laughs in southeast Asian

I've literally quit jobs because the 3 hour daily commute is too much, and I'd rather not go the 5 hour route by public transport.

[tuatoru]

1. Commuters' start and end points are not on highways. The value of commuting is in arriving at your destination.

2. Attention. How does a car ensure that its driver re-engages when required? Imagine what would happen if thirty or fifty cars pull over because their drivers are busy finishing emails.

2A. If self-driving cars become common on a given highway, how do they cope with griefers? (Other drivers who manipulate the software into giving up, or worse.)

7. Having remote drivers is about as expensive as having an on-board driver.

9. Safety. Self driving software is not infallible, and there are other vehicles on the highway, and unusual situations can develop.

9A. Kinetic energy. Major accidents at highway speeds can be lethal.

10. Publicity. Accidents on highways can also be spectacular/photogenic.

11. Liability risk. Anything that can be blamed on the software, will be.

[tennien]

More fatalities happen off highways, where roads are in unusual condition and there are more edge cases.

1. Wouldn’t a 30min commute be 1/3 as worse there’s a 20-25min section you can disengage?

2. It can force reengagement via the same mechanism for L2 highway driving, but there is a period where one need not pay attention. There are also existing car lots on the sides of highways for maintenance, and they could be added and expanded. At worst, the shoulder provides space for hundreds of cars per km.

3. greifers are a problem for any self-driving system

7. The remote drivers need engage only in exceptional circumstances, as the driver would.

9-11. apply to existing geofences, but with far more added complication, including pedestrians and cyclists.

[SkyPuncher]

I have a Hyundai Palisade with L2 driving. Basically every new car coming out has the same features. L2 handles 98% of highway driving needs.

I've driven 500+ miles including through Chicago rush hour traffic only needing to handle lane changes, interchange merges, and highway exits. Sure it's not set-n-forget, but it's close enough.

For me, it's literally been life changing. I have terrible knees from sports injuries. Long drives use to mess me up for days. With L2, I can cruise for days without feeling uncomfortable.

[rthomas6]

How is this different than a bus? This is what buses do right now.

[flerovium]

1. Buses make frequent stops and are much slower, especially in the US.

2. Most Americans commute from suburbs to cities. This requires a few minutes to get to the highway from a suburb, which is spread out too much for a bus to be practical.

[guilhemherail]

The risk for fatalities goes up with speed. When a vehicle is driven at ~65mph on the freeway, you have less margin for error than when a vehicle is driven in a city at ~20mph. If something comes up, a remote operator may not necessarily have time to safely manage the situation, and the risk of having a catastrophic event is as high as it gets.

[tennien]

Most fatalities happen off highways, despite highways accounting for a disproportionate number of miles driven.

Accidents still happen, but it’s hard to believe they would be less safe than human drivers, when most accidents are caused by fatigue, reckless driving, illegal maneuvers, and unusual driving on the part of the operator. It’s even harder to believe that it wouldn’t become safer with effort, when both the road and the behavior of cars is far more predictable.