Even single automakers have enough cars out there to make this practical. (Not counting Tesla.) It seems like the easiest way to update the maps. Google street view gets out of date fast, not too bad for that application but bad for AVs.
They don’t, not nearly. While (let’s say Ford) technically travels enough road to scan everything in a year, it doesn’t cover nearly all the roads (most driving is the same set of trips over and over), and not a single Ford vehicle currently has the technology…it would take decades and decades to mature.
But that’s not why I came back to the thread…I thought this was a good summary of some of the broader AV discussions that are happening.
www.sciencemag.org/news/2017/12/are-we-going-too-fast-driverless-cars
From an economic perspective, L3/L4 probably make sense but L5 requires increased investment with (possibly) diminishing returns.
That last 10% of scenarios/functionality will probably stretch out a long time.
After they ramp, of course. Current Ford cars are not going to do much good. Maybe early adopters will have to take over to record uncharted roads. Maybe Google will sell them this stuff. Maybe they’ll lend a recording car to their dealers and tell them to get recording in order to get AVs to sell.
As for the article, I wonder how useful research on the impact of the automobile would have been before it was introduced. Not very, I expect.
I mean, cross-licensing is pretty common in the industry. There’s no reason one company has to do it all, and it’s mutually beneficial.
Meh. Not the mag’s best review article.
-
It does seem that “level 4” is wide swath. If “certain conditions” means only good weather in very limited locations then it is not fitting what most are expecting from level 4. Most expect level 4 will cover the vast majority of circumstances. My daily commute and errands under all weather conditions reasonably expected for me to experience and visit to my brother’s. We don’t expect it do do off road in a dust storm.
-
Using a short term study of giving people use of a chauffeur-driven car for up to 60 hours over 1 week to predict how people will over time use AVs is just dumb. Give a kid a new toy and the first week they’ll use it a lot. Once the novelty wears off usually not so much so.
-
Yes, the questions of owning your own vehicle vs a service, AV as a service, and regulation, are all open. Many are unanswerable right now. Which does not mean we need them before it proceeds.
I doubt more than 1 in 5,000 people think about Levels at all. And when they do think about AV, they think about going to or from work without intervention, or their friend’s house, or shopping, or whatever. The fact that those can be Level 5 *or * in some cases Level 4 is of no interest to them I’d guess.
When I started the thread I didn’t know about the levels, and I agree that the broad range of level 4 muddies the waters. An airport shuttle that runs a regular loop 50 times a day would be a level 4 vehicle, and I could see those starting to roll out in the next few years. A personally owned car that could take you to 98% of mapped addresses on 99.5% of days (.5% being inclement weather that it doesn’t think it could handle so it refuses to go) would also be a level 4 vehicle, and yet that would be a massive safety improvement, would be perfectly acceptable to most consumers, and I would certainly consider that to satisfy the premise of my OP. In that sense, I think level 4 vehicles are years away and also decades away.
That’s the exact L4 example I’ve been using for a while (the airport loop, with McCarran being a good possibility given NV regulation and weather), and yes, it’s very broad. But it does work as a shorthand as “AV with boundary conditions”…and defining those boundaries will be important for regulatory and insurance reasons. But to your point, I’m not sure the levels are much help to anyone not trying to get this technology up and running.
As for the cars updating the maps …there are overlapping concepts here.
My initial post was actually thinking about the individual vehicle’s private data set developing a highly detailed and recently updated map of the roads most traveled on, and thereby better able to impose top-down expectations on fuzzy information coming in (there was a stop sign there yesterday, now there is a white blob with a bit of red amidst drifts of white that is the height of the stop sign that was there …, there was a pothole in this location yesterday, plan for its still being here …).
Uploading and sharing that information, to a cloud and/or by querying and being queried by other local vehicles (V2V), and thereby having detailed deep information of areas where the most vehicles are most used and increasing information for roads less travelled, is an overlapping item. That one of course comes with increased security concerns. V2V so far is, as I understand, more immediate hazard focused and the more open the system the more hackers can play.
That’s a reasonable distinction I wasn’t considering, though would the AI be local to the vehicle? As long as data is getting sent to the cloud for analysis it might as well be used for the broader purpose.
That reminds me, I’ve seen wildly varying estimates of how much data would get uploaded by every AV every day, but they’re all big numbers. Fortunately quantum computing will come along any day now to save us.
I was presuming local but my ignorance may be showing.
Ars Technica article – self driving cars are here (well – in Phoenix – or more accurately – a subsection of Phoenix)
A long way until even 51% of the country in 51% of conditions, but still a step.
Brian
Apparently, Hyundai is already at Level 4 with their Nexo. Cite.
The Nexo goes on sale in March in South Korea. That’s March 2018.
Your own quote says it won’t be level 4 until 2021 (estimated). We shall see. The description of the highway demo doesn’t sound any different from what Tesla and others currently have.
Let the decades roll on!
:dubious:
I got the magazine from AAA today, and the first editorial was about how they are running the autonomous shuttle in Vegas and how they are partnering with an autonomous vehicle test area.
I can imagine the strategy meetings there about how to have a club for drivers when no one drives anymore. The tow service will still be useful, I suppose.
They seem to be taking this pretty seriously.
Google (Waymo) is going to start autonomous vehicle ride sharing service in Phoenix very soon. Self-driving cars are already here.
Here’s yesterday’s science fiction comic strip Brewster Rockit, with another possible hazard self-driving vehicles might present: Brewster Rockit by Tim Rickard for February 22, 2018 - GoComics
The complications of keeping sensors clean…