EVNow
Well-Known Member
Apparently people quote a bunch of posts for no reason. Bizarre.
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Autonomous Car Progress
- Thread starter stopcrazypp
- Start date
-
- Tags
- Autonomous Vehicles
Knightshade
Well-Known Member
Yeah that was baffling.
Both things he quoted me saying were 100% factual at time of posting
Doggydogworld
Active Member
I'm happy to say I was wrong. I did not think Waymo would open downtown Phoenix to the public so soon, nor expand the service area this quickly. They're only doing it in response to Cruise, but a last-minute response is better than nothing!
Most interesting to me is the airport service. That's actually a workable business model, if they market it.
San Francicso is the better market, of course, at least in the near-mid term. They still trail Cruise in deployment there, but it's early and one Cruise misstep would leave the door wide open.
KArnold
Active Member
Are driver-less cars an option in these cities where offered? Or is it by chance?
Do they charge the same rates with or without a driver?
Do they charge the same rates with or without a driver?
diplomat33
Average guy who loves autonomous vehicles
We can't say that because we don't know if phantom braking would be even worse if Cruise had less sensors. But the fact is that more sensors don't automatically mean zero phantom braking. It depends how good your sensor fusion is. Good sensor fusion can reduce phantom braking due to false positives but it won't necessarily eliminate it completely. It should also be noted that phantom braking can occur for other reasons than bad perception. For example, phantom braking can be caused by bad prediction or bad planning. For example, if the AV thinks a vehicle will pass in front of the car and brakes but the car actually turns and does not cross paths or if the car plans a path and then realizes late that it will hit another vehicle so it brakes to avoid the collision. Tesla FSD used to brake hard for cars turning in front of the Tesla because it did not seem to know the car would get out of the way in time. That was phantom braking not caused by bad perception. That type of phantom braking will not be solved with more sensors. You can only reduce that type of phantom braking by training better prediction and planning.
Phantom braking can never be eliminated completely. Machines do it; humans do it.
Phantom braking happens when the driver senses something that cannot be identified immediately and perfectly and that could contribute to an accident. Then the driver must act to reduce the risk.
If it then turns out to be a misunderstanding, so there was no danger, we call it phantom braking, meaning that the driver sensed a phantom. But, in fact, the driver did sense something.
The question is only how often it happens. If it happens too often, we rightfully blame the driver, machine or human. But completely eliminating it is impossible.
Humans have the advantage of having a lot of background knowledge about the world. Machines have the advantage of being fast and precise. Conversely, humans react slowly, while machines do not know and understand many things. These shortcomings contribute to phantom braking.
Things seem to be progressing. I am pleased to say I am surprised at the rate, I did not think it would. Then again I did not post constantly in a didactic manner and have an inability to say ...hey I was wrong.
Waymo and cruise pushing it all along. Funny stuff. Baidu too I guess.
Do you guys think this is going to be an issue for Tesla ?
Waymo and cruise pushing it all along. Funny stuff. Baidu too I guess.
Do you guys think this is going to be an issue for Tesla ?
diplomat33
Average guy who loves autonomous vehicles
Not yet, no. Tesla and Waymo sell different products. Waymo sells autonomous rides. Tesla sells EVs. Even if people in SF and downtown Phoenix don't buy a Tesla because they can take a Waymo ride, that would be a drop in the bucket compared to Tesla's total car sales. And Waymo rides are great for certain people but there is still a need for personal cars for people who want to go outside the Waymo service areas. And Tesla makes good EVs with a good driver assist. There is still a lot of demand for Teslas. I think the bigger threat to Tesla will come from other EVs with L2 "door to door" systems. Tesla should worry when consumers can buy a non-Tesla EV that is affordable, with good range and good charging, with a good L2 "door to door" system because it will mean less demand for Tesla and less revenue. We are not there yet but non-Teslas are getting better and there are alternative L2 systems like Ultra Cruise and Super Vision that are coming.
But I do think it should be a wake up call. The future is autonomous. Robotaxis will expand to more and more places. And Mobileye is promising "L4 everywhere" on personal cars starting in 2025. GM says they plan to do L4 on personal cars around "mid decade". As the tech gets better, I think we will eventually see fully autonomous consumer cars. So eventually, if Tesla cannot offer true autonomy, I think it will be a big issue since consumers will be able to get autonomy somewhere else, either as a robotaxi or as a consumer car.
AFAIK, Waymo can charge for rides in downtown Phoenix now. And they just recently applied for their permit to charge for rides in SF. As soon as they get that permit, they will be able to charge for rides in all of SF, 24/7.
Most important for Waymo is that once waymo charges they can stop sharing data on a disengagement’s. The permit is the state of CA , like you said they can charge in AZ.
Waymo can expand throughout CA now with a pretty good permit, day and night and high speed.
diplomat33
Average guy who loves autonomous vehicles
Not sure why that would be so important for Waymo. Once Waymo is charging money for driverless rides in all of SF, their disengagement rate would have to be extremely good. Incidents will still happen of course but they should be rare enough and can be handled outside of the CA DMV disengagement report process.
Perhaps give this a read , I found it interesting. I am not an expert so maybe you know more about this permit issue than I.
Robotaxis Are Everywhere in San Francisco Now. But Something’s Wrong.
Blocking transit. Driving on the sidewalk. Speeding away from cops.
Last edited:
diplomat33
Average guy who loves autonomous vehicles
Perhaps give this a read , I found it interesting. I am not an expert so maybe you know more about this permit issue than I.
Robotaxis Are Everywhere in San Francisco Now. But Something’s Wrong.
Blocking transit. Driving on the sidewalk. Speeding away from cops.slate.com
Yes, I read that article. Put simply, the CA DMV disengagement report only applies to testing, before commercial deployment. Once the AVs get a commercial permit to deploy, they are no longer required to report disengagements like they did before. Cruise has had several embarrassing incidents where the cars stalled and blocked traffic. The concern is that once AV companies like Cruise and Waymo get their commercial permits, they will scale to lots of robotaxis and the robotaxis could cause a lot of traffic problems and since they are no longer required to report disengagements anymore, the public won't know about the incidents. So some are suggesting that companies like Waymo and Cruise should be required to continue to report disengagements even after they get their commercial permit so that the public can see how often these incidents happen and regulators can take action if needed. I think another issue is that city agencies like the SFMTA and the Board of Directors feel left out of the regulatory process since they have no control over the deployment of these AVs. Companies only need to get permits from the CA DMV and CPUC to deploy.
I am not against continued regulations of AVs after commercial deployment. And I am on record as saying Cruise needs to fix those issues. But we would need to change the definition of a "disengagement". That's because the traditional definition of a disengagement does not apply to driverless since there is no safety driver to disengage. A driverless car can stall or block traffic and since the autonomous driving was never disengaged, it would not count as a disengagement. And remote assistance often happens without disengaging the system so that would not count as a disengagement either. So reporting traditional disengagements after commercial deployment would not capture the data properly. We could (and should) require companies like Waymo and Cruise to report safety events that are not disengagements like stalls, accidents, remote assistance, with details on the nature and cause of the event and miles driven for context. And AV companies should be required to publicly document how they are addressing any issues. Furthermore, city agencies like the SFMTA could be given a say so that they don't feel left out of the regulatory process.
Last edited:
diplomat33
Average guy who loves autonomous vehicles
Cruise got their permit to charge before waymo. So yes, cruise has been charging for rides already. But cruise ODD is more limited,, only at night and at speeds below 35 mph. It is usually easier to get a commercial permit for smaller ODD.
Doggydogworld
Active Member
FWIW, I've never seen a driverless Waymo exceed 45 mph. They mostly ceased highway testing a few years ago (except trucking, of course), but recently seem to have picked it back up a bit. With safety drivers, though.
Other than the issue of grouping on a dead end road, the problems seem to almost all be from Cruise (and SFMTA has largely only objected to Cruise, I haven't seen objections to Waymo yet). SFMTA seems to be very frustrated their concerns seemed to have largely fell to deaf ears by the state agencies, so much so they have skipped upwards and complained to the NHTSA (thus the investigation).Perhaps give this a read , I found it interesting. I am not an expert so maybe you know more about this permit issue than I.
Robotaxis Are Everywhere in San Francisco Now. But Something’s Wrong.
Blocking transit. Driving on the sidewalk. Speeding away from cops.
Main concern is without a safety driver (or a remote way to move the car from blocking the road, or at least allow a first responder to command it to move), there are too many incidents of blocking traffic and even incident of blocking emergency vehicles. The other concern brought up is Cruise didn't demonstrate they can consistently pull to a legal pick up spot instead of just double parking by default. From up thread there was a recent update that supposedly improved this, but SFMTA wanted that to be working already before they were allowed commercial operations.
Similar threads
- Article
