Slow Self-Driving Car Progress Tests Investors’ Patience
Investors are getting impatient with the pace of driverless-car development, pressuring an industry that has had to scale back plans and retrench.
www.wsj.com
-
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
www.wsj.com
diplomat33
Average guy who loves autonomous vehicles
Slow Self-Driving Car Progress Tests Investors’ Patience
Investors are getting impatient with the pace of driverless-car development, pressuring an industry that has had to scale back plans and retrench.
I get that investors who put millions, if not billions, into autonomous driving, want to see a return on their investment. And certainly, if you believed the promises that driverless cars were just a couple years away then I can see how you would get impatient and think the progress has been too slow. Furthermore, we basically misjudged how close we were to solving autonomy. We did not understand the so-called "march of 9's". And perhaps we are still misjudging how close we are to solving autonomy. But the tech has greatly improved now. We have tools now to solve autonomy that we did not have 5-10 years, like faster computer chips, better sensors and more advanced ML. And I think AV companies have matured and understand the challenges of autonomous driving a lot better now than they did a few years back. It would be a shame to give up on driverless cars when we have actually made so much progress. The first autonomous vehicles were incredibly primitive compared to a Cruise AV or Waymo AV today. And we are seeing real driverless cars now in ride-hailing services. And AVs will improve over time.
I think regulators share some part in the "slow" progress. They need to set standards for what is acceptable safety so that AV companies can know if their AVs are ready for public deployment or not. Part of the reason we are seeing "slow" progress is because of liability. The fact is that from a purely technologically point of view, we could deploy driverless cars at scale now if we wanted to. But there would be accidents. We don't know if the accident rate would be "acceptable" or not. Without clear standards for what is acceptable safety, companies don't want to risk losing big in lawsuits if they are found liable. So companies like Waymo and Cruise have developed their own safety frameworks for what they consider "safe enough" and are methodically following that framework. But that will naturally slow down AV deployment. If we had better standards then perhaps companies like Waymo and Cruise could deploy faster, knowing that their AVs are "good enough".
I do think we need to reset our expectations. Driverless cars are not like a smart phone. They are not something you just invent in the lab and then it is done. Driverless cars are much more complex and also require a lot of extensive testing since they are a safety critical product. Any mistake by a driverless car can potentially inure or even kill so it demands rigorous testing before public deployment. So it will take time to deploy AVs. And we should not expect driverless cars to be something that just happens one day. Driverless cars everywhere will not happen over night. Rather, we should expect intermediary products such as so-called L2+, or L3, or L4 geofenced. We should expect driverless cars to start limited and improve over time and deploy in more areas over time.
The truth is that I don't know if it would have really worked to deploy self-driving cars at scale a few years back. The cars would have been too "robotic". It might be good that progress has been "slow" since it has allowed us to make the tech even better. Now, we are working on driverless cars that aim to be good and natural drivers, not just safe. It is a much more difficult problem but when we succeed, driverless cars will be even better, safer, and more accepted by the public.
Last edited:
diplomat33
Average guy who loves autonomous vehicles
Daniel in SD
(supervised)
It would be interesting to see what preceded this. I wonder how they're programmed to handle being stopped in an intersection when the light cycle changes?
diplomat33
Average guy who loves autonomous vehicles
To be fair, traffic does look rather congested. The Cruise on the far right of the screen was waiting for the bus to pass and does seem to move once the bus had passed by. There are some cars on the left side of the screen that were looking to turn right and were blocked by the bus. Those cars were blocking the Cruise that was trying to turn right. And the one Cruise that was mid turn, could not move because of the other car. So I think there were a lot of cars that were blocked, not just Cruise. I am not sure you can entirely blame Cruise for the whole thing.
Do you have some source for what happened before the video in that Tweet? As the way I see that video everything is caused by the Cruise vehicles that are stopped. But without video prior to that there is no way to know for sure.
Not a pretty site. Hard to say but might be a momentary issue that cleared up toward the end of the video. Some of that clears up sooner if Cruise vehicles could back up. It doesn't look like Cruise is programmed to leave space for intersection cross flow traffic to flow.
diplomat33
Average guy who loves autonomous vehicles
No, I do not have a source for what happened before the video. I am providing my opinion based on what I see in the tweet. But when I look at the tweet, I don't think everything was caused by the Cruise vehicles. The bus was blocking the path. It was causing most of the congestion. The Cruise on the far right which I have labeled "1" is waiting for the bus to move over. It must have judged that there was not enough space to move past the bus which was in its lane. So it waits for the bus to move over. Car "3" is trying to turn but is blocked by the bus as well since the bus is in the process of moving over into its lane. Vehicle "3" is cutting in front of Cruise "2". So you cannot blame Cruise "2" since it has a car cutting in front of it and a bus also moving into its lane. It does not have a clear path. But this means that Cruise "2" is blocking the path for vehicle "4". Vehicle "4" is blocking the path for Cruise "5" which in turn is blocking the path for vehicle "6".
Last edited:
The bus wouldn't have been causing any congestion if it didn't have to go into the oncoming lane to get around the stopped Cruise vehicle, #2. That Cruise vehicle is likely the cause of the entire problem, with the other Cruise vehicles making it worse. Cruise vehicle #5 can't make its right turn because it took it excessively wide, and is turning into oncoming vehicles.
But #2 is stopped, holding everything up, probably with its four way flashers on. (But I can't tell for sure.)
diplomat33
Average guy who loves autonomous vehicles
Maybe. But like you said, we don't know what happened before. And yes Cruise "5" took the turn too wide. If it had turned more sharply, it would have made the turn and not blocked the cars behind it. I am not saying the Cruise vehicles are blameless. But since we don't know what happened before, I don't think we can say that it is all the fault of Cruise. And to me, this looks different than the previous cases where Cruise vehicles got stuck. To me, this looks more like a congested traffic case that the Cruise vehicles did not handle correctly, rather than the Cruise vehicles simply getting stuck for no reason and blocking traffic.
The Cruise in the middle of the intersection looks halted and judging by the cars going around it, that seems to be the logical case. #5 made the turn way too wide (could have changed mind too late).
#1 could also be halted also, but hard to tell.
#1 could also be halted also, but hard to tell.
EVNow
Well-Known Member
May be Cruise is not ready for day time driving yet.
Perhaps, if investors are getting impatient and sell, it is time for us to buy.
Cruise will probably congregate wherever there's lots of pedestrians and shopping centers so the potential is there for reoccurrence. It wouldn't hurt to be a little more nimble and able to resolve situations like that or maybe able to call Mr Wizard.... twizzle twazzle twuzzle twon time for this one to come home.
EVNow
Well-Known Member
Also doesn't Cruise operate in denser traffic areas than Waymo ? IIRC, Waymo avoids the most important/busy areas of SanFrancisco.
It is interesting to see Cruise struggle after so many years of concentrated SanFrancisco focused development. And they didn't have to spend a lot of time on perception. The problem is all on planning side.
I wonder how MobilEye with their REM and planning will do here.
BTW, has Omar ever done his rides in very busy areas of SanFrancisco ? I usually see him ride in not so busy areas / times.
diplomat33
Average guy who loves autonomous vehicles
Map of autonomous testing sites in US:
Autonomous testing by road type:
Autonomous testing by vehicle type:
Source:
Autonomous testing by road type:
Autonomous testing by vehicle type:
Source:
diplomat33
Average guy who loves autonomous vehicles
I know people sometimes ask about how AVs handle things when the HD map is wrong. Here is a little video from Aurora that shows how their AV trucks handle when the lane lines are repainted and are different from HD map.
Similar threads
- Article
