Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

Autonomous Car Progress

This site may earn commission on affiliate links.
I finished
.
"Aug 6, 2023 #CNBC
Trucking is an integral part of the economy, representing over 70 percent of freight moved in the United States. Yet it is dogged by driver shortages, safety issues and supply chain challenges. Pittsburgh-based Aurora is hoping to solve these problem and more by bringing self-driving technology to trucks. While other self-driving companies like Starsky Robotics, Embark and TuSimple have folded or scaled back efforts in the U.S., Aurora is moving ahead and is now delivering loads for customers like Uber Freight and Fedex in Texas. CNBC got an exclusive behind-the-scenes look at its self-driving operation outside of Dallas to see what a driverless future for trucks could look like.

Chapters:
2:54 Ch 1 Aurora
5:58 Ch 2 Self-driving in Texas
9:34 Ch 3 Challenges"
 
I finished
.
"Aug 6, 2023 #CNBC
Trucking is an integral part of the economy, representing over 70 percent of freight moved in the United States. Yet it is dogged by driver shortages, safety issues and supply chain challenges. Pittsburgh-based Aurora is hoping to solve these problem and more by bringing self-driving technology to trucks. While other self-driving companies like Starsky Robotics, Embark and TuSimple have folded or scaled back efforts in the U.S., Aurora is moving ahead and is now delivering loads for customers like Uber Freight and Fedex in Texas. CNBC got an exclusive behind-the-scenes look at its self-driving operation outside of Dallas to see what a driverless future for trucks could look like.
The most interesting thing I heard in that is that Aurora Drive will be sold as a subscription, and they expect it to cost about the same as a human driver. So, the only savings will be fuel, by driving slower, and utilization by being able to drive 24x7. (Though the utilization is somewhat undone by driving slower.)
 
The most interesting thing I heard in that is that Aurora Drive will be sold as a subscription, and they expect it to cost about the same as a human driver. So, the only savings will be fuel, by driving slower, and utilization by being able to drive 24x7. (Though the utilization is somewhat undone by driving slower.)
They'll also sell on safety/liability. I get the feeling that's a much higher cost for metro areas and rural two-lane highways, though, not where these trucks will operate.

24 hour driving is more capital-efficient, and that matters now that cost of capital has risen, but the overwhelming majority of costs are per mile vs. per day. Even team driving, which can get those strawberries from LA to Dallas in day, doesn't cost that much more per mile.

Competition will eventually drive the price down, though. Software wants to be free.
 
The most interesting thing I heard in that is that Aurora Drive will be sold as a subscription, and they expect it to cost about the same as a human driver. So, the only savings will be fuel, by driving slower, and utilization by being able to drive 24x7. (Though the utilization is somewhat undone by driving slower.)
Well that and theoretically there can be "infinite" virtual drivers...and also theoretically the costs could quite easily come down the moment a competitor appears.
 
The most interesting thing I heard in that is that Aurora Drive will be sold as a subscription, and they expect it to cost about the same as a human driver. So, the only savings will be fuel, by driving slower, and utilization by being able to drive 24x7. (Though the utilization is somewhat undone by driving slower.)
as someone that hires truckers....I would pay 20% more for something that shows up when they are supposed to. Literally the least reliable group of workers I have ever dealt with from electricians to IT folks to grocery store employees to landscapers to plumbers. Truckers are the bane of our existence. So I would guess that Aurora has that figured out. Software that shows up. Magnificent.
 
  • Informative
Reactions: Doggydogworld
as someone that hires truckers....I would pay 20% more for something that shows up when they are supposed to. Literally the least reliable group of workers I have ever dealt with from electricians to IT folks to grocery store employees to landscapers to plumbers. Truckers are the bane of our existence. So I would guess that Aurora has that figured out. Software that shows up. Magnificent.
But when AI replaces truckers, it's going to kill a whole genre of country music. Sad.

Unless they also train the network to stream the songs, activate the jukebox at the Waffle House and order products from the radio sponsors.

More collateral economic damage for the AI naysayers to chew on!

Maybe a new song genre will appear:
Robots took my job and shoved it.
Stole my truck, my woman and my self-respect.
My kids say "He don't know nothin', ask the robot".
Dawg Mode take away my last friend in this ol' world...
 
Brad Templeton wrote an article, reporting on the CPUC meeting today about the unexpected stops.

Waymo and Cruise did provide some stats on what they call "vehicle rescue events" or VRE where a human is dispatched to physically help the car.

Cruise offered the following statistics:
  1. They had 177 of these VREs from June 1 to July 18, 26 with a passenger on board.
  2. During that time they drove over 2.1M miles in SF, which they indicated was 10x more than all the other operators in the city (mostly Waymo.)
  3. This rate amounted to one VRE per 79,000 miles of passenger service, and 1 per 11,500 for all types of service, but that this has been steadily improving, and is now 1 per 30,000 miles. There have been no harm to passengers in any incident.
  4. Only 2 of these VREs involved emergency responders according to their notes. In contrast, they noted 168,000 encounters with emergency responders, solving over 98% of them fully autonomously.
  5. The average response time for a VRE rescue was 14 minutes. About 1/3rd involved a vehicle needing to be taken back to depot.
Waymo offered somewhat different calculations
  1. They have driven 3M miles (mostly outside in San Francisco) with 30,000 emergency vehicle encounters. (Waymo was more strict about limiting their reports to events with a passenger on board, because the CPUC authority is strictly over passenger services, not unmanned movement.)
  2. They report 58 “VRE” style events over the last 6 months. They also report that June was 80% less than earlier, implying a strong downward trend.
  3. Most (not quantified) of these events have been fixed and would not happen again in the same situation (presumably tested in sim.)
  4. Average retrieval time for a rescue was 10 minutes, sometimes as low as 2 minutes
  5. They only found 4 cases with 1st responders present, and in no cases where they impeded.

Read it here: Curiosity unlocked.
 
The Truck turned from the right lane because it needed the room. Why didn't the Cruise backup and give the Truck more room to turn. So the car attempted to pass the truck while it was trying to make a wide turn. I put the blame 50/50 on this one. How would an autonomous Semi be trained to handle this situation?
A Cruise employee recently told me (at a publicity event on the streets of San Francisco) that Cruise cabs
don't back up, but that "they're working on it".

A similar situation involving buses, fire engines, etc. making wide turns into streets narrow enough for just one car (it's not rare here) motivates doing something better than "just stop" behavior until the other side gives in.

Backing up is also useful for (1) hapless humans who start going the wrong way down a one-way
street, (2) people attempting a U-turn in the middle of the street who miss the mark only to then convert it into a 3-point turn, or (3) reversing overcreeping into a limited visibility street. (1) & (2) likely are never
initiated by robotaxis, and (3) may be desired by some folks driving a Tesla using FSDb but is manually reversible.
 
Last edited:
  • Informative
Reactions: DDotJ
Here's two recent "local color" autonomous car stories from my hometown.
The story comments section can be enlightening; the first one highlights Waymo,
the second one involves Cruise:


 
Brad Templeton wrote an article, reporting on the CPUC meeting today about the unexpected stops.

Waymo and Cruise did provide some stats on what they call "vehicle rescue events" or VRE where a human is dispatched to physically help the car.

Cruise offered the following statistics:
  1. They had 177 of these VREs from June 1 to July 18, 26 with a passenger on board.
  2. During that time they drove over 2.1M miles in SF, which they indicated was 10x more than all the other operators in the city (mostly Waymo.)
  3. This rate amounted to one VRE per 79,000 miles of passenger service, and 1 per 11,500 for all types of service, but that this has been steadily improving, and is now 1 per 30,000 miles. There have been no harm to passengers in any incident.
  4. Only 2 of these VREs involved emergency responders according to their notes. In contrast, they noted 168,000 encounters with emergency responders, solving over 98% of them fully autonomously.
  5. The average response time for a VRE rescue was 14 minutes. About 1/3rd involved a vehicle needing to be taken back to depot.
Waymo offered somewhat different calculations
  1. They have driven 3M miles (mostly outside in San Francisco) with 30,000 emergency vehicle encounters. (Waymo was more strict about limiting their reports to events with a passenger on board, because the CPUC authority is strictly over passenger services, not unmanned movement.)
  2. They report 58 “VRE” style events over the last 6 months. They also report that June was 80% less than earlier, implying a strong downward trend.
  3. Most (not quantified) of these events have been fixed and would not happen again in the same situation (presumably tested in sim.)
  4. Average retrieval time for a rescue was 10 minutes, sometimes as low as 2 minutes
  5. They only found 4 cases with 1st responders present, and in no cases where they impeded.

Read it here: Curiosity unlocked.
I'll have to take a look when I get a chance. This subject has been on local news lately:
One can imagine similar opposition when more companies want to autonomous testing and operation in SF and other cities...
 
Here's two recent "local color" autonomous car stories from my hometown.
The story comments section can be enlightening; the first one highlights Waymo,
the second one involves Cruise:



Are you saying local news is more favorable to Cruise than to Waymo? The Waymo story seems more negative (focusing on a "stall") whereas the Cruise seems more positive (focusing on Cruise wanting to expand quickly).
 
  • Like
Reactions: loquitur
Cruise says the semi was at-fault
Partially.

What would a human do if on a city street, approaching an 18-wheeler in the right lane, approaching an intersection, truck has its left turn signal on? The correct answer is not "pass on the left", at least around here.

Apparently Cruise does not see or react appropriately to turn signals yet. I'd call this 80% Cruise' fault.
 
  • Like
Reactions: diplomat33
Partially.

What would a human do if on a city street, approaching an 18-wheeler in the right lane, approaching an intersection, truck has its left turn signal on? The correct answer is not "pass on the left", at least around here.

Apparently Cruise does not see or react appropriately to turn signals yet. I'd call this 80% Cruise' fault.

IMO, both vehicles contributed to the accident. The Cruise should not have tried to pass the semi on the left like that. That just puts the Cruise is a difficult spot if the semi does try to make a left turn. But I think the semi was also responsible because it tried to make a really dumb turn and was clearly not paying attention to other vehicles around it. The semi should have seen the Cruise vehicle and known that it was not safe to make that turn at that moment.

On a related note, I think this accident shows why accidents are not as simple as just blaming one party. And AV companies will naturally try to deflect blame by saying that they were not at-fault. This can potentially skew safety data by making the AVs look safer than they really are. We cannot just look at at-fault accidents as a metric for AV safety. Not being at-fault does not necessarily mean safe since there are things the AV could or should have done to mitigate the risk or possibly avoid the accident even if they were not at-fault.

Lastly, I also think this accident exemplifies why semi trucks should not be allowed on city streets. They ae too big and unwieldy IMO. So when they need to make a turn, they cause problems because they take up too much space and often have to use both lanes to make the turn.
 
  • Like
Reactions: cwerdna