Autonomous Vehicles

I will now stop reading or contributing to this thread, as the main contributor is so anti uber and pro waymo that there can never be an open or equal discussion or debate.

Your are correct that I feel strongly about both Uber and Waymo. Based on the information available on both this crash and the way each have developed their AV programmes to date, I would be surprised if you would find many contributors to this or any other respected blog that feel other than pro Waymo and anti Uber. But the field of AV is still evolving and a lot can happen to change one's views. I am open to new developments in this crash investigation and in future developments that each Company takes on the AV journey.

Currently, there are many investigations open as to the exact cause of the fatal crash in Arizona so the legal result is still unknown but in the court of public opinion, there is only one conclusion: Uber could have and should have prevented this accident. Uber is in the proverbial dog house.

We are all saddened by this fatality and for the family of the victim. Sentiment is running highly anti-Uber in the press and in virtually every blog I have read on this subject based on a long Uber history of seemingly flouting the laws. While one can respect the business acumen needed to build its ride hailing taxi service, one can be highly critical of Uber's self-driving, autonomous vehicle programme. It is this programme that I am addressing in this thread.

I believe we would all benefit by your view of Uber and the defense of their AV technology.

I note that for the first time yesterday, the CEO of Waymo commented on the fatal crash. He stated that a Waymo self-driving car "would have been able to handle it" and that Waymo cars are "intensively programmed to avoid such calamities."

According to the Washington Post: "Asked the basis on which Krafcik made his claim, Waymo pointed to its hardware and testing regimen, which includes closed-course testing at night in which pedestrians sometimes unexpectedly enter the roadway, according to the company. The cars operate on public roads and pull data from those situations to create simulations of cities where Waymo operates, such as Phoenix and Austin. Under that setup, Waymo says, the company can repeat a single testing scenario thousands of times."

Comments?

https://www.washingtonpost.com/news...een-able-to-handle-it/?utm_term=.1cdd8281bfa2

and

https://www.forbes.com/sites/alanoh...r-would-have-avoided-pedestrian/#3c6426bf1dbb
 
Having read the story/comments before watching the vid I was a bit surprised, looked like the pedestrian was very hard to see with dipped beams until the car got quite close (I assume it was obeying speed limit), the first part to become visible is the reflective shoes but they might be mistaken for road lines initially(?).
I think there is a good chance the pedestrian would've been hit by a human driver but it does seem weird that the car didn't brake at the last moment once the legs became visible.


Have you seen the video with normal exposure? It is much more typical of city lighting that you would expect. This didn't happen on a unlit country road. https://www.youtube.com/watch?v=1XOVxSCG8u0
 
here's all the uber videos and other car cameras, discussion etc.


only 13miles on average between driver interaction with uber which is rubbish, and wow the safety driver not doing there job, far more time spent looking down probably on a phone.
 
.......

only 13miles on average between driver interaction with uber which is rubbish, and wow the safety driver not doing there job, far more time spent looking down probably on a phone.

@Glaucus. I agree with you that "only 13 miles on average between driver interaction with Uber" is rubbish but will await the final reports, esp from NTSB, to see if this statistic is directly comparable to Waymo's 5600 miles on average between driver interaction as your video, the NY Times article I posted above and other blogs have described. Keep in mind that the Waymo numbers were filed in California (a State requirement) whereas the 13 miles from Uber were never filed publicly but part of an internal memo the NY Times obtained from Uber related sources concerning their programme in Arizona.

For example, we know the Waymo disengagements are for safety related disengagements where the safety driver takes over to prevent an accident. But it does not include situations where the self driving vehicle gets confused in say a construction site requiring the driver to take over, even when there is no danger of an accident. Could the "13 mile" figure from Uber include all disengagements by Uber vehicles rather than just safety disengagements? We do not know yet.

But while we do not yet have this stat we do know that Waymo is now running its Early Rider AV service in Chandler, Arizona without ANY safety driver in the front seat. Passengers hail the service through an app and direct the vehicle to their destination in this Phoenix, Arizona suburb. If their Early Rider AVs (and there are dozens of them operating currently) were stopping unexpectedly in the middle of streets, we would have heard about it by now with the world watching. Clearly Uber was already significantly behind Waymo's efforts in Phoenix before their AV's crash, but according to the NY Times, were desperately trying to impress their CEO in a visit planned for next month and to launch their own commercial ride hailing service by year end.

As your video points out, Waymo CEO stated three days ago that they are confident that their vehicles would have "handled the situation" and avoided the crash. Their CEO has defended this statement based on their testing for this very type of instance in thousands of instances with slightly different variables of pedestrians walking across roads with bicycles.

We also have the CEO of Velodyne's LIDAR group (a leader in LIDAR) stating that they are "confused" as to how, with the LIDAR they supplied Uber, this accident was even possible. In other words, Velodyne are saying their sensors would have detected the pedestrian to give enough time for the car to avoid collision.

And today we hear from the CEO of Mobileye, a self driving vehicle company that Intel bought recently for $ 15 billion, saying that its computer vision system too would have seen the pedestrian and avoided the Uber crash. Mobileye was critical of Uber in describing them as one of a number of new entrants that have not gone through the years of development necessary to ensure safety. Uber began their programme in 2015.

https://uk.reuters.com/article/us-a...een-pedestrian-in-uber-fatality-idUKKBN1H22LM
 
Bloomberg: "Uber disabled Volvo SUV's standard safety system before fatality"

A standard Volvo XC90, the vehicle Uber used that struck the pedestrian, comes equipped with standard collision avoidance technology. However in the specific Volvo used by Uber, this standard equipment was disabled by Uber, according to Aptiv plc, the supplier. Uber supplies its own sensors and software for its AVs. Aptiv clearly wants to separate itself from the Uber crash.

According to Bloomberg: "Aptiv’s radar and camera system using Mobileye chips and sensors helps power the Volvo XC90’s driver-assistance system, which provides collision avoidance, lane-keeping aid and other safety features." But this system only applies to the commercially available XC90, not to Uber's AV.

https://www.bloomberg.com/news/arti...-suv-s-standard-safety-system-before-fatality
 
Well it isn't too surprising that they'd do that, having another system control the car defeats what they're trying to achieve. No doubt uber was expecting their vehicle could stop in such a situation and tragically it has failed.

The safety driver has some questions to answer too tbh..
 
Would love too know how companies are expected to test and develop vehicles when they are not allowed on roads.

Since cars first came on the roads in the late 1800's there have been tens of thousands of deaths, but no one has banned all vehicles from the roads

If a person gets killed by a car driven by another person there is not this level of outrage its just counted as part of life

Should be exactly the same if a computer is driving

In fifty years time when there are thousands of these vehicles on the road, there will still be deaths, I absolutely guarantee it, I just hope by then people add a a bit of common sense into everything.
 
While we await the report from the NTSB and other regulators investigating the fatal Uber crash and consider whether the Governor of Arizona may have seen an early version of the report in making his announcement this morning concerning Uber's "unquestionable failure", what other factors at Uber might have contributed to this fatality? (Or as the video supplied by Glaucus yesterday states, the "perfect storm" suffered by Uber). And whether these factors added such intense pressure on Uber to "win at all costs" the race to creating the most successful AV unit that it severely cut safety corners? Can it ever rebuild trust in AVs on its own now?

There is an Op Ed in the NY Times today that I believe is controversial but raises a few interesting statistics about Uber's current business model that might be considered in answering the question about other factors that influenced events leading to the Uber AV fatal crash.

https://www.nytimes.com/2018/03/26/...-left-region&WT.nav=opinion-c-col-left-region

As it is behind a paywall and a subscription is needed, I provide a few quotes that I found interesting and others might too.

1. "Uber’s business model: Its modus operandi is to subsidize fares and flood streets with its cars to achieve a transportation monopoly. In city after city, this has led to huge increases in traffic congestion, increased carbon emissions and the undermining of public transportation."

2. "Most customers who love Uber don’t realize that the company subsidizes the cost of many rides. This is likely a major factor in Uber’s annual losses surging from 2.8 billion in 2016 to $4.5 billion in 2017. This seemingly nonsensical approach is actually Uber’s effort to use its deep pockets to mount a predatory price war and shut out the competition. That competition is not only taxis and other ride-sharing companies, but public transportation."

3. "Ridership on public transportation is down in nearly every major American city, including New York City (which recorded its first ridership dip since 2009). This is hurting the revenue that public transportation needs to sustain itself. Uber passengers and public transportation users alike now find themselves stuck in heavy traffic for far longer because of what’s been called “Uber congestion.” In Manhattan, there are five times as many ridesharing vehicles as yellow taxis, which has caused average speeds to decline by 15 percent compared with 2010, before Uber."

4. "The company’s new leadership continues to deny that it is contributing to these ill effects. Mr. Khosrowshahi even insists that Uber can help solve congestion by adding a small number of electric cars, and that it could start using flying taxis in five to 10 years (which is preposterous — Uber doesn’t even have a prototype)."

The remaining parts of the Op Ed digress into issues such as regulations requiring a fee for hire to address congestion, prohibiting Uber from fare subsidies, better driver background checks, etc, which while interesting and controversial, do not directly relate to the potential pressure points on Uber in its race to reach AV independence. I do wonder what regulators in London and Newcastle, who are reviewing Uber's license, think of some of these issues?

A loss making business model that exists at Uber creates pressure on them to change the model. The cost of the driver adds pressure on Uber's margins and eliminating the driver could turn the loss into a profit. With its AV fleet now grounded indefinitely, the competition will gain further ground on Uber. Will the indefinite grounding lead Uber's new CEO, who was originally dubious of continuing Uber's self drive efforts, call a halt to their programme entirely? Keep in mind that Uber has invested in significant numbers of AVs on the road already, has a staff of 400 people in their Arizona self drive unit, has staff and operations in other US cities including Pittsburgh and Boston and has placed an order a few months ago to purchase 24,000 specially built (for self drive) Volvos for delivery starting early next year. How much of the $ 70 billion private market valuation for Uber is based on them "winning" the race to autonomous vehicles? How has Uber's desire to launch an IPO next year been affected by the crash?

The new Uber CEO has the unenviable task of assessing whether their Uber self drive programme has been irreparably damaged and can no longer survive as a stand-alone effort. As I have suggested previously, would it not make sense for them to consider a partnership with Waymo, a substantial Uber shareholder, putting Waymo's "driver" in all future Uber self driving vehicles? Negotiations might have already begun as Uber's CEO had already hinted at such a possibility before the fatal Uber accident. With Japan's Softbank now one of Uber's largest shareholders will they add pressure on Uber to enter into such a partnership (they recently arranged the deal announced between Grab and Uber to combine rather than compete in many South East Asian markets after Uber had spent nearly $700 million to create its presence in these markets)? Waymo already has a non-exclusive agreement with Lyft, the number 2 player in the US ride hailing market and has led several rounds of financing for Lyft. Bringing Waymo's "driver" into a partneship with Uber will require negotiations over data sharing, mapping, profit sharing, etc.

One of the more intriguing profit opportunities with a self driving vehicle (esp for Uber with or without a partner like Waymo) will be the opportunity to "sell services" when fleets of cars are connected. With 4G and soon 5G connectivity, and with the vast majority of cars on the road still "unconnected", car manufacturers and fleet owners see the ability to sell services from a platform they create in their vehicles at high margins to interested parties. What services might a manufacturer or fleet owner charge for? Would a transport company or online retailer be interested in shipping on board connected, secure vehicles that operate 24/7 where there is considerable unused space (often hundreds of cubic metres of unused space)? Would a local government or retailer be interested in accessing a crowd-sourced HD map maintained by a network or fleet of vehicles? What would they be prepared to pay for such data? Surely a profitable service could be built around such demand.

Will Uber's CEO make some major strategic decisions soon?

Comments?
 
Last edited:
Uber and every other AV developing companies should be allowed full unfettered access to all roads, at any time of day and night, to test and develop their vehicles in the real world situations they will be working in.

Only then will they ever truly see real world circumstances, and only then can the testers develop solutions to avoid the issues that lead to the unfortunate situation.

You can test as many billions of miles as you want in labs and simulators, once vehicles are on the road deaths WILL happen, there is no avoiding them.

If we shackle companies with unnecessary rules & regulations etc then they will never be able to fully develop systems in real world situations.
 
Surely the cars we drive today have to be tested on the road at some point as they did in the past, and surely by the simple fact that there are a mix of cars, bikes, and pedestrians on the roads and pathways there are always going to be circumstances where people die. I can see where Entai is coming from in quite a few of his posts.
 
Yes but a human test driver can't be affected when a car's systems go wrong or by dodgy coding, they will sense that something isn't right and just stop. Not silently fail and mow down cyclists like this "AI" did, because it can't look at itself and go "hmm, this part is ****ed".
 
Last edited:
Uber and every other AV developing companies should be allowed full unfettered access to all roads, at any time of day and night, to test and develop their vehicles in the real world situations they will be working in..........
If we shackle companies with unnecessary rules & regulations etc then they will never be able to fully develop systems in real world situations.

You will notice that the Governor of Arizona only singled out Uber for suspension, not any other company testing AVs in that State. When the Governor of a State describes your AV efforts as an "unquestioned failure" and there is a Federal and State safety investigation going on with your vehicle that has crashed and caused a fatality, when your competitors have publicly chastised your AV efforts and stated that their computer vision systems would not have resulted in a crash, there is no benefit to telling the Governor that you should have "full and unfettered access to all of the roads" in that State. Instead Uber's response correctly was to accept his decision, cooperate with the investigators, remove their AVs from public roads and keep its head down. An aggressive response by Uber would have been widely chastised. That is the reality of crisis management. And Uber's self driving vehicle programme is in a crisis.

Following my comments about the declining value of Uber's self driving business, I thought others might find this article from Ars today of interest. The article is : "Why it's time for Uber to get out of the self driving car business/Uber's self driving car project would be more stronger outside of Uber"

"More fundamentally, the project needs to earn back the public trust after last week's fatal crash. That's going to be hard to do if the project remains under the corporate umbrella of Uber, a company that doesn't have the best reputation for honesty and respect for the rules. Selling the project to a new owner could give it a fresh start in the public mind."

A new owner, particularly one perceived as honest and truly focused on safety might be a better home for Uber's self driving car project and allow it to test its technology on public roads, which is your point.

https://arstechnica.com/cars/2018/0...ect-is-struggling-the-company-should-sell-it/
 
Just announced: Waymo and Jaguar Team Up with 20,000 car self-driving Fleet/Bloomberg

Full steam ahead at Waymo. Waymo CEO has announced this alliance that will allow them to ramp up self driving vehicles in the next two years.

Waymo is buying the new Jaguar I-Pace, a battery powered hatchback introduced this month by Jaguar and that will be produced in their plant in Austria. Waymo will import them into the US. They will adapt the I-Pace with their own radars, cameras and sensors.

These vehicles will be used by Waymo in their Phoenix ride hailing service that will begin before the end of this year. Waymo intends to launch services in other cities soon. Eventually the Company says that they hope to provide one million rides per day.

https://www.bloomberg.com/news/arti...ar-team-up-with-20-000-car-self-driving-fleet
 
as already said elsewhere

That was a crass statement, unworthy of a CEO. Two things:
  • In business, you lose customers by talking down about your competitors, so the general practice is to talk about what you offer rather than to denigrate someone else - so Krafcik just tagged himself as that other kind of business guy

  • Unless John Krafcik has crash data and ran it through his system to demonstrate how his system would have done something to save the lost life, he needs to STFU - out of a sense of decency
Waymo could do better, if it wanted.

also this

On the other hand - the way that we’re going about it now, as a society, is outrageous and absurd. There’s no top-end government-run program to consolidate research and set safety standards; there’s no demand for companies to share data and information and improvements. It’s the wild goddamn west of technology - where any company that wants a slice of the pie can just start putting their autonomous vehicles on the road, as long as they can afford to cover their minimum legal requirement for financial liability for accidents. The name of the game is winner-take-all, and the potential payout is gigantic, so of course these half-baked arrangements with human blame-absorbers will be preferred: they’re cheaper than spending more money on R&D and top-of-th-line hardware and software....
...

That's not how it works. Per mile driven Uber kills thousands of times more people than human drivers. If there were as many self driving cars as there are normal cars there'd be thousands of times more deaths.

as it suggests what are the safety standards for the antagonists, have either of them volunteered to participate in such standards development.
(in the RAC dazzling lights I drew from a 200page document about adaptive car light qualification ... that is banal, but is there an analogue for AV's ?)
 
Just announced: Waymo and Jaguar Team Up with 20,000 car self-driving Fleet/Bloomberg

Full steam ahead at Waymo. Waymo CEO has announced this alliance that will allow them to ramp up self driving vehicles in the next two years.

Waymo is buying the new Jaguar I-Pace, a battery powered hatchback introduced this month by Jaguar and that will be produced in their plant in Austria. Waymo will import them into the US. They will adapt the I-Pace with their own radars, cameras and sensors, but of note, these changes will be made at the factory level in Austria rather than when the cars arrive in the US. This is in contrast to how Waymo currently adapts the Fiat Chrysler Pacifica minivan---built in Ontario and then adapted by Waymo in the US (incidentally, Uber followed this same practice with its Volvo fleet).

These vehicles will be used by Waymo in their Phoenix ride hailing service that will begin before the end of this year. Waymo intends to launch services in other cities soon. Eventually the Company says that they hope to provide one million rides per day.

https://www.bloomberg.com/news/arti...ar-team-up-with-20-000-car-self-driving-fleet

In a follow up interview, Waymo CEO confirms that what attracted them to the Jaguar I-Pace was the large battery that they fully expect will allow them to drive the car continuously throughout a 24 hour day with only a brief recharge. According to Jaguar Land Rover, "I‑PACE’s 90kWh battery is made up of high energy density lithium-ion pouch cells. Its design and state-of-the-art thermal management system supports longevity and periods of sustained maximum power."

He stated that a fleet of 20,000 self driving I-Pace will be able to provide 1 million rides per day. In other words, each I-Pace will be capable of offering 50 rides per day.

https://www.jaguar.co.uk/jaguar-range/i-pace/index.html
 
Last edited:
Yes but a human test driver can't be affected when a car's systems go wrong or by dodgy coding, they will sense that something isn't right and just stop. Not silently fail and mow down cyclists like this "AI" did, because it can't look at itself and go "hmm, this part is ****ed".

So why are there thousands of deaths on the roads if human drivers will always see what is wrong and stop in time ??
 
Back
Top Bottom