[avisidhu]

Originally Posted by amitoj

.............
And speaking of Uber, I have doubts about the longevity of the company itself. Most of its rides are subsidized by venture capital funding
The True Cost of Your Uber Ride Is Much Higher Than You Think
......
Not to mention the recent bad press, even though it was fleeting.

Well, Amazon was in a similar situation not so long back no? This particular business anyway will be very tough to replicate globally by any competitor, once Uber starts charging more. They are out of red in many countries already, and even the fares back home have gone up substantially.

Add to it the wealth of rider information already obtained by them, and the company looks more robust than what might appear from the surface. Besides, such high valuation has to have some logic behind it.

[Joxster]

Uber's self-driving fleet suspended again, this time over an accident that kills a pedestrian crossing the road.

An Uber vehicle operating in autonomous mode in Tempe, Arizona in the United States, struck and killed a pedestrian crossing the street on Sunday night.
The Tempe Police Department said in a statement that the vehicle, which had a human operator behind the wheel, hit 49-year-old Elaine Herzberg as she was crossing the street outside the crosswalk. She was transported to the hospital and later died from the injuries.
Uber has suspended pilots in all cities following the deadly accident, and CEO Dara Khosrowshahi said in a tweet that the company was cooperating with police in its investigation.

The U.S. National Transportation Safety Board also said in a tweet that it is also investigating the accident.

The ride-hailing company began testing its autonomous vehicles in Arizona in February 2017. One of its vehicles in autonomous mode crashed in March 2017, shortly after the pilot began, and the company briefly suspended testing. Uber vehicles returned to the road after an investigation found it was not at fault.

First fatality of its kind

The accident is the first reported death caused by a vehicle intended to be fully self-driving. In May 2016, Joshua Brown was killed when his Tesla Model S crashed into a semi-truck while operating in Autopilot. The system is semi-autonomous, and an NTSB investigation found that Brown was using the technology outside of its capabilities.

Though Herzberg was not in the crosswalk at the time of the accident, self-driving safety advocates say her position should not put her at fault.
"'Crossing outside of the sidewalk' was never a valid excuse for traffic deaths, and it provides no cover for autonomous mobility companies," Janette Sadik-Khan, former New York City Department of Transportation commissioner and a transportation consultant, tweeted.

The accident will likely affect all manufacturers developing autonomous vehicle technology, and the results of the investigation could influence consumer acceptance, experts say.
"This has the potential to severely impact public perceptions of autonomous technology, and should be handled with utmost prudence by regulators, authorities, and the industry alike,” said Akshay Anand, an analyst at Kelley Blue Book, in an email.
Further fatalities or an inadequate response to Sunday's accident could deal a crushing blow to years of autonomous vehicle development.
"If we have more tragedies like the one we saw in Arizona, the public is going to recoil from this technology and are not going to get in these cars, and any safety benefits that could be realized will be lost," said Peter Kurdock, director of regulatory affairs at the Advocates for Highway and Auto Safety. "Certain segments of the industry that have placed rush to market over making sure these things are safe."

The crash also comes as Congress considers a bill that would allow manufacturers to deploy tens of thousands of autonomous vehicles without meeting certain federal safety requirements. The legislation is currently pending as five Democratic senators have placed a hold on it.
"This tragic accident underscores why we need to be exceptionally cautious when testing and deploying autonomous vehicle technologies on public roads," said U.S. Sen. Edward Markey D-Mass., one of the legislators blocking the bill, in a statement. “If these technologies are to reap their purported safety, efficiency, and environmental benefits, we must have robust safety, cybersecurity, and privacy rules in place before these vehicles are traveling our roadways to prevent such tragedies from occurring."

Volvo Cars, which supplies XC90 crossovers to Uber and has been working with the company on integrating the technology into the vehicles, said in a statement that it was notified of the crash.
“We are aware of this incident and our thoughts are with the family of the woman involved,” said a Volvo spokesman. “We are aware that Uber is cooperating with local authorities in their investigation.”

Legislative action

On Friday, a large coalition of public interest groups urged Secretary of Transportation Elaine Chao to toughen what it called “weak and unenforceable” voluntary guidelines for autonomous vehicle developers.

Revised guidelines issued by the Trump administration last summer eliminated requirements for automakers to seek regulatory approval before launching autonomous technology on roadways and urged states to leave motor vehicle performance requirements to the federal government. Automakers and technology companies do not need to include “unnecessary” design elements, such as steering wheels, from safety self-assessments filed with NHTSA, the nation’s motor vehicle safety regulator and a DOT sub-agency.

The self-assessment letters are a way for companies to publicly disclose information about autonomous technology undergoing testing.

At the time, critics faulted the proposed guidelines for not requiring more information from automakers about autonomous vehicle systems and requiring performance standards for the software and sensors controlling the vehicles.

“The Department and ... NHTSA have chosen to be detached spectators instead of engaged safety regulators during one of the most crucial and critical times in the history of automobiles,” more than two dozen groups and individuals said in a letter to Chao.

“We strongly believe that DOT has a legal responsibility to evaluate and regulate technologies that are ‘safety winners or losers’ before they even enter the marketplace. This is the most effective and assured approach to prevent unproven and potentially dangerous technologies from being sold to the public and allowed on public streets and highways across the country,” the coalition said.

The Takata airbag recall crisis, General Motor’s faulty ignition switch, Toyota's acceleration scandals and Volkswagen diesel emissions cheating demonstrate why the industry shouldn’t have a free hand with autonomous vehicle deployment, the safety advocates said.

Source

Cheers

[volkman10]

Toyota temporarily suspends U.S. public AV testing after Uber fatality.

Toyota Motor Corp. said it was temporarily suspending public road testing of its advanced autonomous driving technology after a self-driving Uber test car killed an Arizona pedestrian in the first known fatality involving a fully autonomous vehicle.

The pause in testing affects vehicles operating in chauffeur mode — Toyota's in-house term for fully autonomous driving.

The decision suspends only operations in the U.S. and was ordered by the Toyota Research Institute, the Silicon Valley-based unit researching autonomous driving and robots. Toyota did not say when it would resume public road tests.


Link

[arunphilip]

Video of the crash as released by the Arizona PD. Not to worry, the released footage pauses before impact.

The interesting thing to note is that the pedestrian had already crossed an entire lane (the one on the left) and crossed the dividing line into the right lane. It is therefore interesting that the non-visual sensors (LIDAR, RADAR, etc.) did not detect this at all, and that too on a straight stretch of road. It's not like they popped out from behind a bush or something.

Looks like we had a distracted safety driver as well.

https://www.youtube.com/watch?v=pO9iRUx5wmM

[speedmiester]

Originally Posted by arunphilip

Video of the crash as released by the Arizona PD. Not to worry, the released footage pauses before impact.

The interesting thing to note is that the pedestrian had already crossed an entire lane (the one on the left) and crossed the dividing line into the right lane. It is therefore interesting that the non-visual sensors (LIDAR, RADAR, etc.) did not detect this at all, and that too on a straight stretch of road. It's not like they popped out from behind a bush or something.

Looks like we had a distracted safety driver as well.

The car was on low beams and understandable if it was a human driver behind the wheel.
But autonomous cars are fitted with sensors to pick up movement through non visual sensors. The car did not apply brakes at all.
This is a definite fail and will have long standing repercussion due to this accident.

[amitoj]

Originally Posted by arunphilip

Video of the crash as released by the Arizona PD. Not to worry, the released footage pauses before impact.

The interesting thing to note is that the pedestrian had already crossed an entire lane (the one on the left) and crossed the dividing line into the right lane. It is therefore interesting that the non-visual sensors (LIDAR, RADAR, etc.) did not detect this at all, and that too on a straight stretch of road. It's not like they popped out from behind a bush or something.

Looks like we had a distracted safety driver as well.

Originally Posted by speedmiester

This is a definite fail and will have long standing repercussion due to this accident.

After seeing the video, I have to agree that this was a failure of the autonomous car, and the safety driver as well. The car should have been able to not only detect the pedestrian, but also extrapolate the pedestrian's travel into its own path. If an autonomous car can't even do that, then what are we talking about??

The distracted driver highlights another huge implication for autonomous vehicles. That is, that humans can't be trusted to make up for what the vehicle fails to notice. When even regular car drivers are so distracted, what hope can we have from drivers in autonomous cars?

[arunphilip]

Originally Posted by amitoj

The distracted driver highlights another huge implication for autonomous vehicles. That is, that humans can't be trusted to make up for what the vehicle fails to notice. When even regular car drivers are so distracted, what hope can we have from drivers in autonomous cars?

Very true.

There's this saying "Go big, or go home". Similar to that, self-driving cars would probably have to reach SAE Autonomy Levels 4 or 5 for them to be rolled out safely for typical car buyers.

Distracted driving is already a problem today, and this might only get worse with any form of partial automation (especially when trumpeted as automated driving or self-driving and not an additional safety measure like proactive braking).

SAE Autonomy Levels 3 and lower do expect an engaged and aware driver, and it would not be safe to put that form of reliance on a regular driver.

There is a real risk that something will slip in the crack between a car that couldn't handle it, and a driver who wasn't situationally aware to react in a timely manner. It will turn out similar to the story of somebody, everybody, nobody and anybody.

[srishiva]

I am not sure the driver would have been able to avoid the accident even if we as aware as a active driver.
But then we wanted to avoid such accidents using autonomous driving, right ?
I dont see it helping the driver because it will be too late when the driver realizes something is wrong (he is not going to be 100% involved similar to manual driving)

I can see a complete environment controlled where it can work. probably with walls on either side of roads and all vehicles able to communicate with each other. Also, I dont see any veering off by the car. So its binary, everything or nothing.

[amitoj]

Originally Posted by srishiva

I am not sure the driver would have been able to avoid the accident even if we as aware as a active driver.

The low resolution camera probably does not give a true picture of what the human eye can see.

[benbsb29]

Originally Posted by srishiva

I dont see it helping the driver because it will be too late when the driver realizes something is wrong (he is not going to be 100% involved similar to manual driving)

I agree. What is bound to happen is that the driver would remain in a passive state of concentration, and by the time he/she realizes that the machine has failed to react as appropriate, it's too late, and all you can do is brace for impact. There is no way a human can react super quick under those circumstances.

It's kinda similar to how people try to slow down on seeing a speed camera, but the reality is that by then the speed has already been determined by the camera and evidence collected. Slowing down based on human reaction has effectively failed to serve any purpose at that point.