[blackwasp]

Waymo - the self-driving car unit of Alphabet Inc., has started public trials of their autonomous cars in Phoenix, Arizona.



As part of the early rider program, residents of the Phoenix Metropolitan Region will be able to summon the cars through an app. For now, users won't be charged for using this service. Waymo plans to use this program to understand the way people outside their test facilities will use these vehicles. The company is studying user experiences and usage patterns to estimate the potential revenue for such a service in the future. It plans to run up to 500 such vehicles as part of this program.

The company is using the Chrysler Pacifica minivan for this program. So far, only testers and contractors have been able to experience such autonomous cars on the public roads. During the earlier test runs, a driver (operator) was always seated in the driver's seat. However, with the launch in Phoenix, there will be no one at the wheel.

According to the company, since 2009, when Waymo was part of Google, it has done over 3.5 million miles (56 lakh km) of testing on public roads in 20 cities combined, along with additional testing of over 20,000 individual scenarios at its testing ground.

https://www.youtube.com/watch?v=aaOB-ErYq6Y

Link to Team-BHP News

[nishsingh]

Waymo has already been testing these on public roads without any drivers for close to a month now!

Waymo started this mid-October, and given that there is no news around this, it seems that the exercise has been pretty successful. (Linux nerds will remember the motto - No news is Good news!)

Waymo intends to invite people for rides as well. Waymo intends to start doing this shortlywithin a 100-square-mile area of the town of Chandler, a suburb of Phoenix. The next step according to Waymo's CEO John Krafcik is a commercial ride-hail service, in which riders can hail one of the company’s autonomous minivans via an app like Uber or Lyft. “People will get to use our fleet of on-demand vehicles, to do anything from commute to work, get home from a night out, or run errands”.


Source: https://www.theverge.com/2017/11/7/1...ler-autonomous

[Jeroen]

Within one hour of service an accident:

https://www.theverge.com/2017/11/8/1...dent-first-day

Looks like it was the other car's fault or rather the driver of that car.

Jeroen

[SmartCat]

I wonder if these driver-less cars can detect left/right indicator or hazard light of a human controlled car.

And in a hypothetical city with 100% driver-less cars, I wonder if indicators are still needed!

[sushantr5]

Now the real world testing will be really difficult phase because fellow human road users will try to fool these cars. Some examples, when the other human driver sees the self driving car then, he will try to fool it by sudden turning in closure. Humans will try tactics of fearing other which are being used on other drivers. Some bad guys will try deliberately bringing these cars into accidents because there are really high stakes involved.

So this will be real testing, real negative testing. Testers are humans and human tendency will help in bringing better self driving cars. If this is successful then we will see sudden acceptance to new technology.

Because to run a cab company, a car comes at €500 per month leasing cost and driver comes at approx €3000 per month for 8 hours of uptime. Imagine this for 24 hours of runtime in these developed countries in Europe and US. Self drive car will come probably at €1000 per month leasing cost and thats it. This is sufficient reason that will bring massive acceptance to this self drive technology.

You will see self drive cab companies really sooner than expected. Whole business dynamics will change suddenly. Car ownership concept will die suddenly because hiring cabs will become really really cheap compared to today. I think in these 3-4 years lot is going to change, really lot.

India will not be far behind.

[PetaWatt]

I believe Google got enough money to develop the car from scratch. Then why there is steering wheel in the car? Lack of confidence?

It will take a while for the final product to hit the market. Automated cars will have to perform better than the humans and will have to show clear advantages. On top of that, evolution will be gradual, so these cars will have to live with the cars driven by smarter humans for some time. Hence, they will have to understand a few things when humans are on the driving seat of the other cars on road:

• Speed depends on the mood
• Indicators are given depending on the mood
• Car can be parked even in the no parking area
• Lanes are only for self conscious cars

This list is longer than this, but I think I indicated what I mean to say.

So, driving along with humans will not only need few sensors, controllers and rule based algorithms but will also need artificial intelligence to understand human and hence their driving. I am not sure at what stage Google is now with this development, but steering wheel in an autonomous car is not a good indication.

[sushantr5]

Originally Posted by PetaWatt

I believe Google got enough money to develop the car from scratch. Then why there is steering wheel in the car? Lack of confidence?

It will take a while for the final product to hit the market. Automated cars will have to perform better than the humans and will have to show clear advantages. On top of that, evolution will be gradual, so these cars will have to live with the cars driven by smarter humans for some time. Hence, they will have to understand a few things when humans are on the driving seat of the other cars on road:

• Speed depends on the mood
• Indicators are given depending on the mood
• Car can be parked even in the no parking area
• Lanes are only for self conscious cars

This list is longer than this, but I think I indicated what I mean to say.

So, driving along with humans will not only need few sensors, controllers and rule based algorithms but will also need artificial intelligence to understand human and hence their driving. I am not sure at what stage Google is now with this development, but steering wheel in an autonomous car is not a good indication.

I am sure it has functional artificial intelligence which has following advantages
1. Accuracy: It can estimate accurately distance, trajectory of self and multiple surrounding objects and activities.

2. Response Time: It will have far superior response time in the case of unforeseen events and sudden erratical movement of others. Humans have this as a biggest disadvantage. We take almost 0.7 seconds to realize things and almost 2 seconds to act based on what eye sees. In case of automonous cars it will not be exactly realtime but almost realtime realization and response.

3. Understanding: AI will be more capable to understand the capability of car it is controlling. It has correct idea of speed, brakes, momentum, stopping distance, brake force and more important it will have better watch on multiple objects on the road. It will not miss any road sign. It will not divert attention. It will not be tired. It will not be angry. It will not miss running boy on the side of road, it will not miss bicyclist travelling by the road. heck, it will not miss the crow flying on its head and it will be definitely capable of saving itself from shit of crow .

4. Database: It will have access to database of everything, road signs, risks, density, weather, road condition. It will have statistics for deciding strategy for speed, evasion and much more.

All these things make Autonomous cars more superior to human being when completely developed. As we say today monkeys are unfit to drive cars because humans are superior, tomorrow autonomous car will think(if it gets emotions) humans are unfit to drive cars. That day is not too far.

If you think situation from India perspective, it is distant dream. But if you see driving patterns/rules and general sense of US Europe people (at least German which I see) we feel that it should have been possible long time back. Why so late?

[Durango Dude]

You can have Driver Less cars or Driver Driven Cars but it seems both can't coexist in the same environment as it would be a recipe for accidents (minor and major). Bots and humans don't think alike, it seems.

[sushantr5]

Originally Posted by Durango Dude

You can have Driver Less cars or Driver Driven Cars but it seems both can't coexist in the same environment as it would be a recipe for accidents (minor and major). Bots and humans don't think alike, it seems.

Definitely they can co-exist but not in India. Now autonomous cars are confined to boundary of 100 sq km but it will increase substantially and exponentially. Economics will enable this scale.

And Level 2 autonomy can achieve these results about autopilot in Tesla then you can imagine what will Level 4 in Waymo's case.

https://www.youtube.com/watch?v=Ndeb1pMAsh4

[NetfreakBombay]

Originally Posted by smartcat

I wonder if these driver-less cars can detect left/right indicator or hazard light of a human controlled car.

Yes, almost all major car platforms (level 3, 4) now do this. Waymo's patent numbers are US20170039435 A1 , US9690297 B1.

This is one of "assistance" to drive in snow that covers lane markings. Vision systems try to locate lanes by observing rear lights, indicators of other cars (and marry this with gyro data for validation).

Nvidia and Baidu's platforms have good documentation on this.

https://www.autoblog.com/2016/04/11/...tonomous-cars/

[Technocrat]

Originally Posted by PetaWatt

I believe Google got enough money to develop the car from scratch. Then why there is steering wheel in the car? Lack of confidence?

I would say its more of a fallback option, its always desirable to have a manual override for a key function.

Most of the Modern Airplanes these days can fly on their own yet they have a Yoke (steering wheel)

[Holyghost]

Originally Posted by Technocrat

I would say its more of a fallback option, its always desirable to have a manual override for a key function.

Most of the Modern Airplanes these days can fly on their own yet they have a Yoke (steering wheel)

Most of the Modern Airplanes these days can fly on their own yet they have a Yoke and a Pilot too for passenger aircrafts. Though we have to wait and see how long the pilot will be there behind the yoke.

[A M]

Originally Posted by Technocrat

I would say its more of a fallback option, its always desirable to have a manual override for a key function.

Most of the Modern Airplanes these days can fly on their own yet they have a Yoke (steering wheel)

Reaction times massively increase if you are not paying enough attention. By the time humans will realize that they have to take over from the car, the accident would already have happened. And it is very difficult to keep paying attention while you are only a spectator to whatever the car would be doing.

[Technocrat]

Originally Posted by A M

Reaction times massively increase if you are not paying enough attention. By the time humans will realize that they have to take over from the car, the accident would already have happened. And it is very difficult to keep paying attention while you are only a spectator to whatever the car would be doing.

That is just one of the fallback use cases, there are many others, like ability to steer in loss of power, ability to steer in case of failure of steering motor etc, i am sure there are many more.

[PetaWatt]

Originally Posted by Technocrat

Most of the Modern Airplanes these days can fly on their own yet they have a Yoke (steering wheel)

And there is a trained pilot behind the Yoke. In a similar way, if a driver is needed in an autonomous car, can it be really called autonomous car? I think it will be more accurate to call it car with assistive driving functionality.

My point is, presence of steering wheel indicates technology is yet not enough fool proof to be used in day to day life without involving a human.