[gostel]

Long before we allow or disallow autonomous cars in India, we will have to go through development and testing several times over to make the autonomous car viable on Indian roads. A blanket ban and a quick quip, that is the last response one would expect. At this juncture, the concerned authorities need only to 'allow' research in the field. Just look at how many vehicles on Indian roads, have an auto transmission. In the least, Indian Autocars will have to have special pothole detection and avoidance (oops) abilities, as an example.

[The Brutailer]

Support the decision. The car innovations/trends are basically an anomaly. They are killing what a car is! A car, though originally was intended for a faster means of transport became much more..like an extension of one's self. Well, this site itself is a big example. It works only because of our love for cars.

Auto driving car? What the heck is that! People getting this lazy? Unbelievable. First electric cars, now this.

And let's face it, too much reliance on tech isn't safe. We Indians don't trust even a rudimentary auto park feature, let alone letting the car driving all by itself when senseless beings can pop in on road at anytime.

[civic-sense]

Indian roads will be safer with self driving cars than real people driving. At least there will be some discipline.

And even a robot can do a better job than these dumb ministers.

[sparky@home]

Originally Posted by Nohonking

Hi folks,

Just got this piece of news:



Link to full article

The plea given is, it will take away driver jobs. I think this is reminiscent to once upon a time saying "introduction of computers will take away jobs".

What are your views?

Thanks!

I think its good news, we are a long way off from implementing this technology. When it’s a mad house out there, a driverless car will be a big hazard for the occupants. Most of the roads are just not made for these marvels.

[needforspeed88]

In a nation where driving in no entry is not exactly uncommon in bylanes we are not even prepared to dream about autonomous vehicles on our roads. In a nation where there are still areas where there is no proper road connectivity, it is insane to justify spending the amount of resources needed for improving our road infrastructure to allow plying of autonomous vehicles.

I am not against autonomous vehicles in principle, but for our country, our priority should be on improving our roads and enhancing awareness about safe driving habits than autonomous vehicles.

[T1000]

Originally Posted by pratika

Would I travel in a fully autonomous car. Absolutely.

Would I let my family? Yes, of course.

If a machine is less likely than humans at making erroneous judgements then its fair game.

FWIW most modern airplanes use computers to fly. So most of us have already trusted machines to keep us safe over oceans.

I guess the context of discussion is driverless taxi's.
I restrict my comments to technology and wish to refrain from making political comment.

Im a techie and its interesting to know many here supporting technology.
Driver-less taxis are atleast few decades away, i was only saying, this may not be for the current generation.

[locusjag]

The biggest payoff to be gained by developed countries by switching to driverless cars is the expected savings on human and economic costs associated with accidents. Most accidents in developed countries are due to distracted driving, since road infrastructure and vehicle maintenance are not major contributing factors for accidents in those countries. So, by extension, distracted driving is the need that driverless cars will fulfill. Distracted driving is the one loose end that has not and can not be controlled in the developed countries. For everything else that contributes to accidents significantly in developed countries, there are already a myriad of laws, structures and institutions that all normatively shape road safety.

What do we face in India, on the other hand, that all cause accidents to different extents?
1) Lack of a proper, centralized RTO that issues licenses only to qualified individuals
2) Lack of a points-based centralized database, whereby individuals can eventually lose their license based on their driving record or based on which insurance rates really bite people
3) Different states have their own demotivated and unprofessional traffic police forces that don't enforce rules pertaining to vehicle maintenance or operations. This is why our venerable truckers can drive or park or abandon their malfunctioning heavy vehicles in the center of the highway with no warning lights, brake lights or lights of any kind, or why they can overload their vehicles and chug along slowly on highways.
4) Lack of proper city/town-planning, which leads to clogged roads with poorly planned developments along the roads and lends itself to cross-road movements by organisms and objects of all types and at all times.
5) A burgeoning population with ever-changing migration patterns, and with probably unprecedented population density in many urban pockets!
6) Many different corrupt govt bodies and contractors who build substandard roads and fail to maintain them
7) Stretches of roads sans road markings or stretches where roads cease to exist altogether.

Let me stop right here and ask you - do you really think our drivers can afford to engage in distracted driving?

And as others have pointed out so far, it will probably require computing power and data capture that's orders of magnitude beyond anything available to mankind currently, in order to observe and train machines to cope with all this anarchy on Indian roads. Even safely assuming that the required hardware comes to be (with Moore's law), the cardinal rule in machine learning is, you need to make sure that the machine doesn't find itself in a situation that it cannot anticipate. If it does find itself in such a situation, then there needs to be a handoff to the human operator - and it is at this point of handoff that multiple Tesla owners have crashed and died in the USA, since they didn't pay attention to the car's alerts. One guy was reading a book, another was asleep and another was reaching into the backseat when they crashed and died...

In the Indian context however, there'll simply be too many handoffs and it'll simply drive people to the obvious conclusion that drivers may as well just keep driving the whole time!

I simply don't see the need for driverless cars in India, nor do I see how it'll work.

[T1000]

Originally Posted by T1000 View Post
One accident anywhere in the world is enough to lose trust on the technology.
Self-driving cars are not for the current generation.

Originally Posted by extreme_torque

Look at how far phones have come in the last decade. If ten years ago I would have told you that phones will be as powerful as most computers around and will be gps devices, portable music players, run apps just as standalone computers do, portable cameras that you use all the time and even hear you and respond to you, would you have believed me? I simply speak to my phone now to set an alarm and it doesnt amaze me at all. Its second nature.

Thank you for sharing your views in detail, i appreciate it.
Let me clarify on what i meant,
1. What happened to samsung note 7, just one incident is enough for a recall and sales dropped and samsung had to withdraw the model.
we still use phones but not note 7
2. Same goes with a aircrash, till the reason for a crash is known, the entire fleet is grounded - Concorde is one example, one crash affected the biz and aircraft was permanently grounded.

if one lakh driver-less cars are on the roads and if there is an accident, i expect a similar situation.
One lakh cars will be grounded, till an investigation is done, if the technology is the culprit then it has to be fixed in all 1 lakh cars.

Rebuilding trust on the technology will be a challenge.
Samsung could not do it for note 7, though the issue was with a battery. failure was associated with a brand note 7.
Handling public perception is a challenge

Originally Posted by extreme_torque

Aircrafts fly mostly pilot less after they are airborne, changi airport Singapore already moves people between terminals using driverless trains. There will always be use cases where a human being present in the cabin is not too big a deal but that does not mean that technology cannot conquer it.

My comments were in context of driverless cars and not tech assisted vehicles.

Even after 40/50 years of evolution, auto-pilot is only for cruising not for takeoff/landing, Even at cruising, auto-pilot disengages under extreme weather conditions.
The reason i believe is, lifes of 100's of passengers cannot be given to technology however superior it may be.
Driverless trains between terminals/even driverless metro trains are possible as metro run on a relatively guarded environment without much obstacles.
Metros dont have level crossing and unguarded stretches.

Again the context is driverless car, whats the point of making a driverless taxis and keep a human as a backup. it doesnt make commercial sense at all.
i believe the context of discussion here is drivers are not needed and drivers losing jobs.

Originally Posted by extreme_torque

A self driving car does not mean that it has to be self driving all the time. You can drive it when you are in mood. I would love to have a car which can drive me to work everyday and does not cost me a moon (Tesla 3 will make that happen).

More than fun, an "assisted auto driving car" will be of great help in a bumper to bumper traffic scenario. i let the car drive in heavy traffic.
My comments were in the context of driverless cars.

Originally Posted by extreme_torque

Proof of technology is machine learning technologies, self driving car is an application of this technology.

Very true, Among all the application of machine learning, self driving car concept is the most complex, because of the number of situations/variables it has to handle and handle all in fraction of a sec.
If this technology is mastered, it simplifies the rest of the applications.

One example, A self driving car has to differentiate between a bird flying across the road and a dog crossing the road and take decisions like apply brakes or not , that too in milli seconds. it requires huge computing power and complex algorithms.

Originally Posted by extreme_torque

Most of the advances in the mechanicals of a car trickle down from F1 engineering. You ought to read up more.
http://www.bbc.com/news/technology-12691062

true, I meant F1 engg only.

Thank you

[T1000]

Originally Posted by Lobogris

One accident anywhere in the world and you would lose faith in this technology? Millions of accidents have happened with drivers in cars and yet you trust them? Of course there will be accidents in autonomous cars but the rate will be far less than human driven vehicles. Would I trust my family to be driven by the usual Indian taxi driver? No way. I would prefer a self driven car any day.

Cant compare accidents by humans and machines.
If an accident is because of a bug, and when you know 1000's of cars are running without an update. what will be the reaction? what will be the trust level?
Driverless taxis has to run with stickers like "Melissa bug fixed, roamware updated" or "last update with date".

Have tried to give more details here, request you to check,
http://www.team-bhp.com/forum/indian...ml#post4241198

Note: my comments were in the context of driverless cars , without the need for a human at the controls.

[audioholic]

Originally Posted by T1000

One example, A self driving car has to differentiate between a bird flying across the road and a dog crossing the road and take decisions like apply brakes or not , that too in milli seconds. it requires huge computing power and complex algorithms.

Thats not a valid example. Please read a few articles on sensor fusion. A basic article is here: http://www.eenewseurope.com/news/sen...omous-vehicles

A sensor setup with a Stereo camera, Radar and Lidar can easily distinguish between a dog and a bird unless the bird is as huge as a dog and standing on the road or we have a dog flying across the road as high as the car. Technically speaking, with the current generation of sensors even without a Lidar we already have information related to the dimensions of an object, its boundaries with respect to the car. In addition to this, you also get detailed object classification of things in front of the car such as Car, Bus, Truck, Trailer, Motorbike, Cyclist, Adult, Child and so on. Hence, the decision making system has enough information to alter its decisions based on the type of object that is being considered as an obstacle. With the addition of a Lidar in the coming generation of cars, the information of short range objects will improve further. Hence, its not true that a bird flying across a car will trigger emergency braking. The system is powerful enough to calculate the time and point of impact of that bird with respect to the car. If it is true that the bird would have flown past the car by the time the car reached that point, there will be no reaction from the car at all, except for a audio visual warning. Hence, it is not as simple and easy as going wrong over a bird. These systems do fail, but in scenarios that might have not been thought of at all or in some conditions which would have been difficult to even replicate. That is what is being overcome these days slowly.

[T1000]

Originally Posted by extreme_torque

Most of the advances in the mechanicals of a car trickle down from F1 engineering. You ought to read up more.
http://www.bbc.com/news/technology-12691062

FIA to impose speed limit on F1 drivers after Jules Bianchi crash
https://www.theguardian.com/sport/20...-bianchi-crash

Formula One regulations
https://en.wikipedia.org/wiki/Formula_One_regulations
Its interesting to see engine restrictions like, CC, variables like fuel flow rate, fuel pressure, max rpm are capped to a limit, which indirectly means controlling engine output power.

2013 Engine Restrictions

Size 2.4 litre
Type of Engine V8
Fuel limit per race No limit (around 160 kg)
Fuel flow rate No limit
Fuel injection pressure limit No Limit
Maximum engine RPM limit 18,000

2014 Engine Restrictions

Size 1.6 litre
Type of Engine V6 (Turbocharged)
Fuel limit per race 100 kg
Fuel flow rate 100 kg per hour above 10,500 RPM
Fuel injection pressure limit 500bar
Maximum engine RPM limit 15,000


When technology evolution becomes dangerous to humans, regulators step in and the technology evolution is controlled/stopped.

Though the speed limit in F1 is not about max speed, some design changes to the wings were done to reduce speed during cornering.

Google glass is an amazing piece of technology, why it was killed?

Will be interesting to watch how driverless cars evolve around these problems.

[T1000]

Originally Posted by audioholic

Thats not a valid example. Please read a few articles on sensor fusion. A basic article is here: http://www.eenewseurope.com/news/sen...omous-vehicles

A sensor setup with a Stereo camera, Radar and Lidar can easily distinguish between a dog and a bird unless the bird is as huge as a dog and standing on the road or we have a dog flying across the road as high as the car. Technically speaking, with the current generation of sensors even without a Lidar we already have information related to the dimensions of an object, its boundaries with respect to the car. In addition to this, you also get detailed object classification of things in front of the car such as Car, Bus, Truck, Trailer, Motorbike, Cyclist, Adult, Child and so on. Hence, the decision making system has enough information to alter its decisions based on the type of object that is being considered as an obstacle. With the addition of a Lidar in the coming generation of cars, the information of short range objects will improve further. Hence, its not true that a bird flying across a car will trigger emergency braking. The system is powerful enough to calculate the time and point of impact of that bird with respect to the car. If it is true that the bird would have flown past the car by the time the car reached that point, there will be no reaction from the car at all, except for a audio visual warning. Hence, it is not as simple and easy as going wrong over a bird. These systems do fail, but in scenarios that might have not been thought of at all or in some conditions which would have been difficult to even replicate. That is what is being overcome these days slowly.

Yes, its not impossible. there are multiple sensors/radars to analyse obstacles thats how the prototypes have been working.

It was one simple example to quote the complexities around the system.

there are many companies working on this technology.
some of them have made it work only if the car is run at 5kmph or less.

Another complex use case(Since its a discussion, im stretching it too far, but the use case is possible) is, in the event where an accident is unavoidable
and the situation involves a dog, a pedestrian and 4 occupants of the car or just a material damage.
How will a computer take evasive action?

hence i feel driverless cars are not for the current generation, and it may not be commercially(as taxis) successful .
For it to be commercially successful, the cost of the technology has to be less than the cost of a driver for 5 years atleast.

[audioholic]

Originally Posted by T1000

Another complex use case(Since its a discussion, im stretching it too far, but the use case is possible) is, in the event where an accident is unavoidable
and the situation involves a dog, a pedestrian and 4 occupants of the car or just a material damage.
How will a computer take evasive action?

hence i feel driverless cars are not for the current generation, and it may not be commercially(as taxis) successful .
For it to be commercially successful, the cost of the technology has to be less than the cost of a driver for 5 years atleast.

For the first part, evasive action will be taken if there is a path which the vehicle can take to avoid these obstacles. If there is no way out, then as any human would do, it will slam the brakes as hard as possible. The difference here is most drivers will stop their reaction at the point of slamming the brakes and in some cases steering away. The machine will also do the same since it will continue its assessment and then will be able to change its output.

Coming to the second part here, that's where the main misconception lies.
For the commercial market, driver-less cars are not targeted for personal buyers. They are meant for shared transport systems. Hence, even if the cost of these cars are high, it wont be an issue. The other advancement that is happening for personally owned cars is different from driver-less cars. These cars offer highly automated driving, BUT with the presence of a driver at the wheel. Hence, the system in either case reacts differently and is designed in different ways. Also, these driver less cars are currently being introduced in limited locations only, even if its a foreign country. This rollout will be gradual. Even if it happens in India, it wont offer you a go-anywhere driverless car. Rather it would be a limited place like the CBD of a city, Arterial roads and so on.

[Mafia]

How relevant is the opinion?

To be realistic, the driverless tech is simple too nascent to be viable in a non system traffic situation like our country.

It is decades away from implementation in everyday life for us.

Driverless tech can be implemented in very controlled scenarios like giant software parks, where they can be used for local conveyance.

I don't see this tech taking away jobs in my life time considering the current traffic situation.

[wildsdi5530]

Most of the discussions here are about how erratic the driving conditions are and why driverless cars are not feasible. However I beg to differ. If all the drivers are replaced by driverless cars, where is the question of erratic traffic? Roads will be closed environments. All cars will follow traffic rules. Just watch the scenario in Bangalore- 5 years back, taking a taxi was a luxury. Now there are more taxis than private cars on the roads. Autos are being systematically eradicated. Driverless cars are here to stay. And they are coming sooner than we think. Driving will be relegated to specific roads and become a hobby.
And commercially driverless cars are more viable than a driver. An average taxi needs to be serviced once a month, so software updates shouldn't be a problem. I would peg that by 2020, we will have driverless cars in most of India.