Do you think that autonomous cars will decrease the accident rate?

[deehunk]

Originally Posted by ashis89

I voted No.

Working in IT sector, I know that a software can't be flawless and there's always going to be a bug or downtime.

So our roads will have computer controlled pre-programmed cars, old school man-controlled cars (helmed by sleep-deprived/drunken drivers), some stray cattle, bikers and a few random people crossing the roads.

Exactly, they will be like unwanted Pop-ups who cannot be blocked at the click of a button. I have setup test rigs in labs and have a fair knowledge of how these autonomous cars will be built. The cost and nature of DAQ required to build a flawless system in each car will be very expensive and complex.

[centaur]

Originally Posted by StarrySky

However, so it can't notice/predict cars coming into my lane and can only react once a car is in the radar range. So, I can react to lane changes by other cars faster than the system in my car. But then, mine is not an autonomous car. An autonomous car should already be able to detect such events with its sophisticated radar/camera system.

Definitely agree they might be faster now than before but still they are not as good as humans when it comes to making split second decision IMHO. For your last statement in the quoted post above you may want to read this.

Tesla crash happened in exactly the same scenario that you mention when a truck was moving into the lane in which the autonomous car was driving. It can be written off as a one off case but the point being there still is a chance of it happening. Probably it would get ironed out in the future but I would still say that no matter how top notch the software is, if there is no discipline by other road users, it may not be able to help much.

[roy_libran]

I voted "Yes".

Scenario 1:
One Autonomous vehicle, in a melee of thousands of Human Driven vehicles, would perhaps have a horrendous time, keeping up with the numerous ways in which humans can (mis)behave.
For the Autonomous vehicle to really succeed in this scenario, and especially in a country like India, would require the artificial intelligence to be highly matured and backed with enormous amounts of computing power, since you really wouldn't want the hourglass symbol coming up when a lorry is about to jacknife you and there are kids stranded in front of your vehicle.
This is symbolic of situations where humans are usually better and/or resilient, and may continue to be.

Scenario 2:
A few human driven vehicles, in a melee of thousands of Autonomous vehicles. This is where the majority of nodes, behave as they should, so that the bulk of computing power at your disposal, is available to account for random events induced by the few humans.
Here, the autonomous vehicles, with the technology of TODAY, will thrive, and hugely bring down the number of accidents, and this can happen TODAY!



So, since the question in the POLL, asked us to consider the potential to avert accidents using this technology, my answer would be YES, and it is possible TODAY.
All it needs is to make the substantial majority of vehicles on our roads Autonomous quickly, and the immense benefit will follow.

Of course, we wouldn't like it as much, would we?

[GTO]

Voted for 'can't say' as autonomous cars in India are a long, long time away.

A smart BHPian (don't remember who) had posted on another thread that autonomous & human-driven cars cannot co-exist on the road, especially in India with its poor road manners. If the roads have 100% autonomous cars, then perhaps, yes - they will reduce road accidents. But again, it's going to be decades before we see that here.

One basic thing that self-driving cars need are lane markers. Even in the commercial capital of India, most roads don't have them.

[IshaanIan]

Having interned at a company at the forefront of developing autonomous driving solutions, I have voted 'yes'. As much as I hate how the future of cars is looking, there is no denying that autonomous driving will decrease time spent in traffic (if all cars can sync up and start as soon as light goes green, instead of waiting for the car ahead to start first, that itself will hugely help traffic situations), will be more reliable (manufacturers can get millions of kilometers of data about cars and check for patterns and failure rates more accurately thereby increasing the reliability of cars), and more safe (autonomous cars would not enter a road without looking or be oblivious to lane discipline and right of way rules).

To all those working in the IT sector, let me remind you that autonomous driving has still not yet been fully achieved. Level 5 autonomous driving is not even being tested out right now. There are industry wide standards being set to prevent any danger in case of failure or malfunction of components. Some of the rules dictated by the autosar standard which manufacturers have to abide by, dictate that each component would require its own sensor to monitor its functions and that backloops are created for every task in the case of critical failure. Apart from the worry of hacking, this is no joke and things are being taken very seriously.

[KarthikK]

Voted for No.
The endless combinations of situations which lead to accidents and incidents on the road simply make it impossible to code all the (avoiding) intelligence into the controlling software.

Maybe in a country with totally predictable vehicular/driver behaviour and where infrastructure can be taken for granted, the accidents can be decreased by self-driven cars where human error can be overcome.

In our country we are not only plagued with too many variables - rogue drivers/riders with no lane discipline, rage-filled cabbies, helmetless moped riders, tractors and rickshaw wallas, we also have to deal with cattle, livestock and dogs jumping across the expressway dividers at any point, unmarked gigantic speedbreakers and barricades, dividers appearing abruptly, massive potholes and the list goes on. I would rather bet on experienced human intervention in stopping/slowing a vehicle than letting some computer algorithm flip my car over (in a pre-coded panic move) for every silly reason like a sheep jumping across the car

[libranof1987]

In a perfect world, yes.
In the realistic world, no. Not yet.

In developed countries, may be.
In India, there won't be autonomous cars in the foreseeable future.

There are so many unknown variables in driving that we take for granted because of human intelligence: we "notice" things, we learn from experiences, have intuitions. I wonder if AI is developed on the mass scale such that it can replicate the human brain.

Let's take the example of airplanes: it was only after significant accidents that resulted in tremendous loss of life that the world learnt lessons that we take for granted today. For eg. TCAS (Traffic Collision Avoidance System) developement was rushed following a mid-air plane collision over the Grand Canyon.

Solutions to problems are found typically only when the existence of problems is realized.

All said and done, I'd imagine the learning curve to be really steep. A friend has traveled in a self-driving Uber quite a few times already in Arizona.

[deehunk]

Originally Posted by IshaanIan

Having interned at a company at the forefront of developing autonomous driving solutions, I have voted 'yes'.

To all those working in the IT sector, let me remind you that autonomous driving has still not yet been fully achieved. Level 5 autonomous driving is not even being tested out right now. Apart from the worry of hacking, this is no joke and things are being taken very seriously.

Good to hear about Level 5 term, since I had an opportunity to work with both hardware and software components simultaneously and later went through the pain of interlinking them, I agree with your statement. Is it not true that all inventions were made with serious intentions? I want to understand the risk of the monopoly of technology by few companies if they ever achieve it and its adaptability to a diverse global market.

[Engine_Roars]

If the cars are automated, they can be hacked and if they can be hacked, we have many nut cracks roaming around who wish to kill people. I am all for automation but something here makes me uncomfortable. Call me old school but I am not liking all this advancement in terms of technology. In this race no matter which organization wins, the ultimate looser will be humanity.

I just got google home mini as a gift, and I found myself talking to it when alone at home. Once I was done talking, I could not wrap my head around it. I was talking to a device programmed to talk by humans.

Done with the rant, gotta go, my google home is waiting for my reply.

[StarrySky]

Originally Posted by centaur

Tesla crash happened in exactly the same scenario that you mention when a truck was moving into the lane in which the autonomous car was driving.

I think the Tesla crash is a bad example. Even though Tesla calls it "Autopilot", it is not a self-driving/autonomous car but an advanced driving assist system, which requires the driver to be paying attention. It is, of course, an issue that "the system gave far too much leeway to the driver to divert his attention to something other than driving" - based on comments from NTSB chairman.

A self-driving car would be the likes of those being tested by Waymo, which claims to recognize even hand signals from cyclists. The fact that these are still being tested means that they are not yet ready to be released to the market. So I agree that the self-driving systems remain to be further improved, thoroughly tested and proven.

As a side note, from the Tesla crash report:

The NHTSA also found that crash rates for Tesla vehicles had dropped by 40% since Autopilot was first installed.

[lamborghini]

I voted yes, primarily due to response times and field of view.

A car with sensors will be able to detect what could be in the driver's blind spot. Just for example, the dashcam on my car has a wider angle of view than what I can see at a given moment - such cameras/radars would cover these blindspots at all given times!

Secondly, the response time would be quicker and it would leave more margin for error than a human driver - look at adaptive cruise control technology - a boon on the highways when the car in front reduces it's speed, and Volvo's city safe as an example which auto brakes when it feels it is close to a collision.

I got rear ended once on the highway because I slowed down to a halt when the car in front of me did (traffic), but the car following me was slow to realise it and didn't maintain an adequate distance and crashed into my bumper.

With autonomous cars, the vehicle behind would have maintained a safe distance and there would be no room for driver error, and with proper radars - it would notice the chap trying to cut in before an average driver probably could.

But at the same time, I would say the technology needs another few years to be fully developed and reliable given the chaos that is our roads.

[Coolman]

Voted "Can't say"

Here's what I feel!

The developed nations possibly have an edge when compared to the developing ones. Is that edge sustainable? A very difficult proposition!

• To have autonomous cars go the distance, we are essentially betting on replacing what the humans have been doing for years - all that experience, all the knowledge, the decision making abilities - translate into thousands of test cases, if not more. Keep aside building the test cases, it would be quite a challenge for anyone to possibly understand all of them.
  
  
• Every case that isn't built makes the system less autonomous and what % of that is really acceptable? How safe would the system then claim to be? Potholes, "take a detour", "men at work", anticipatory braking, objects falling from the top are a few that I found intriguing!
  
  
• Assuming that the system is at a point where it is deemed fit to be termed "autonomous", will the artificial intelligence built in, be equipped enough to help the system maintain, evolve and sustain?
  

Autonomous and maybe 100% - there may be implementations in controlled environments - like a vending machine or an assembly line etc. OR even robotics! But I haven't honestly seen "autonomous" systems that are 100% in other areas which are possibly less complex than "cars on road", and so I am not able to convince myself that "autonomous" cars will do the job for us!

Related Link: http://www.team-bhp.com/forum/intern...ml#post4171684

Thanks,
C_

[ashis89]

Originally Posted by GTO

autonomous & human-driven cars cannot co-exist on the road, especially in India with its poor road manners. If the roads have 100% autonomous cars, then perhaps, yes - they will reduce road accidents. But again, it's going to be decades before we see that here.

Exactly what I had tried to explain in my post earlier. Different 'systems' co-existing on the same road will lead to utter chaos. And autonomous car could be be a success if Govt. brings in a blanket ban on all the existing man-driven vehicles, forcing everyone to buy autonomous ones all at once. For a country like India, this sounds like a dream.

[deehunk]

Originally Posted by ashis89

Exactly what I had tried to explain in my post earlier. Different 'systems' co-existing on the same road will lead to utter chaos.

Why different systems? The difference within the system itself will create chaos. Handling interference of signals and prioritizing them in real time is not as easy as it is perceived. Having an engine start-stop button is a great feature, restricting it to top variants makes the car luxury but using fingers to switch it ON is still the fundamental reality.

[ashis89]

Originally Posted by deehunk

Why different systems? The difference within the system itself will create chaos. Handling interference of signals and prioritizing them in real time is not as easy as it is perceived. Having an engine start-stop button is a great feature, restricting it to top variants makes the car luxury but using fingers to switch it ON is still the fundamental reality.

Sorry for not making it clear. By different systems I meant, 1 group of autonomous cars, 1 group of human driven cars, 1 group of riders, 1 group of pedestrians who can jump onto the road anywhere and lastly, the cattle.

Since each of them have a mind of their own, the programs on the autonomous car would be lost in this crowd.