[kousik]

Been to the streets of Mountain View last week. Was on a 2 way street with no divider. Saw the left turn indicator of oncoming vehicle and became dead slow to let it. However the driver changed his mind in last moment and allowed me to continue ahead. When passing, noticed Google's self driven car was just behind that car, which had to break abruptly, and noticed in rear view mirror when the car ahead finally took left, it resumed its journey ahead. Must be pretty routine exercise for it.

[Acharya]

I always wonder the real need for self drive cars, especially when there are already multiple alternatives available for it, such as public transport, chauffeurs and cabs.
I am not a great fan of machines doing Cognitive work in real life, moreover driving is such a pleasure if done in right conditions!
Why would anyone want to give it away 😃

[Karthik Chandra]

Ford fusion autonomous vehicle

https://www.youtube.com/watch?v=62v-NmGK0zk

[mayuresh]

A couple of interesting articles on the topic:

1. Computer, Drive My Car!
http://ieeexplore.ieee.org/stamp/sta...number=7807160

Divides the future into 3 eras:
- Present - 2020 when autonomous cars will be just about reality
- 2020 - 2040 "mixed era" when they will be maturing
- beyond 2040 it might be illegal for humans to drive! The accident fatalities would drop to near 0 from millions per year globally today

2. Ted: What a driverless world could look like
https://www.ted.com/talks/wanis_kabb...ould_look_like

While the whole talk is good and has nice graphics, see position 10:30 with a graphics showing how traffic would flow without signals at busy squares (and the speaker mentions Mumbai as one of the examples of chaotic traffic here!)

[Guna]

Not sure if this was posted somewhere in the forum.

First time in India: Tata Elxsi to test driverless cars on Bengaluru roads

http://www.business-standard.com/art...1701566_1.html

[audioholic]

Originally Posted by Guna

Not sure if this was posted somewhere in the forum.

First time in India: Tata Elxsi to test driverless cars on Bengaluru roads

http://www.business-standard.com/art...1701566_1.html

Its more of a 'even we can do it' strategy'. They had showcased a similar prototype last year inside their whitefield campus. But really looking forward to see this vehicle on the road. IIRC its a Range Rover.

[locusjag]

Foreword
I had written about my experience with Usage Based Insurance (UBI), over here in the USA a long time ago (Link:http://www.team-bhp.com/forum/indian...-snapshot.html). In it, a lot of questions came from fellow team-bhp users on the topic of self-driving vehicles. I was stumped and my responses were anything but clear or educated. I felt bad about it, but then it struck me that economists, workers in entire nations and industries are all stumped by what's coming and even they do not fully know how they will survive in the new age of Robotic Automation and Artificial Intelligence (AI).

What is AI?
This is a little complicated. AI was actually born when Alan Turing devised a computer in World War 2. This is to say that wherever there is significant machine-driven computing power, there is some form of AI. Our Personal Computers have exceeded humans in a narrow range of tasks - such forms of AI are called Artificial Narrow Intelligence (ANI). If you're scoffing at my definition of ANI, you are simply abiding by Tesler's theorem which states that people tend to call whatever is not achieved yet as AI and that people tend to disregard or take for granted any actual AI form in existence! In any case, today's Industrial Robots that manufacture heavy machineries are forms of ANI. So today's Robotics is just an early form of AI in existence.
On the other hand, Artificial General Intelligence (AGI) is the capability that would enable computers to become self-aware and excel at a wide range of highly cognitive, subjective tasks. In other words, the arrival of AGI is when almost every employed white-collared worker will likely be unemployed and when humanity will see a significant degree of upheaval.
Superintelligence will follow the onset of AGI. People like Elon Musk and Bill Gates are worried about the survival of humanity, since out of the global hundreds of AI researchers and unknown secret military AI researchers, only 6 are concerned with the development of friendly AI. A "Terminator-ish" scenario is not at all unlikely.

Why talk about Robotics and AI at all? Isn't this spiel supposed to be all about self-driving vehicles?
What are self-driving vehicles, if not a manifestation of Artificial Narrow Intelligence (ANI)? Self-driving vehicles of today are Robots, as they excel over humans in driving safely and nothing else. Moreover, self-driving cars are products of intensive machine learning, which is at the heart of AI today.

The "when"
"AI" first became a "thing" during a conference in New Hampshire, USA, in the '50s, when scientists foresaw a future with intelligent robots. But since then AI has gone through many winters and boom-periods. It is full boom-time for all things Robotic and AI today, thanks to a few hardware successes (mostly NVIDIA's success with GPUs) that enabled fast computing and Deep Learning. It is forecast that AGI will take over anywhere between 20 to 80 years down the line. In general, according to the law of exponential growth that computing power grows by (due to Moore's law), it is feared that AGI will arrive sooner than we suspect. As computing capability keeps doubling itself every year, the inevitable is really inevitable. And in case you're thinking that Moore's law might be nearing its end, don't lose sight of the ball here - there's quantum computing, computing based on the connectome (i.e., computing based on the human brain's connections and structure) and other breakthroughs in hardware still in the works, waiting to be proven.

Why humans will certainly be caught unawares by AGI
Computing power has been doubling itself every year according to the Moore's law. The Moon rocket by NASA had lesser computing power than today's washing machines. AI researchers estimate that when exascale operations are made possible for computers (giga is 10 raised to the power of 9; exa is 10 to the power of 18), AGI is most likely to arrive. This is since human brains process information at an exa-scale.
Anyway, let me illustrate how this exponential growth will catch humans off-guard. Exponential growth simply means that with every step, there is a doubling of whatever you're measuring. Let's take a football stadium and fill 1 drop of water in it in 1 minute. The next minute you fill 2 drops, and 4 drops in the next and so on. The football field would have a thin coating of water somewhere around the 40th minute. The field would be inundated with a few feet of water by the 45th minute. But the entire stadium would be inundated (including the stands) by somewhere around the 47th minute. In the last 2 minutes, any over-confident spectators who were watching this experiment from the stands would be doomed. Many AI researchers are stating that we are in that last stretch when our doom is near. I too doubt if many of us realize that most of our jobs are going to get taken away all too soon.

Enough about AI. Let's talk specifically about Self-Driving vehicles. What's the prognosis?
Right. In 2004 during the grand DARPA challenge, the best self-driving vehicle that geeks could create was a Humvee that broke down in the Mojave desert hardly after a few miles when it hit a rock. This was the humble beginning of self-driving vehicles. Today, we all know how Google drove its self-driving cars a million miles before its first accident (which was more due to human error on the part of a bus driver than anything else). There are 3 states in the USA today where their laws allow self-driving cars. The state of Nevada allows self-driving trucks to deliver goods. Uber's trucking division has already been delivering goods commercially. In Europe, truck manufacturers like Scania have been experimenting with caravans of self-driven trucks that have plied on motorways safely with other motorists around them. And the city of Paris has just begun dabbling with a self-driven city bus.

The forecast by the Wharton School of Business is that by 2030, 47% of all jobs in the USA are going to get automated; you have to keep in mind that Truck driving as a profession is the most common job description in every state of the USA. Overall it's a given that every bus, truck and cab driver is going to find themselves out of their professions eventually in developed countries. The highly evolved highway and road infrastructure in these countries is only going to hasten the arrival of self-driving vehicles.

Of course, self-driving vehicles are so very good at averting accidents that this is the carrot that everyone is getting at. I shouldn't keep lamenting the coming economic singularity, when jobless economic growth will come to stay. The (in)human and economic costs of having humans drive vehicles is all too well known; India probably has to be the worst of the lot with an estimated road accident death every 4 minutes. That is appalling indeed. For sparing humans this misery alone, self-driving vehicles are probably worth it.

What are the unknowns surrounding self-driving vehicles at this point?

• Since the automobile insurance industry will be decimated by accident-less vehicles, where will they get their bread and butter next? The estimated loss in business for the overall insurance industry might be between 40 to 60% in revenues.
• If and when a rare accident does occur with a self-driving car, who ought to bear the liability for it? The OEM that's assembling it or should it be a root-cause driven attribution to a component manufacturer? Does this mean that this is where insurance companies will next earn their bread, with product liability insurance products?
• Uber CEO Travis Kalanick is betting on self-driving cars to effectively end the era of car ownership. Without the lure of driving in a future where manual driving may even be outlawed, perhaps people won't bother buying cars anymore?
• The self-driving vehicles of today rely on having a human operator who would ideally "take over" when a "new" situation presents itself. There are limits to machine learning and certainly as of today, every self-driving car will hit a situation where it won't know what to do. I think it was Tesla has faced with a serious situation where a customer was inattentive at that precise moment when he was needed to take over and drive his car. He crashed and died. Does this mean that these cars need even more Intelligence whereby they wouldn't hand over control to a noticeably inattentive driver? Or is it that the cars themselves have to get better enough to not need any human intervention?

Conclusion
I try not to lose sight of the coming human commotion and upheaval in developed countries when driving jobs go away first, as early as the year 2030. Other job categories will fall too, but not as quickly. Universal Basic Income from taxation of corporations that use robots or AI might help pay off people's needs but will that keep people happy? We know about resource and cash-rich Udta Punjab. And the citizens of Switzerland in 2016 voted a resounding 77% "No" for Universal Basic Income, in exchange for which they wouldn't have to work. They feel they would much rather work and earn their living.

Voltaire's dictum with work was that work is necessary to deal with man's needs, boredom and vices. Queen Elizabeth the First did in fact ban a knitting machine in her time, since she feared an outbreak of lawlessness among her subjects. Short-sighted governments in the future might ban AI and automation, but such economies will fall by the wayside as the other adopters may chug ahead on the back of their new-found productivity increases. Or maybe these other countries will descend into anarchy and maybe the survivors will just go and live in cabins in the woods...

In any case, I feel that automated or self-driving vehicles' adoption in India will lag the developed countries since Indian road and driving conditions present higher challenges by many magnitudes. But what has to happen will happen eventually even in India.

[msdivy]

Uber has taken its self-driving cars off the road after one flipped over in Arizona

The pictures depict the aftermath of a three-vehicle collision that occurred Friday night (March 24) in Tempe, Arizona. Local police say there were no injuries and Uber’s vehicle was not responsible for the incident. Another car failed to yield to Uber’s vehicle, police told Bloomberg, causing the Volvo to flip on its side. Uber’s car was in autonomous mode with two designated “safety” drivers up front at the time of the crash, the company told Reuters. The back seat was empty.

Learning pains, I suppose

[Rehaan]

Interesting to see this being open-sourced:

A Chinese internet giant just made a big move to compete with Tesla in the self-driving-car space

Baidu, one of China's biggest technology companies, said Tuesday it would open source its self-driving-car software in hopes of accelerating progress.

The move shows Baidu is serious about competing with the likes of Tesla as it looks to release the vehicles as part of a shared shuttle service in 2018 and to mass produce the cars in 2021.

Tesla has its sights set on China, which is becoming a more lucrative market for electric-car makers as...

Read more

[sprocket]

Thought we should have a dedicated thread for autonomous vehicles in India.

Here is the news from Infy today - http://www.ndtv.com/india-news/visha...-plans-1724627

[blackwasp]

US: Congress panel to consider self-driving car legislation

A subcommittee appointed by the House of Representatives of the United States of America in Congress will vote on Wednesday regarding a proposal to allow car manufacturers to have up to 1,00,000 self-driving vehicles bypassing the existing safety laws. The legislation will also prevent states from implementing any rules regarding the technologies of the self-driving cars.

This is said to be the first legislation that will help faster introduction of self-driving vehicles into the market. While it will require manufacturers to submit safety assessments to US regulators, it will not require an approval for new technologies to be used in self-driving cars before being introduced to the consumers. The draft revealed that manufacturers would need to demonstrate that the cars "function as intended and contain fail-safe features", and the transport department could not condition the launch or testing of such cars based on existing safety laws. The current safety laws (over 75) assume that a driver will be in control of a vehicle at all times. The draft also bars states from blocking the use of such vehicles for up to 1,00,000 vehicles per manufacturer. It also directs the transport department to introduce rules that alert the driver to check rear seats - in an effort to prevent cases of children being left behind as well as rules for performance of a car's headlamps.

The states could however set rules on other aspects like licensing, registration, liabilities, etc. for self-driving cars. Companies with semi-autonomous cars or those involved in their research have tried hard to lobby against the government to exempt various rules that are slowing down the introduction of self-driving cars in the market. With this legislature, testing and introduction of self-driven vehicles can happen all over the US rather than being limited to certain areas. Manufacturers have also said that unless there is a change in current regulations, further development of autonomous vehicles could shift to the Europe or other viable places.

The draft of the bill is available here.

Link to Team-BHP News

[14000rpm]

The Auto Industry is fast gearing up towards an Autonomous driving future. While some have their foundation set, many are still gearing up while others are merely waking up or peeping out of the window into this future landscape. All auto OEM’s, courtesy of their investment in connected cars, tend to agree that the car is no longer Sheet Metal and Rubber tires. It is all Embedded Electronics. In the not-so-far future, buying decisions will be based on user experiences built on connected services.

I'm sure many fellow members could already be working with clientele on this, others could be curiously watching in the lead-up to be consumers.

Fun fact : some of the cars we drive are already at Level 1. If you have cruise control, you are already at level 1. More on this later.

But first, some academics.

Autonomous Driving by definition–"a motor vehicle that uses artificial intelligence, sensors and global positioning system coordinates to drive itself without the active intervention of a human operator." (source – Wikipedia)

SAE categorizes Autonomous driving in the 0-5 scale with 5 being fully autonomous with human intervention.





source : www.sae.org

Below should help you further understand what you need to know about levels 0-5. The biggest difference is that, starting at Level 3, the automated driving system becomes able to monitor the driving environment.

Level 0: This one is pretty basic. The driver (human) controls it all: steering, brakes, throttle, power. It's what you've been doing all along.

Level 1: This driver-assistance level means that most functions are still controlled by the driver, but a specific function (like steering or accelerating) can be done automatically by the car.

Level 2: At least one driver assistance system of "both steering and acceleration/ deceleration using information about the driving environment" is automated, like cruise control and lane-centering. It means that the "driver is disengaged from physically operating the vehicle by having his or her hands off the steering wheel AND foot off pedal at the same time," according to the SAE. The driver must still always be ready to take control of the vehicle, however.

Level 3: Drivers are still necessary in level 3 cars, but are able to completely shift "safety-critical functions" to the vehicle, under certain traffic or environmental conditions. It means that the driver is still present and will intervene if necessary, but is not required to monitor the situation in the same way it does for the previous levels. Jim McBride, autonomous vehicles expert at Ford, said this is "the biggest demarcation is between Levels 3 and 4." He's focused on getting Ford straight to Level 4, since Level 3, which involves transferring control from car to human, can often pose difficulties. "We're not going to ask the driver to instantaneously intervene—that's not a fair proposition," McBride said.

Level 4: This is what is meant by "fully autonomous." Level 4 vehicles are "designed to perform all safety-critical driving functions and monitor roadway conditions for an entire trip." However, it's important to note that this is limited to the "operational design domain (ODD)" of the vehicle—meaning it does not cover every driving scenario.

Level 5: This refers to a fully-autonomous system that expects the vehicle's performance to equal that of a human driver, in every driving scenario—including extreme environments like dirt roads that are unlikely to be navigated by driverless vehicles in the near future.

However, the building blocks leading us into a driverless mobility future are:
Connectivity and Advanced Driver Assistance systems (ADAS)

Industry leaders call the connected cars development as the next auto revolution. The pace at which each company in the eco-system is investing in this already stands to tell. The image below shows an easy to understand rendition of the same.

[mayuresh]

Ethical issues with autonomous cars have been discussed on this thread before also.

One more recent ted talk on this topic. Besides autonomous cars, the safety of overall AI systems, particularly what should they learn from and how do we know it is safe, is being discussed in various forums.

https://www.ted.com/talks/iyad_rahwa...ess_cars_makeo

Should your driverless car kill you if it means saving five pedestrians? In this primer on the social dilemmas of driverless cars, Iyad Rahwan explores how the technology will challenge our morality and explains his work collecting data from real people on the ethical trade-offs we're willing (and not willing) to make.

[swissknife]

Here is one from our own backyard from Maruti. From the looks of it, it's far from being on the streets any time soon, if at all.

[rsidd]

I was reading two old (July 2017) posts by Rob Landley, techie known for busybox/toybox and other free software. (1, 2). Worth a reading IMO. Quotes:

Most cars are driven less than 4% of the time (they sit there parked the other 96%), but shared cars get driven more often. (The self-driving shared car people are expecting ~40% usage, and keep in mind half the day people are asleep.) This means maintenance becomes a much larger issue. An individual owner may take 5 years to put 100,000 miles on a car, a shared car can get there in months. Fleet vehicles wear out WAY faster, in that context a car expected to last 500k miles beats a car expected to last 150k miles hands down even if it _wasn't_ cheaper to buy and operate.
...

According to the 2010 census 80% of the us population lives in urban areas (I.E. in or around cities). For these people, over the next 5 years, "summon car" becomes a button on your phone, their car payment replaced by a much cheaper monthly subscription fee. They never have to clean out their car (although they can't use it as extra storage either, but that's just a question of getting used to a different assumption). They don't wash it. They never fill it with gas. They don't change its oil. If it gets a flat tire, the app just summons another car (while the car summons a tow truck for itself, which is the service's problem).
...
let's focus on the holdouts. The people opposed to this brave new world, who WANT to keep their landlines and film cameras and vinyl records and their own personal car that they drive with pedals and a steering wheel. And the message for them is: you can still ride a horse today. You'll just be restricted to horse trails, instead of riding around downtown in a modern city.

Learning to drive a car in 2017 is like learning to ride a horse was in 1901. This skill will be obsolete within a generation, and nobody will need to do it anymore outside of closed course hobby contexts. And eventually, nobody will be able to except on a closed course.

Self driving subscription service means no car insurance, no speeding tickets, no calling a tow truck when you're locked out or need your battery jumped. If the car blows a tire you summon another one and continue on your way, the _service_ tows it. You don't even need a driver's license: no extensive training, periodic retesting, or loss of mobility when you get old (or are too young). And if 99% of the population doesn't need or use those things, the infrastructure to provide them to human drivers will atrophy. (Hint: the failure mode of "speeding tickets are no longer a profit center so the police force stops devoting officers to traffic control" isn't "and now you human driver can go 90 in a school zone!")