Clearing the path for hands-free, automated driving

I am working on a Strong AI system that offers a very different solution to automated driving but my system is at least ten to fifteen years away. This approach using more conventional computing and intensive sensors is looking more and more promising and its starting to look like we will be seeing real working systems within a few years - very exciting. I?m always happy to see Strong AI miss another boat.
Another really good idea is using a bus as a platform, a much better shape (than a car), nice and stable, and there?s plenty of room for all the equipment.
One idea I would like to share from my specification is establishing a standardized warning system telling other drivers that a vehicle is under automated control. Something like a police or emergency light, roof mounted, maybe blue or green, but different to the emergency services. Another is a requirement of an operator as an on-board human presence, for dealing with interdictions and arguments etc if nothing else.

This same type of automatic drive control was in Mechanics Illustrated before I retired in 1988. With the increase in cars on the road today and the speed which they mantain it would probably save a lot of lives today. I also think that cameras on the higways would make a lot of money for the counties in catching cars exceeding the speed limits.

For over 18 years I drove back and forth to work in Silicon Valley (San Francisco/San Jose/Bay Area). Often bumper to bumper thinking wouldn?t it be nice if they put sensors within the barriers along the inside of the freeways that smart cars could use to determine position, location (GPS), and to actually and safely drive the car too.

Like wake me up when the car is near my off ramp (close to work or home)!

And how about using nuclear power to also move electric cars via these barriers? Gets you there then your car is all charged up for the last 40 or 50 miles (or however long the batteries will take you).

Make it a toll road with automatic smart passes. Only a few cents per mile and cheaper and less polluting then gas.

I guess some psychic scientist sensed my wishes as this all is about to come true!

Enjoy the ride and your computer will wake you up as you near your off ramp or maybe even drive you all the way!

What I would like to see is a combination of techniques: 77 GHz radar, video sensors (i.e. cameras) with image processing to steer a vehicle between lane markings, and accuracy-augmented GPS to not only keep a vehicle on course, but also decide when to apply throttle, when to let up on the throttle, when to use regenerative braking or conventional braking, adjust engine parameters such as cam phase angle, et cetera, to yield the smoothest and most economical performance.

There are so many gaping holes in this idea, I'm not sure where to start... Has Zhang actually thought this through, or is he just focused on winning respect in the geek community? Can-be-done and should-be-done are two very different things.

Here we go : First, who's going to pay the ENORMOUS expense of retro-fitting every car, as well as the staggering number of sensors and magnets needed to cover the roads? This isn't even close to being practical.

If you want to read or snooze regularly, why not take a bus or a train? If the goal is to reduce accidents and emissions, it would make far more sense to expand and improve public transportation, which would benefit everyone- including those who don't own a car. Perhaps we should improve driver's ed- getting licensed in Germany, for example, costs $1500, and involves a lot of training. We hand them out to just about everyone who demonstrates even a modicum of skill. No wonder people are dying, and why we're stuck with this lame 55 mph limit.

Many people- including myself- absolutely love driving cars. Do you really think that the guy who just bought a $90,000 Mercedes wants to sit back and let a computer drive? I predict that most people won't want to give up control of their vehicles, which would make this system a huge waste of time and money.

Finally, what would happen if the system suddenly crashed? Even if it sounds an alarm, will you wake up in time to get your bearings, and regain control before it's too late? Imagine the pile-up that it could cause, since everyone was sleeping, reading ,or just generally distracted? And who's at fault; who's insurance company foots the bill? What an all around mess.

This concept has too many dangerous flaws, and a lop-sided cost to benefit ratio. Better and cheaper results could be achieved by making public transportation more desirable, and requiring a lot more training before issuing driver's licenses.

Unfortunately, I agree with "ddferrari" in that there are too many ways for this to be a VERY expensive disaster. Even if they initially limit this to the urban areas, think about the traffic disruptions to retrofit those roads where they intend to try this!
There was a book or short story many years ago about this very topic, "automated" highways. I think it was by Heinlien, but it might have Asimov or Clarke. It has been so long since I read the story that I don't even remember the title. It ended with the fatal crash of the patrol officer hero. Would this version be any different?

The problem isn't the technology, it is and always will be the people behind the wheel. IF you have ever seen a car in the express lane that did not belong there you know the problem. If you ever had to pass someone on the right you know the problem. People die in car accidents every year, not because they did something wrong, but because we failed to remove the stupidly arrogant. Repeat drunk drivers, texting phone users, hurry-hurry white rabbits, all need to be prohibited from ever being behind the wheel again. That won't happen and therefore people will die in that automated system. Killed by the stupid arrogance of a driver thinking, "I, I, I, ..."

This discussion is over 10 years old. As a Federalemployee I worked with the graduate students and their Advisor at the University of Minnesota on similar projects starting as early as 1989. 3M, being a Minnesota company discussed embedded magnetic technology in the mid 1990's, but the Mechanical Engineering Department at UM preferred GPS location, radar for nearby sensing and tactile feedback. These ideas were presented to Ford and Chrysler in the mid-late 1990s, but their reps al were concerned with legal implications in the event of a crash. They said the victim's lawyers would focus on failure of the technological device )or the abstract interface of equipment to human) as the cause of the crash. The conclusion was that foreign companies would install these technologies long before American manufacturers would.

They were doing what appeard to be "successful" trial runs around San
Diego a decade ago.

Some mean-spirited people have a deep-seated hatred (or fear?) of
individual liberty and privacy, and there are others who will do anything for
a buck. They apparently want other people surveilled and controlled every
waking moment. They love mass transit, where people can be packed
together like sheep and more easily tracked and controlled. They hate
individual automobiles and helicopters and planes with which people can
go when they want where they want and avoid being easily tracked and
abused.

Ethical US STEM workers, OTOH, knew a couple decades ago that these
systems could be implemented, but refused to do so. Apparently, they
either don't teach professional ethics to engineers and software
developers, anymore, or professional ethics have been total subverted by
academia or the flood of cheap 3rd world labor.

I heard some freak, just this morning waxing gleeful about the accursed
seat-belt roll-out and imposition. Of course, a 4- or 5-point harness,
helmet, and brace would be much better if your actual concern were
safety, and people have always had the option to take such extra safety
measures. Many people pointed out when they were hatched that the
power-mad feral federal government would not take long to impose them,
and we were correct. It's the same with this abomination.

Cars will not be retrofitted for this, rather new lines of cars will feature this ability in 15-20 years. Eventually in main thoroughways people can use autopilot and in rural or open areas they can drive manually - just choose. Remember that scene in iRobot where he switches his car from automatic driving to manual driving?

About safety, computers can be built with multiple layers of fail-safes, which is much safer than the one layer of human fail-safe that exists today (make one mistake or become unconscious, guaranteed accident). With multiple layers of fail-safes, it will be like a hundred people with perfect sensory and driving ability automatically taking control if the one before makes any kind of minor mistake. MUCH safer than what we have today with one semi-skilled driver at the wheel of each vehicle.

There will still be control over the vehicle, in fact - you can take manual control at any time in any place, drive it wherever, and the computer will only come in when there is a near-accident and then you got full control again. It will be like a videogame where you can drive wherever, but it doesn't let you within one inch of the guardrail.

First, we need to think positive. How can we make things work
and solve the problems we currently have? All the negative
arguments dismiss the fact that our current transportation
system has real problems.

Healthy discussions can lead to advancements. No single
person can design or even comprehend the complexity needed,
so no single person should think they are smart enough to
point out why it won't work.

Our current transportation system is in decline and we need to
find a new foundation for moving forward. I see a possibility
growing out of the new Segway vehicle called the PUMA. A two
wheeled (side-by-side), two seater, half the size of a Smart car.
It can turn 360 degrees in its own space.

We could drive them on to truck or train platforms for public
transport to destinations and drive them off when we get there.
This would address a lot of the issues brought up here. The
immediate naysayers will have their comments on why it won't
ever come to pass, but again they don't have any answers.

I like to think of it as the glass half full or empty. Think about
it. If Henry Ford brought out the Model T today people could
point out all kinds of reasons why it wasn't reliable, safe,
comfortable etc., but because the people supported it there
have been incremental improvements that addressed the
problems.

In the 21st Century a concerted, supported effort could take us
to the next level. The big difference is we aren't starting from
nothing. We have to transform a system that is ubiquitous and
vital. So much so that we don't seem to be able to seriously
regard its ever growing problems.

The truth is we have so much more knowledge, technology,
experience and ability to communicate that we could do this if
we could just focus on meeting the need. This is an individual
problem that must be resolved at the society level requiring
change on the part of individuals. If it accomplished
successfully we will see improvements at both the societal and
the individual level.

One does not need to understand the complexity of this system to see that it isn't the answer, any more than they need to be a doctor to know that cancer is a bad thing.

This "automated driving system" is an expensive band aid, not a solution. It does nothing to reduce the number of vehicles on the road, which IS the problem...it simply delays the inevitable: we WILL run out of room in the not-so-distant future.

This system may very well be do-able, but it's not the answer. The solution to pollution and accident reduction lies in a massive acceptance and shift to the use of public transportation, not sensors and magnets. This can be achieved by making trains and buses a lot more appealing, i.e. safer, cleaner, and much more convenient.

The solution has been here all along- and it's not high-tech wizardry: exercise more personal responsibility behind the wheel, and make the small sacrifice to take a bus or train whenever possible.

Public transportation is for those who want to relax and not drive.

This is an expensive solution to a problem already solved. Technology for the sake of technology without doing anything to reduce highway congestion and traffic pollution.

This is not a new concept. GM had this concept in their World of Tomorrow exhibit at the 1984-1985 World's Fair in Flushing, New York. Granted, maybe the technology is now more possible, but I agree with ddferrari on who is going to pay for all this drilling, installation, maintenance, etc. May I point out that in snow-belt states, they already have enough problems with freezing/thawing that causes hols, potholes, etc in road surfaces. That's all they want is more holes drilled in the roads for more places to allow water to crack and destroy the road surfaces.

We can't even fund repairs of the currently failing bridges and roads in many of the very heavily used interstate system. Add to the fact that many states are now doing the grind and repave with recycled asphalt. Oops, there went the sensors. This is nice concept, but obviously not well thought out. I have been an IT professional for 40 years and believe me, there will be system crashes.

There is a saying I heard long ago, that seems appropriate:
"Nothing is impossible for the man that does not have to do the work or pay for it".

Driving on the interstates in the Nashville, TN metro area at rush hour is a hassle and stress inducer, that I avoid whenever possible. Too many people talking on cell phones, not paying attention, too many jack rabbits weaving in and out across 4 lanes of traffic. No enforcement of the HOV lanes. People driving 55 in a 70 zone the fastest lanes, pickups towing trailers going way too slow to be using the interstate, etc, etc, etc. In the morning i don't ever use the interstate, too much chance of getting in an accident (and there always is one that shuts down the interstate) with nowhere to go, just sit and wait.

I have a coworker who was rear ended by an idiot not paying attention after everything came to a halt one morning due to an accident.

I agree with the person who said "if you want to sleep or read or relax on the way to work, take public transportation",

I used to ride the express bus in Indpls, IN when i worked downtown and live in the suburbs. I could nap, read the newspaper, talk with others, etc. and arrive at work or home unstressed and relaxed. The I had a job change and had to drive to the work location. Bus commuting for 1.5 hours with transfer was just was not feasible.

In the Nashville area, they are serious lacking in public transportation options for people in the suburbs, so thousands just drive. Not that many carpool/vanpool. There was a expres bus route just started from my town to downtown Nashville, but it does not help me, I don't work downtown.

It's called public transport folks, just get in, sit down and shut up. No brainer.
If it ever comes to that in personal vehicles, shoot me. Where is the skill? the fun and enjoyment? Personally i'd hang on as long as i can to my old hands on driving experience, even if those vehicles were banned.

Setting aside arguments for mass transit versus energy intensive individual people movers, I feel that a sensor-based approach has significant limitations beyond the cost and time to get it installed in cars and roadways. Why provide the benefits of automation only on major roadways?

If we used a camera plus AI technology plus GPS plus transponder, we could have so much greater, more versatile vehicle control. Furthermore, camera plus AI technology, applied to traffic lights first, would provide enormous benefits in saved time and energy.

Traffic lights could communicate in a wireless mesh network, optimizing thruput of all light-controlled roads. No more stupidity of sitting for up to three minutes at a light with no cross traffic, then getting the go-ahead just as cross traffic pulls up. Incredible!

Many of the new traffic lights I see going up are getting cameras. Why not intelligent traffic lights with cameras. No more digging up roads for coarse sensor-based control.

Now on cars, too, rather than sensors, the cars would have cameras, one each (or two, for redundancy and greater coverage) fore and aft. (It could even be handled with a single, centrally mounted pod camera, looking up with a 360 degree anamorphic mirror above it.)

AI processing would recognize other cars, lanes, obstacles, signage, etc. Navagiation would use GPS supplemented by signage, and would learn your preferences from spoken commands or by observing (and backing up) manual driving.

Transponders such as on aircraft, in some form, would let vehicles communicate with each other. At the least, they could signal that each vehicle was under automated control, and groups of such vehicles could link up in highway "trains." At a higher level they might be able to communicate what they had learned about driving conditions based on recent observation.

You would not need centralized government or toll authority funding for this, because even in unit quantities it could provide benefits for early adopters. You wouldn't need critical mass in order for individuals to benefit from it.

But I suppose that level of AI processing of real-time, highway speed images is pretty much pie-in-the-sky, far in the future. *Sigh*

Thought of the lane departure system and active cruise control. Use the highway lines to automate driving only on the highways. Local roads are all manual. No need for magnets since all highways are lined already.

Other approach is to apply it in a mass transit system. Airport trains has already unmanned trains. Just make it bigger that can go farther.

Sensors in the road? What is this, 1999?

Please see the accomplishments of the THREE DARPA Grand Challenges,
in 2004, 2005, and 2007. Oh, I see, Stanford won, not Berkley.

Done on any road, using GPS and sensors. No retrofit for old cars,
the computers watch out for all of you stupid, backward "enthusiast"
drivers, while the rest of us get from A -> B, safely, comfortably,
and easily.

We have tried letting people do the driving for 100 years now, and
it results in over 30,000 deaths per year in the US. Let Stanley
drive.

http://www.darpa.mil/grandchallenge/index.asp