Chuck, the thing that strikes me is the number of sensors. How many sensors does the current technology, the human driver, use? Not many. I really think these researchers are barking up the wrong tree. They probably only need an accelerometer and a stereo vision systems like the Kinect. Perhaps GPS would be useful as well.
I can't really speak to the issue of whether they could do this with stereo vision and Kinect. But I can vouch for the fact that they do have a lot of sensors on these vehicles. To know where they are, the vehicles combine GPS data with "low-G" accelerometers and gyroscopes.Then they have to filter that data through dual- and quad-core processors. To avoid obstacles, they use infrared cameras and, in some cases, LIDAR systems containing as many as 64 separate lasers to create a "point cloud" of obstacles ahead. I suspect that as they get better at this, we may see the number of sensors drop, but right now they still feel they need a lot of sensing capability to handle this chore.
Its really amazing how technology changes our perception on communication and transportation. Look at that above article, it was all an interesting matters that might benefit as all. Who doesnt want to own this own? This is such a big dream to each and one of us, don't you? I'm really a fan of modern facilities like this especially vehicles with lost of special specs and good quality. If I will purchase another car one thing I had to have in that new auto is all the best security benefits. I wanted to make sure my family would be secure no matter where we go. That is why I went to a great site. It showed me the very best automobiles for security and helped me get a car loan. When have you expected additional info about a new or pre-owned car and turned to an internet search on "boa auto loan?" Your search is over, whatever that you need is at CarDealExpert.com! Need to know more about Car Deal Expert? Get auto loan info here!
Please do not do hijack a discussion thread again for crap websites. Granted discussion threads almost always drift away from the original topic, but what you did is outside the realm of good behavior here.
I think that if you really analyzed it, you'd find that the human body has a large number of sensors all attached to a very sophisticated "inference engine". Subtle changes to the pressure on your hands tell you a lot about how hard the wheel is being turned. The part of your body in contact with your seat back, in conjunction with your inner ear, senses acceleration, peripheral vision picks up rapid movement just outside your field of vision, your eyes are constantly adjusting to varying light conditions and so on. Replicating the human driving experience through sensors hooked to a computer is seriously complicated business.
A phrase I hear all to often where I work is "All's you gotta do". It's what people say when they think something is going to be a simple task. It's usually said by people that will NOT be implementing the task. After all, if people can do it (drive a car, and some of them quite challenged in the brain-power category), then surely machines can do it if we add enough sensors and processors. Yeah, right. And that's just the technical side (this is an autonomous, hopefully intelligent, safety critical Robot we are talking about).
Wet hardware (humans) is truly an amazing instrument. Capable of taking far more information than we are aware, while filling in the blanks for missing/conflicting/incomplete info, and adapting to unforeseen circumstances. This is truly a deceptively simple task.
The other side is legal. Will Toyota or Audi accept the legal liability for auto accidents that occur when "self-driving"? In reality, they must since it's their "brains" behind the wheel. But, when I have to buy auto insurance, how is my provider going to bill me? Or will they bill the auto manufactures? Or will the government do what it seems to be best at, which is grant immunity to big business leaving us to fend for ourselves against 2 ton death machines designed to be as cheap as possible.
When I first saw the article title, the first thing that came to my mind was that Toyota was trying to re-brand its issue with unintended acceleration (really, it wasn't unintended ...)
@naperlou--perhaps there is a relationship between the sophistication of the processor and the number of sensors required. The human has hands and feet as actuators (and maybe voice for some functions) and eyes, ears, pressure, and touch (for vibration, g-force, etc.). However, the human has the brain. Fast microprocessors in the autonomous system can't match the intelligent processing of the brain. Solution? Add more sensors to break up the input into more digestible chunks and write multi-variable models to try and account for everything.
The human brain has too many connections to sensing mechanisms to expect a series of man-made devices to duplicate. Therefore if we expect to successfully develop a driverless car, we need to establish a control environment for it to operate within.
Imagine a freeway system designed to accept driverless cars. It must have destination lanes, lanes allowing the vehicle to enter the correct traffic lane. Lane changes must be controlled by establishing a speed and spacing control for each traffic lane. All of this can be controlled by satellite communications. There is an immediate problem with a factor known as peristaltic effect, or adding to a fixed series of moving lines without spreading them. The new vehicles must combine with the existing lines without making them wider. This requires controlling the beginning spacing based on the anticipated additions. Destinations must be entered in advance so that all calculations of traffic flow can be control planned. A reservation system could be used for a fixed length of travel but it must be maintained by each traveler.
Problems require solutions and with our current calculating capability, I am certain a solution could be forthcoming.
tnek--you are right on that a combination of vehicle technology and adpating infrastructure is the ultimate solution. This is, in part, what V2I (Vehicle to Infrastructure) is about. I also agree with your physical solutions--design lanes to feed in cars correctly, let them out correctly, control spacing, etc.
Here in Colorado they have tried, during the peak ski season when there is excessive traffic going to or coming from the mountains, to set up caravans with police leading and trailing a group of vehicles moving at constant speed. I haven't heard much lately on what they learned from the trials, but how much simpler would that be with intelligent vehicles, V2V and V2I communications?
I missed this at CES but I was only at the show for a day.
As a Californian, I'm NOT looking forward sharing the road with self-driving cars. Our Governor signed a bill authorizing driverless vehicles in 2013. Maybe I'm old-fashioned about it but I think human skills are superior to sensors.
I agree with Nadine, that the idea of cars driving themselves entirely--sophisticated sensors or not--is a bit daunting. I like the idea of a car being able to park itself in a lot and perhaps come back and pick me up, but driving on its own I'm not quite comfortable with. Maybe it would be OK as long as someone was always in the car to override any error--but wouldn't a self-driving car make a driver lazy? And I don't know about anyone else, but I love driving...there is something soothing about it, especially long drives with good music. I have used long drives as therapy! Have we become so preoccupied by other things that we can't enjoy or accomplish this simple act?
At many airports you ride in driverless trains, and have been doing so for decades. While a much simpler case, there is still the potential for error and human loss. Somebody figured out the legal side of that, along with the technology.
I agree that driving can be fun; on the other hand sometimes a bus or train ride is really fun to relax, do some reading, have a real conversation, etc. and still get where you need to go. Think about all the other technologies that have become "personal" (i.e., all the power in your smart phone, for example) and consider that many people will welcome the idea to hop in the car and tell it where to go, then sit back. For them, it will be more enjoyable than the bus.
Good point about the driverless trains, but you have to remember that is on a track and there aren't other trains (or cars) around also driving to create possible interference or cause an accident. I guess I didn't think about the idea of a commute...it seems like a good idea if a car can just be programmed to go on a set path and follow a route while the driver can kick back and relax. But then I see a future with cars all moving on tracks and everything being automated and predictable...again I tink it takes the fun and spontaneity out of driving. But that's just me!
Hi Elizabeth. OK, I have to admit I like driving my Jaguar. I used to have an AWD Turbo Eagle Talon and really DROVE that.
But, I think most of the time most of the urban miles are commuting, so this technology could have a real impact. Also consider that with a self-driving car, you can optimize the behavior for maximum efficiency, especially if you communicate w/other vehicles and plan ahead to minimize speed changes. The comments from another writer are dead-on regarding adapting infrastructure to facilitate this; but hey, we have HOV lanes, bus lanes, light rail along the highway, distributor/collector lanes, traffic lights to pulse in volume onto busy roads, etc.
I'm with you, I hope that we are still "allowed" to drive in this future. Think Will Smith in I Robot.
While not all driving is for fun, cars are not the right solution for driving that is not for fun. The main purpose of cars is simply for the feeling of power. If you simply want transportation, mass transit makes far more sense. Self driving individual cars make no sense at all.
The only point of self driving would be for cargo transport off the public roads, such as in Alaska or tunnels.
@Rigby5--I would have to disagree. Strip away your world view of a "car" as you know it today, and for the purpose of commuting what if you have a transportation module that you can get in whenever you want, go whereever you want, and it does so autonomously? Would not that be very attractive? People want convenience, ability to change plans spontaneously, and flexibility. While mass transport works for many, why are there so many cars even in cities (like, say, Shanghai) where it can take 2 hours from one side to the other?
Eafpre, the reason there are so many cars is because people enjoy getting into a massive 1 ton machine, and having it amplify their every hand or foot movement with hundreds of horsepower. Basically cars are human transformers, that make us all feel superhuman for just a bit.
No one want to just get into a little capsule and turn up somewhere.
That would make us claustrophobic and bored. For the tremendous cost of a vehicle, we would all much rather save that money and simply use mass transit, which already is the transportation module you mentioned, except you may have to transfer once or twice, and don't have to waste thousands a year to buy, fuel, and maintain.
If you did not get the feeling of power from driving a car yourself, no one would want one. They should build a subway in Shanghai.
For those of you who would feel uncomfortable driving on the freeway with driverless cars, THINK. You are now driving on the freeway with cars without a competent driver, driving too slow, speeding, changing lanes constantly, tailgating, and other careless and inconsiderate actions. For those who fly, you have been riding with driverless planes since the 1960's. Fortunately, they do have a driver to take off and land. Now it is possible for the pilots to land, hands off, as well. Those who are young enough, will see hands off driving in the not-so-distant future. The technology is there, the development is not. You will enter you starting co-ordinates, taken from a GPS reading, and your destination from a log book of familiar places, start the engine, and relax. Your controlled trip will be much more efficient, just like traveling on cruise control. Next available is an engine that runs constantly at its most efficient speed, without cooling or lubrication, no changing gears.
Two problems I see are litigated responsibility for accident injuries and consumers giving up their spot behind the wheel. We're in love with our cars and it will take a lot to get us to turn them over to computers.
In the last month, Toyota agreed to pay about $1.1 billion to settle the class action lawsuit over unintended gas-pedal acceleration. Wonder if that will indirectly factor into the development of this technology.
There is an application for this in places that have private roads in wilderness areas, tunnels, etc., but NOT on public roads.
Sensors, decision making software, circuits, etc. are never going to be reliable enough, and the self accelerating Toyotas should have already convinced everyone to minimize electronic involvement. Everything that can be done by humans, should be done by humans. It is likely to be more than a century before this could ever be both reliable and cheap enough to be of use on public roads, and it may never be acceptible.
"There is an application for this in places that have private roads in wilderness areas, tunnels, etc., but NOT on public roads"
Rigby5, I think it's a general mindset of peoples, about the security concerns. I don't think there is any need of fear about it. Auto navigation systems will take care about colloid free journey even in busy streets.
Cars are not multi million dollar planes or even million dollar trains. Cars are not maintained, get old, are abused, not services, or worse improperly serviced.
So auto driving cars will never be acceptible on the public road. Sensors are incredibly inadequate, and the only way planes do it is with mandatory transponders that you will never be able to require pedestrians, pets, trees, etc. have.
All it takes is one spectacular death, and it will cost the manufacturer billions. It does not matter if it is safer than normal driving. The juries will award billions.
But it will also be a century before automated systems even begin to approach the skills of a human. There are thousands of problems they have not even though of yet, such as radar interference, optical illusions, rain and snow, etc. People don't even know how to test software systems this requires, much less actually creating it. Defects go up exponentially with program complexity. There is no question is it beyond our reach presently.
Charles, quiet sometime I read the same from Google in Canada. Any updates about that. I think such vehicles will helps to reduce accidents and can offer safe journeys. So far we had seen only driving assisting systems only.
at least initially. Yes of course, driving is fun. Why should I leave this to a computer? But we all know that most western societies are aging quickly and those who can enjoy their old age in good health, are fortunate exceptions. Dominant biographies include glasses in one's forties, a stroke in the fifties, diabetes and hearing aids in the sixties and artificial hips or ankles in the seventies, while still getting 100 years old. That makes some 30 years without possibility to participate on individual traffic. And public transport is rather rare in the rural areas. Think about the increase in quality for perhaps a third of one's life.
Cars that humans drive will become toys. There is no need to own cars. Order a driverless car to come get you at X;XX time and it can be there. After you get out you get charged and the driverless goes on to park itself or to pick up the next passanger. Utilizaton of cars will go up there would be no need to have two cars or even one car to a family. If soccor mom or dad needs a van, order one up on you cell phone. Same if you need trucks. Think of all the cars parked at airports parking lots. What a waste of the transporation resource! Life cycle of driverless cars can be shorter because of higher usage rates, meaning faster upgrade to newer technologies.
I acturally think I rather have a predictable driverless car on the road than a unpredictable teenager driver texting or a hung over driver. Driverless cars can also network with each other to know what cars are around them, much like what the GPS based air traffic control would be like but simplier because there is only two dimensions instead of three.
I agree with you, sysdesign. Some of the experts I've talked to also agree. In certain situations today, they say autonomous vehicles can do better than humans. In the DARPA Urban Challenge six years ago, for example, it was said that human drivers were repeatedly making mistakes -- rolling through stop signs and failing to use turn signals. Driverless cars never made those mistakes.
Charles, I agree that human drivers are making repeat mistakes, but very rarely. But they have some logical thinking for reaching the destiny or taking suitable route depends up on traffic situation and information’s in front of them. But when comes to auto navigating system, the entire decisions are based on the info feeded and there are chances for wrong decisions too, because of the lack of right information.
Many good points but I believe that the first adopters should be those who view driving as a necessary evil and who are most likely to be engaged in other activity while driving. Urban environments where chronic congestion is widespread is an area where Self-Driving Technologies will be a benefit provided they are not used as an alternative to the continued development and upgrading of our mass transit systems.
Security and the integration of regional traffic information data aquisition into the SDTs will be addressed as part of the product development process and the collection, transfer and delivery of secure accurate information will become an integral part of our transportion infrastructure.
I envision a day when those who do not desire to bear the burden of piloting a motor vehicle where other options are not available will not have to do so. I also believe that when this comes to pass that there should be more rigorous licensing requirements for those who do want to have the privilege of driving autonomously. What a glorious day that will be for those of us who enjoy driving! No more left lane hogs, no more right turns from the left turn lane! Too many examples to list. I am excited.
Pblanche, driving is a major activity, where atmost care has to take but now a day's the most soficated and advanced in-house entertainment systems are diverting driver attention up to an extent. While driving, personally I won't use such systems for a better attention.
As usual driverless car depends on software and networking which will become an easy target for terrorist to create chaos, damage to property and lives by hacking their software and introduce malfunctions.
In my opinion, it is good to have them in ideal world but not in today's world where terrorism is a big menance.
Prakash, all coins have two sides. Like that there are chances for misusage, but hope for the best. Technological developments are for human advancement and to make the lifestyle smoother/ easier. So there are chances for misuse too, but will you think that will make cease further invention, No. it's a part of the journey.
The comment about humans making mistakes illus6trates exactly the fundamental fault that all of the driverless cars will be built with, which is the ability to think. The very best that a driverless caar can do is follow a set of programmed rules, it really can't do any better than that. So if there are no cars within a mile, it will still stop at every stop sign, even if you are attempting to rush to a hospital to save a life. The driverless software will never consider the relative risk, the lawyers have already decided that it will be way more cautious than my grandmother was when driving. The other fault is even worse, which is that with the billions of lines of code controlling the car, if there ever is a problem, the carmaker will have to refuse to admit the existance of a problem, because of potential liability concerns.
So even if there were no other problems with the driverless car, those two problems amount to an immovable show-stopper.
There's one simple fact that proves your point perfectly, William K: Even the most ardent supporters of driverless vehicles have admitted that the technology's not ready for use in crowded, complicated settings, such as Bombay, India. The human brain is much better than electronic controllers at dealing with the unexpected.
Yes, my point exactly. And unfortunately, it seems that the most accidents happen where there are lots of cars, and that is where they are being touted as such an improvement in safety. And as for meeting unexpected conditions, I come across those about half the times that I go driving. So really, a driverless vehicle is "a solution looking for a problem". They would probably be handy for convoys, where driver fatigue has been a problem, and they could probably help make intercity trucking a bit safer, but I would not want thenm in my neighborhood.
I don't find driving relaxing at all. My commute is on the Interstate highway system and being vigilant and aware is always a must for me. The death toll on Tennessee highways last year (2012) was 1,012 people. Most of this was due to carelessness, elevated speeds and inattention. In my opinion, this is a ridiculously high number. The number of people taking public transportation has remained steady but most families are multi-car families with mom and dad working and they both need transportation. This fact along increases the probability a greater number of accidents will occur. I would say the development efforts of Toyota and Audi are well -placed and should continue simply due to annual death rates on our public highways, local and interstate. Even with this being the case, we are probably years, maybe decades, away from cars Charles mentions in his article.
A new white paper by the Association for Advancing Automation says that increases in industrial robot shipments correlates positively with increases in US job growth based on Bureau of Labor statistics. The result could be new types of manufacturing and engineering jobs.
Focus on Fundamentals consists of 45-minute on-line classes that cover a host of technologies. You learn without leaving the comfort of your desk. All classes are taught by subject-matter experts and all are archived. So if you can't attend live, attend at your convenience.