Automated Cars

[lonetcrystal]

Automatic cars are much easier to get used to, especially if you are a new, elderly or learner driver. Some people prefer them because there is no need to worry about which gear to select or the actual manual gear changing itself, the clever automatic transmission does it all for you, this is especially advantageous when driving through traffic or busy towns and cities as you don't have to worry about constantly changing gears as you would with a manual transmission.

----------------------

Alice

 

Guaranteed ROI

[CaptainPanic]

Thanks! It is great to see that girls also understand something about cars.

 

However, if you would have read anything other than just the thread title, you might have replied with something different

[npts2020]

waitforufo: The winners of the DARPA challenge for automated vehicles used a radar sensor similar to what you describe, so it is technologically feasible. Furthermore, those vehicles ran off-road over pretty rugged terrain and had no other human input once the race was begun. A transit system, even with many more vehicles, should have fewer variables to consider and keep track of. A guidance sensor can but need not be the same as a detector for foriegn objects on the roadway. IMO an imbedded grid using optical or electromagnetic sensors for travel (a relatively inexpensive way of doing it) and a separate radar for hazards is the way to go. On an enclosed system stray objects should be rare and if the vehicles self-monitor for performance breakdowns should be as well.

[waitforufo]

I get the IEEE Vehicular Technology journal because I work in wireless and as I mentioned this journal covers mobile wireless topics as well as automated cars. I read the automated car articles because it is an interesting engineering problem. I hope you did not get the impression that I am against this idea. I am not.

 

As I mentioned, several years ago many automated car articles were on the topic of radars. The ones I read, and I read many, always seemed to find more problems than solutions. Many of these problems involved objects that humans could easily identify as harmless, caused radar based vehicle control system to stop the vehicle or at least disengage the automatic system. One I did not mention in my previous post was oncoming traffic particularly on curved roads. Such oncoming traffic would easily pass but would appear as a quickly approaching potential accident. One also has to consider the liability of system failure to the manufacturer. If GM or Toyota built such a system and an accident occurs with the system on, will they be willing to accept the liability?

 

In recent years great strides have been made location services. GPS is but one. Roadside tags and vehicle to vehicle communications systems will move this technology ahead quickly. Combining these systems with radar should provide a means to overcome many of the radar only system problems of the past. As I said however, I don't see articles on this topic.

 

On a positive note, the IEEE Vehicular Technology society journal continues to publish articles in this field. The research is well funded. The corporations investing in this area must see some promise, or they would not continue funding. I expect in a few years advanced cruse control systems that incorporate location services and vehicle to vehicle communications. I don't see a day when you can take your hands off the wheel.

[npts2020]

waitforufo: It is not easy to accumulate information on this topic, partly because a lot of it is proprietary. One of the ways that the Darpa vehicles dealt with the misidentification problem was to have an array of (optical?) sensors to give better triangulation. One of the differences in the view I have and what you are talking about is that in my scenario the vehicles are basically drones that dont do anything until a central processor tells them what to do. This will allow traffic to be better coordinated. If the system is enclosed only rudimentary avoidance technology will be necessary anyway. RF tags could be embedded in the roof, along with solar panels and wind mills. It unfortunately seems likely that the federal government is the only entity that could ever finance, complete, and take on the liability possibilities of any such project.

[Flashman]

Call me a Luddite, but I'd be deeply suspicious of most automated systems, even ABS tries to kill me practically every other time I drive a car with it. Driving completely safety is as much about prediction and projection as it is about reaction, I think it's an AI complete problem. Thinking of some of the accidents I was nearly in, I'm thinking that about half of them could have resulted in impact with any automated tech foreseeable in the next 5 years. These were situations where an absolute minimum of clues were available, just the tilt of another driver's head, the slight angle of another car in it's lane, seeing an object rattling loose on a truck in front, hence I performed almost gut feeling instinctive reactions that kept me out of trouble, possibly up to a full second in advance of what any reactive system would be able to achieve. I don't see an automated system being able to react to mere nuances and predict behavior like that. It would wait until vehicles were actually out of their lane, or objects off trucks actually hit the roadway.

 

However, I could see the development of a kind of "roadtrain express" lane on highways/freeways, where you could link up to a close coupled group of vehicles, maybe even docking magnetically, and being pulled by a master unit tractor that had inductive pickups or something to run electrically. To join up, you'd match speed in the next lane over 100ft behind maybe, switch to auto, and it would join the train. This lane would be exclusive to the system and would have sensor networks along it in case there was accidents that blocked the lane, allowing forewarning to the master unit.

[Mr Skeptic]

I think that most people (myself included) would find it unacceptable for their car to be controlled by a remote computer. There are just too many things that could go wrong. An onboard computer would at least prevent someone from crashing every car on the network, or from the cars being lost if someone sabotages the transmission.

[npts2020]

Well the humans on the space shuttle give up control of their vehicle every time they land it and to my knowledge computer difficulty has never been an issue. At any rate it seems unlikely that free driven cars will ever completely go away, even if the government stopped building and maintaining roads for them (e.g. NASCAR). It has only been the last few years that computers have been able to consistently beat the best human players at chess, thereby surpassing the raw computing power of the human brain. This plus the fact that computers have been shown to be much faster than humans for reaction times, makes me skeptical that any human could outperform a well designed machine in avoiding an accident. When one thinks about accidents, I don't think anyone would argue the vast majority of them are caused by distracted or intoxicated drivers (90% sticks in my mind but am not sure how good that figure is). How often would a computer be distracted or intoxicated? Many of the other things that cause accidents should not occur on an automated system. If someone is sabotaging a car they will be caught on camera doing it, if every car has them for an avoidance system of some kind it is a relatively simple matter to also use it as a surveillance camera as well. Carrying an uncovered or unsecured load of any kind should be impossible if the vehicle encloses them. Furthermore, freight would only use the slowest lanes of travel (still more than fast enough to cause a serious accident) and could require inspection, permit, or whatever if the freight is not completely enclosed. If vehicles self monitor performance and are regularly inspected, mechanical failures should almost never happen to the point of causing an accident. Nobody can guarantee that no accident will ever happen but it should be a much rarer occurrance than are major airplane crashes.

[big314mp]

Lot's of freight already requires inspection, so that isn't a radical change.

 

I think what flashman was objecting to, was that the little nuances of driving won't be picked up by a computer. Basically, human experience trumps raw computational power. And I'm inclined to agree with him. However:

 

1) If every car on the road is computerized, theoretically there shouldn't be any driving nuances to keep track of. One computer can just talk to another and be very explicit as to what exactly it's intentions are. npts already covered several aspects of this in his recent post.

 

2) The computers that beat Kasparov relied on...human experience. The programmers wrote code that allowed the computer to analyze human played chess games, and then decide what were the best moves. Basically, you program the machine to learn, and then you teach it. Instead of modeling road condition, you model a human mind, and then you get modeled human experience.

[npts2020]

What we are discussing in this thread, the control system, is pretty much the only technical obstacle remaining for automating highways and railroads. My contention is that we already have the technology, it just hasn't been used for this particular application. The automated systems being built and existing now (I'm thinking Heathrow Airport and Morgantown, W. Va.) are interesting but IMO will never be practical, as they are, on a large scale. Neither system is likely to ever achieve the speeds I would like to see on a national transit system, the vehicles on both are unnecessarily large for personal transit, and I find it hard to believe that either system would be able to handle the volume required (single tracks), if the vast majority of travellers use it, without major modification. My quest is to figure out the best way to modify those current systems for widespread application.