The incredible challenge of designing self driving cars

[FrodosPet]

Humans look at things, and based on a combination of appearance, context, and prior experience, are good at understanding what they are and how they may affect their moving vehicle.

Computers can make educated guesses about objects based on those factors. But keep in mind that the world is made up of a lot of different things with a lot of different shapes and colors and lighting conditions. To be able to figure out what that constantly shifting blob ahead of you is going to do takes enormous processing power. First you have to find areas and edges and key points. Remember, at a freeway speed of 70 mph (112.654 kph), you are in constant motion, travelling at 102.667 feet per second (31.293 meters per second). The car has to understand what in the world is going on in a wide cone ahead. It needs to understand the world, track potentially hazardous objects, such as vehicles, people, animals, and debris, and calculate an optimal path at least 20 times per second, which is still 5.133 feet of travel.

Optical systems (cameras) face a lot of challenges in performing their part of that task. Dust and dirt either in the air or on the sensors. Fog, darkness, rain, snow, shadows, lens flare, washout from headlights and streetlights and the sun all make the challenge of detecting shapes based on like colors and edge detection even more difficult.

The consequences of design flaws and inadequate resources are major. We need to get this right as much as humanly possible.

[Binkie The Clown]

1. Put them on one-way rails enclosed in tubes.

Put east-west tubes on a different level than north-south tubes.
There are no intersections so no red lights to run, or T-bone collisions.
Head-on collisions become impossible.
Collisions with pedestrians become impossible.
Running off the road in into a building becomes impossible.

[brush]

5. Too costly, especially considering all the infrastructure improvements already needed

[FrodosPet]

2. Chris Urmson: How a driverless car sees the road

[longship]

3. And fucking snow! And dirt roads!

Both of which are common where I live and both of which any self-driving car would fail on.

Grass Lake Road alone would have any self-driving car end up IN Grass Lake!

[Logical]

8. You way underestimate how technology advances! Nt

[longship]

10. Sounds like Ray Kurzweil.

Yet another Matrix wet dream adherent.

People over estimate future technological projections.

The first death by driverless car occured this past week. I know, we've learned by our mistakes and can repeat them exactly.

I want the next DARPA challenge to navigate Grass Lake Road, with random oncoming vehicles. In fucking winter!

We drive on these roads for years. The reason why nobody dies on them is because humans have much better cognitive abilities than limited and silly silicone.

Give it up. I want to see a driverless car navigate Cajon Pass during a white out, not uncommon in winter.

Then you can claim victory.

Until then, you have nothing but wet dreams.

Winter is an utter deal killer for driverless vehicles.

[Adrahil]

18. How many does in human-driven vehicles.

And many of them dies due to human error. It is tue that people are WAY WAAAAY more tolerant of accidents caused by themslves than they are of accidents caused by automation errors, but as someone who works in the automation field, I think you have no idea what's possible and what's coming.

[longship]

19. Extrapolation is just wrong.

What technological advances will occur is not known by anybody and only invites being found foolish when one makes such predictions.

Duplicating what the brain does while driving an automobile is at best orders of magnitude more complex than what any putative technology can accomplish today.

Note that all tests are being made on well paved and dry roads either in urban areas or interstate highways. The conditions are no where near the conditions that people here drive in, especially in winter when the roadway is not in any way visible, or even rain storms.

And the narrow unpaved roads here are no place for autonomous vehicles, even in good weather.

The burden of proof is on those making the claim, not the one refuting it.

I stand by my posts here.

[Adrahil]

20. Yeah, we engineers have no idea what we are talking about...

It's all magic and nobody knows what's coming.

Sheesh!

[longship]

21. I see no reason to debate this further.

My claims stand. Autonomous vehicles cannot handle winter let alone unpaved, narrow country roads.

I will modify my claims if I see evidence to the contrary.

And no, I don't think you know what you are talking about.

So let's just leave it there.

My best regards,

[Adrahil]

56. You're right, with 30+ years working in vehicle automation...

I have no idea what I'm talking about.

[longship]

62. With five decades of computer tech experience...

I just want to just ask you to show me the successful winter test drives of autonomous vehicles, or those on curvy, narrow, rural, unpaved roads.

And the DARPA Challenges don't count, as they are more properly characterized as pure research, not practical.

That's right. There aren't any, at least none that I've ever heard of. I will change my position when there are, however the acid test has got to be driving over the Grapevine (Tejon pass, I-5) during a snowstorm. I don't think that would be allowed.

I stand by my posts until proven wrong.

[Adrahil]

64. Yeah... So it must be impossible, right?

I work mostly in aviation R&D. I have personally been involved with 11 projects that culminated in doing something that had never been done before. That's what R&D is all about.

Winter driving will not be the biggest challenge for automated cars. Not by a long shot.

[longship]

67. No, it is just fucking difficult.

Difficult enough that it isn't going to happen very soon. The snow alone would be the end of autonomous vehicles here.

I have not seen a winter DARPA challenge. So they aren't even to square one on the winter problem.

Good luck with getting a vehicle over the Grapevine (Tejon pass, I-5) in winter.

[Glassunion]

68. Not square 1... But there is still work to do.

[longship]

71. On a well layed out highway, not a country road.

Where the whole road can just disappear under the snow.

So far, your autonomous vehicles apparently cannot even tell the difference between a horizon and a white sided semi-trailer.

Sorry, I don't buy that we are going to have autonomous vehicles anytime soon.

I wish you luck, however.

[Glassunion]

72. I think we will within about 10 to 15 years.

It was not an autonomous vehicle that struggled with the horizon and a white sided semi. The Tesla is not an autonomous vehicle, instead is uses a driver assist. Basically, the driver should have also been paying attention to the road as well. Just like with adaptive cruise control, you should still brake when the car doesn't.

Anyway, back to my 10 to 15 year feeling on autonomy. I feel that it will not start as full autonomy, but piecemeal features gradually being introduced year to year as driver assist features. There are technologies of the past present and future that will all contribute to an eventual autonomous car. There are always limits to technology, but there are also limits to human ability. The trick is, if we can get an autonomous vehicle that will not kill over 30 thousand people, and injure over 2 million a year. If those numbers can be reduced via automation, then I will personally count it a success.

Technology can have a greater sense of surroundings, and obstacles when compared to a human. However the human can deliver a more effective driving strategy on those surroundings that it is aware of. We have a better ability to gauge stupid and prioritize our surroundings. Example: I ride a motorcycle a lot, and I assume any car turning left in front of me, or at the intersection is going to pull out into my path. Simply because folks are not good at being aware of their surroundings and seeing a motorcycle. So, I hover my right foot and hands over the brakes, and my left hand over the clutch because I'm anticipating their stupid, and my possible need to slow or maneuver quickly.

We may have autonomous vehicles in the near future, however they may make you drive up and down a dirt roadway, or when the road conditions are not satisfactory. Then again, these systems might also learn and share information so it knows where the road is even when snow covered because of all the other vehicles that have driven and mapped that road while conditions were optimal.

[longship]

74. How are they going to share information?

The only INet here is cellular and dial-up. The cellular is dodgy here, and I presume that you aren't suggesting that we string landlines to these vehicles, not that dialup would have sufficient bandwidth, let alone cell service here.

The big elephant in the room is that, in order to have a national system of highways safe for autonomous vehicles, there are going to have to be some rather large infrastructure changes. And in winter, things can go screwy rather quickly. One car driving autonomously on a well-defined highway does not change that.

Try Tejon Pass in winter for a real challenge. That's I-5 in SoCal, often called the Grapevine. Don't forget. There are many very large trucks on that route. I don't want an autonomous vehicle anywhere near that place when I drive on it. The Interstate is admittedly very wide there, but nevertheless it remains a very scary place to drive. In winter, it is outright treacherous.

My best to you.

[Glassunion]

76. Data can be queued, and uploaded when a good signal is available.

Could basically use a network similar to, or piggyback on networks like ON Star, Sync, BMW Assist, MBrace, etc.

I'd not count the entire success or failure of a system based on a 40 mile pass on a single road, out of the near 4 million miles of roads here in the the U.S. I'd have concerns about other passes as well. We used to do the Donner Pass to get to Lake Tahoe from Sacramento, and it can be insane in the winter as well. It once took us 7 hours to get over the last 3,000 or so feet of elevation. These things can be overcome, and perhaps in the winter, when the Tejon Pass is actually open, they may make require all vehicles to not use their automation features.

Weather is not a factor in 74% of motor vehicle accidents. Snow and Sleet only factor in for 2.7% of accidents.

[longship]

77. On the Tejon or Cajon passes, weather is a huge issue.

And on any damn Interstate highway from coast to coast. If autonomous vehicles cannot handle it without phoning the mother ship they don't belong on the highways. What happens if the mother ship doesn't answer? That's right, you could be screwed. Hopefully it won't happen while you're traversing the Grapevine.

The principle "fail safe" comes to mind here. Presume that you have no comm link. What happens then? I suspect mayhem. Either that, or go to manual controls.

This has turned into an interesting discussion. Thank you.

BTW, everybody ought to experience the Grapevine once in their lives. It is an astounding engineering job. But scary as all Hell with all the damned huge trucks and heavy traffic. I don't like it much.

[Glassunion]

80. Phoning the mother ship, or not being able to connect would not be an issue in my scenario

that I brought up. Basically it is simply to update system maps. Sort of like my Google GPS. The GPS is in a constant state of updating, and gathering aggregate data. This updates the system to give up to date traffic information. If the last 50 cars to drive down a road are going slow, it will update the system to notify of traffic. This system that Ford is working on, is basically one that creates 3D maps of the road, that can in turn, update individual systems for use when conditions are less than perfect. The more vehicles drive down a road, the more accurate the autonomous system gets. Basically building a 3D map of a road for that vehicle and others to use. Sort of like how you drive down a road 10 times, and you know exactly where the potholes are, or where the edge of the road is even when snow covered, because you're using the snow poles, mailboxes, and street signs to approximate where the road surface is. If the first 10 cars create such a map, all others vehicles can use it as well as expand upon it. The vehicles would store this in it's memory just like you would after having memorized the road and its conditions.

Redundancies should be built into any automated system. I work with systems all day, and none of them fail to utilize a redundancy of some sort of another. I'll give you an example. My car has sensors galore in it, that control several different things. I have speed sensors that take measurements along the crankshaft, torque converter, transmission, driveshaft, differential, and finally wheel speed. If any one of them do not line up, the vehicle makes an assumption based on whats happening. It may assume I have lost traction in one wheel, or my transmission may be slipping, etc. I recently had one of the sensors on one of the two rear axles that finally STB (got 160k+ miles on the car), and my dashboard lit up like a Christmas Tree. The effect of which was not catastrophic, however the vehicle went into a core fail safe mode. I had no ABS on only the right rear (I noticed when braking hard), and my ATS/DTS system shut off, and the transmission would not utilize sport mode. However the vehicle drove completely fine for the next few days while I waited to get the service scheduled.

You'd have the same fail safe modes in automated systems as well. If the vehicle has a failure, it will perform a series of built in, predefined functions, compounded with the redundancies built in, you'd simply get a warning to service the vehicle to repair the failed component, while another part of the system could pick up the slack of the failed component. Like overlapping fields of view on the external sensors. So if a sensor fails, the two on either side of it will continue to scan the area of the failed sensor, and the vehicle will continue to operate as normal, or in a more critical failure, safely move the vehicle off the road or until the driver assumes control.

But the basic command and control systems of safely operating the vehicle would be internal, and not reliant on a connection to any sort of external system. It would supplement its core knowledge by periodically learning new information from that external system. Basically a learning system, that does not use a single data set, but hundreds or maybe even thousands to gain an accurate picture of the road, its conditions, or its quirks.

[longship]

81. You have some good points there.

But GPS would be useless where I live. Many of the roads are not well mapped, and few are paved. Grass Lake Road is treacherously narrow and so close to the lake that no GPS is going to keep you from driving INTO the lake on one of multiple tight curves. And if an oncoming car approaches, what does your automagic car do then? I assure you it will be wrong.

People here take Grass Lake Road because it's a very pretty drive right through the national forest. There are houses there, too. People live there. It is a rather short cut that cuts out a several mile round about way on the also narrow paved roads here.

What I am saying is that there will be no autonomous vehicles here for some time.

Early adopters will be in the always warm CA valley cities, mostly the gentrified areas where people are wealthy enough to afford such a luxuries. But they will not go auto over the mountain passes, especially in winter. If they try that, good luck to them, the damned fools.

Once there is experience with them, there may be expansion into other areas. But I don't see it becoming wide spread quickly. It will take decades.

That's my opinion.

[Glassunion]

85. Autonomous vehicles do not rely heavily on GPS.

Which is a good thing, as civil and commercial GPS is only accurate about 95% of the time to within 3 meters. 3 meters is huge if you are driving, as that could easily put you in the oncoming lane. Currently, fully autonomous vehicles (Tesla is not) use a myriad of technologies to achieve the goals of driving.

One of the fail safes built into the current vehicles, is that the driver may have to take control in certain circumstances. From my understanding the Tesla (semi autonomous) crash was caused by the software being confused by a white truck on a white sky background, as well as the driver not paying attention. The Tesla autopilot feature is still in beta testing, and drivers need to be aware that they at a moment's notice have to take control of the vehicle. I think Tesla made this clear.

Anyway, the software confusion of white on white, tells me that the system was simply relying on a camera system to keep the vehicle in its lane and to watch for obstructions. In vehicles that are fully autonomous, like the Google vehicle, they not only leverage cameras, but they also use lasers and radar to gain a more complete picture. These systems as a whole will paint a picture of a given road within an accuracy of centimeters. Initially, autonomous vehicles would not drive down unpaved, unmarked roads until it has an accurate picture of the road. As each capable vehicle is driven down a road by a driver, the cars will gather 3D images of the road for eventual autonomous use. I'm sure there will be some roads that no autonomous vehicle will be able to drive itself down.

This however will not be a measure of success. Success will be measured by a reduction in motor vehicle accidents and deaths. Just this morning, as I rode my motorcycle in, a van pulled right out in front of me. I saw her approach the stop sign, come to a complete stop, look right in my direction, and then pull out onto the road right in front of me. I went from about 55 to 25 and dropped down 3 gears as quickly as I could, but when I ran out of braking room (was going to hit her), I leaned it over into the empty oncoming lane, and swerved around her. Now if her van had been equipped with technology currently available, her van would have seen me where she did not. It could have warned her that she was about to pull out in front of a vehicle that was at speed, or even forbade the vehicle from pulling out into the road all together avoiding what could have been an ugly accident. If there was an oncoming car, I would have been left with the only option of hitting her.

You see road conditions and weather play such a small part in causing accidents. Of all accidents (2.1 million) where there was a fatality or an injury, the critical reason that the accident occurred, Driver Action accounted for 94% of accidents, Vehicle malfunction or improper maintenance accounted for 2%, and lastly the physical environment (weather, road conditions, etc.) accounted for another 2%. The remaining 2% were unknown.

Now in the Driver Action category, a recognition error was the primary reason (41%). Either the driver was inattentive, distracted, or failed to properly survey their surroundings. The next highest was decision errors, which account for 33% of these accidents. This would include speeding, or too fast for a corner, falsely assuming others' actions, making illegal maneuvers, or misjudging a gap or speed of another vehicle. The remaining primary causes were performance (11%) and non-performance (7%) where a driver would over compensate, have poor directional control, or non performance, where the driver was sleeping or tired.

These are all things that technology can eventually address or already do currently. Taking a look at what exists today in newer vehicles: collision avoidance, lane awareness, blind-spot monitoring, driver monitoring, etc... is already having a positive impact on reducing collisions. We cannot make folks drive better, so if we continue to work to take the driver out of the equation, this will have a huge impact on reducing collisions.

[longship]

86. Very good, but we won't see auto vehicle commonality for many years.

They will be limited to places with clement weather and no winter. The desert Southwest, including CA valleys. Also, auto vehicles are likely to be priced rather high, like today's EVs.

And again, they will never be useful where I live.

[Glassunion]

87. I'm sure the initial to-market products will be expensive.

Technology always works that way.

Like I said in earlier posts, I think they'll overcome the issues with snow no problem using internal 3D mapping systems. The folks at Google say that they should have the snow and heavy rain issues sorted out within the next 3 or so years.

The folks at Google are aiming for a 5 year deadline to have the issues resolved, tested and ready for sale to the public.

[longship]

88. Well, as I wrote, many roads here are not well mapped.

This is a national forest area. No cities nearby. The closest large town is 15 miles.

The roads here are well serviced, just not well mapped. I don't know how an autonomous vehicle is going to be able to stay on the right of way by using 3D mapping, when there is none here. And I stand by my claims on winter. Autonomous vehicles are screwed in a snow storm.

Best regards.

[Glassunion]

90. They can get "mapped" easily enough.

I'm not talking GPS maps, or even the old Rand McNally county level Atlas, or even the USGS topographic maps. I'm talking about user (vehicle owner) created maps.

If you have a vehicle that is equipped to be autonomous, it will have all of the sensors required to create a map. So you (the driver) are not using the autonomous features of the vehicle, and are actually in control of it. You make use of the road, as do 5 other equipped vehicles that day. Now your vehicle, even though you were driving, as well as the 5 other vehicles used their sensors to passively create a 3D map of not only the road, but every sign, pothole, tree, intersection, utility pole, etc... Then the next time all of your cars get a wireless signal, it uploads that information to a central location, where the 6 maps are combined into a map that is accurate within centimeters. The more vehicles that travel down that road, the more accurate it gets. Now in a snow storm, the vehicle does not need to "see" the road surface at all. It's internal map already knows exactly where the road surface is because of the vehicles orientation based on the things it can see, such as mailboxes, utility poles, etc. all of this to within centimeters. Which is far more accurate than a human operated vehicle can perform under the same circumstances.

[longship]

91. I have no interest whatsoever in having an autonomous vehicle.

And those trumpeting them are making all sorts of presumptions which will likely not pan out as they presume.

In other words, I am against them being put on the roads at this time. Except for research, obviously.

[Glassunion]

92. To each their own.

I have great interest in having autonomous vehicles.

I admit, that I cannot be certain the outcome until we get there. But at the end of the day, I feel this technology will save lives, prevent injuries by reducing accidents. You cannot make people drive better, however you can make a computer drive better.

[longship]

93. As long as it isn't running on Microsoft!

Make it Linux!

[Glassunion]

94. Agree 100%

Thankfully: https://www.linux.com/news/linux-leads-self-driving-car-movement

[A HERETIC I AM]

111. Drove down it yesterday in an 18 wheeler

Frankly, the Grapevine ain't squat! Try Cabbage Pass in Oregon on I-84. 7 miles of 6%. Or Vail Pass on 70 in Colorado. 7 miles of 7%. That's about as steep a grade you can find on the interstate system

The grapevine is marked as 5 miles of 5%, but it's closer to 6%

[The Second Stone]

112. The Grapevine is quite a feat, but nothing compared to 80 over Donner Pass

Donner Pass also has, parallel to 80 in spots, the Central Pacific's route across the Sierra. These were stunning engineering feats.

[longship]

114. I've driven on both of them.

But I found the Grapevine quite terrorizing mainly because of the crazy heavy traffic. White knuckles all the way.

At least the Donner Pass has the decency to have somewhat lighter traffic. Just don't want a big truck behind me.

[Glassunion]

61. There are already automated vehicles in use on narrow unpaved roads.

Particularly in mining operations.

[Egnever]

51. Sorry but that just is not true

here is video from 11 years ago of the darpa challenge where the goal was 130 miles of anything but dry well paved roads. We have come a long way since then.

[longship]

63. Watched that one. Those were research vehicles.

Many did not finish. They exposed many problems.

Again, a good portion of the country experiences something called winter, when one has snow and ice covered roads. Where I live the roads, mostly unpaved, are narrow and have no shoulders.

There's no way in Sam Hell an autonomous vehicle is going to know where to go.

I will stand by my posts.

[Egnever]

70. Those were research vehicles 11 years ago

We have come a long way in 11 years.

You can stand by your posts all you like it wont change the reality that these cars are coming.

Snow..

https://www.wired.com/2016/01/the-clever-way-fords-self-driving-cars-navigate-in-snow/

[Travis_0004]

101. You don't think auto makers can improve on college kids 11 years after a challenge?

Yes, there were many problems. Self driving cars were not ready in 2005. They are not ready in 2016. And I get it, you have unpaved roads, and snow. If you can figure out how to drive in the snow, so can a computer, and in time do a better job. It might take 5 more years, nobody is saying the technology is ready today.

[longship]

102. I get it, but there are quite a few problems left to solve.

And then there's the problems that early adopters will still have to deal with all the human drivers on the road.

And the cars communicating with each other won't work until the roads are filled with autonomous vehicles, or are at least more common.

Five years is far too short. IMHO.

[Logical]

28. LOL, people ike said man would not fly for 1000 years 2 years before the wright brothers did......

if you need some other examples let me know.

Technology waits for no man.

[Hassin Bin Sober]

98. First, the guy that was killed was in a driver assist car, not driverless. He fucked up.

Not to mention had the truck had the same system, it would have seen the car and the accident wouldn't have happened. The human professional truck driver REALLY fucked up and may be guilty of reckless homicide.

Second, you take the most extreme example of some pass with whiteout conditions and claim failure for the whole concept...

... When the vast vast majority of driving, and people killed while driving, is done in normal run-of-the-mill vanilla regular street and highway driving.

Doesn't the program make sense if three hundred people run off the road on some winding road in bumfuck Michigan or Moosejaw Alaska, while 20,000 or 30,000 people are saved by run of the mill "dumb" accidents?

Thirdly, this (see below) is beta version of driver assist that doesn't even claim to be able to handle rural roads, yet it does - for the most part.

I've driven in whiteout conditions. Once caught by surprise on the way to Buffalo and another foolish decision to run to the Southaven Michigan Menards during a blizzard last year. If I ever have the unfortunate occasion to do it again, I hope it's in a driverless car that has the road mapped out to the square inch AND has radar to avoid the cars ahead. I know those roads like the back of my hand and was in a 4WD truck. The problem was the other cars spun out and/or stopped... And the car that just about rear ended me when I stopped to avoid the stopped spun out cars.

I don't know why you think your memory of twists and turns of your local roads will be superior to the memory of a gps system that's been down that road. Or why you think your eyes will work better in whiteout than radar with positive recognition of other cars. Or why you think your management of steering, skidding and braking on slippery surfaces will be better than a computer that thinks thousands of times faster than us. We are already halfway there on current vehicles with antilock brakes, traction control and stability control.

https://m.

[brush]

4. I don't get why it's needed. Too many road condition variables, plus. drivers like to . . .

drive and do it better. Those who don't like to drive, other transport methods are available.

[Travis_0004]

6. Drivers don't do it better.

Perhaps they do it better than some cars that are still in development, but self driving cars will easily be better than humans.

Also, I think a lot of people want self driving cars. Not necessary everyday, but for a lot of the time.

I have to drive about 6 hours later this week. I would much rather hop in the back seat and relax if it was possible. If its raining, I can have my car drop me off in front of my work, instead of 1/2 mile away where I park. If you have teenagers, (14-15), you could let them drive themselves to a friends house.

There are lots of benefits.

[brush]

14. Your post sounds like musings about flying cars we used to hear back in the day

A Tesla driverless car just killed its rider because it couldn't recognize a tractor trailer and drove right into it.

They are years away from being perfected.

And we still don't have flying cars.

[Glassunion]

22. And 3,286 people were killed the same day by divers.

[brush]

23. I'll take my chances driving my own car.

Tesla, the leader in this area, tested a driveless car and it killed its rider. Not good. One test, one death.

Driverless cars, like flying ones, are still years away.

[Glassunion]

24. That's fine.

You are correct, it is years away. However I feel the driver-less car will probably be a fixture within the next decade. I personally cannot wait. I trust my own abilities behind the wheel, but I trust other drivers as far as I can throw them. They kill 10's of thousands a year, and injure more than 2 million people a year.

[Orrex]

27. It's not clear that the car was at fault

http://www.democraticunderground.com/10027976327

Of course, having to copilot one's "driverless" vehicle sort of defeats the purpose, but if the vehicle performed as it was programmed to do, then the fault may lie in the driver's failure to meet the performance requirements.

[Logical]

29. give me the number of years you think. I would love to bookmark it. nt

[brush]

34. With the number of driver cars still on the road, with tens of thousands continuing . . .

to be made, driver resistance and bad publicity like the Tesla driverless car death, I'd say a generation before they are even a fourth of the cars on the road.

[Logical]

35. LOL, 1/4 would be amazing in 10 years. So you are thinking it is possible. thank you! nt

[brush]

38. A generation is usually considered to be 20 years.

[Logical]

42. In 20 years every new car will be self driving. Maybe in only 10. nt

[brush]

45. That's only an opinion, and not backed by any actual data or facts.

[Travis_0004]

25. the tesla is not a self driving car

Its a level 2 on a 4.0 scale.

It has driver assist features, but is not intended to be self driving.

The tesla may have missed the semi, but so did the driver. Perhaps he assumed the tesla is in control, but that shows that drivers sometimes lack judgement as the car is not self driving, and the driver should always be in control.

[brush]

26. What? You mean there's margin for error and doubt as to whether the driver or the artificial

intelligence is in control at a critical moment?

Hah!

I'll keep driving.

[Logical]

44. "I'll never giv up my horse and buggy". LOL, 33,000 americans die because of driver error a year....

but something tells me you had no idea about that stat.

[brush]

49. You actually think people don't know people die in auto accidents?

You think that's a newsflash?

[Logical]

78. Apparently you think 1 death vs 33000 if an issue.

[brush]

82. Yeah, go with that. Please proceed, governor.

[Logical]

84. LOL, yea, you are making quite an impression. nt

[Logical]

32. Wow, how many companies are working ob flying cars vs. driver-less ones? Do more reading! nt

[brush]

36. Try re-reading my post. I reference stories "back in the day" that promoted flying cars.

[Logical]

39. "back in the day" 20 companies were not working on self driving cars or flying them 15,000.....

miles a week like Google is.

Wow, are people so uninformed on this whole area.

[brush]

43. Obviously you are "informed" on it, for what that's worth.

That doesn't mean driverless cars are going to be dominant on the roads anytime soon — driver resistance, bad publicity from deaths in testing, much, much more testing and perfecting needing to be done, hundreds of thousands of driver cars still on the road and tens of thousands of driver cars being manufactured every year and for years into the future.

They are not just around the corner by any means.

[Logical]

46. On the lots in 6 years. Bet on it. nt

[brush]

47. Could be. Dominant of the roads in 6 years, not hardly.

[whatthehey]

37. You've been corrected on that lie before

Tesla's Autopilot is not considered "driverless" or even approaching it by anybody with two brain cells to rub together.

[brush]

40. Why call it a lie? The artificial intelligence was not overridden by the co-driver . . .

to stop it from driving right into the 18-wheeler.

So actually, the AI was driving at the time the co-driver was killed.

[whatthehey]

48. Because it's a lie. Autopilot was not designed, described or implemented as being driverless

or anything like it. This is no more a "driverless" car than cruise control which has been around for decades and will also run you into a truck if you are fucking stupid enough to ignore Tesla's clearly written instructions to stay alert with your hands on the wheel at all times. This is nothing at all like Google's autonomous cars which have gone well over a million miles with only one even questionably at fault minor scrape. All Autopilot is is a combination of lane awareness and adaptive cruise control, two very familiar technologies.

[brush]

50. Who needs it then if you still have to "drive"?

[whatthehey]

55. Actually there I concur. Don't see the point.

A lot of Tesla software updates are kind of silly to be honest. But this was not even an attempt at at delivering a car anywhere approaching driverless. The S simply does not have the sensors that Google cars use, and they are a few years away from deployment even so. Idiots killed themselves when cruise control first came out too. This is nothing more than a redux.

[Logical]

7. Lol you sound like "I want to keep my horse" people when the car came along

[brush]

12. I would like to keep my car. Yes. Especially since a Tesla driverless car just killed it's rider.

[Logical]

30. LOL, And drivers killed over 100 the same day. Nice try. nt

[brush]

31. One driverless car, one killed, not good odds. Like I said, I'll keep driving.

[Logical]

33. One, have you fucking researched this at all. You are extremely uninformed on this topic......

From googles monthly report......

"We’re currently averaging 10,000-15,000 autonomous miles per week on public streets"

So your "One driverless car, one killed, not good odds" comment is silly and a fly out lie.

[brush]

52. Obviously this is your "thing". It's not mine and most other peoples

If driverless cars prove viable in the future they'll make inroads, but many will still prefer driving.

Why so argumentative about it? They are not hardly about to take over the roads.

[tonedevil]

53. In all honesty...

the one your are replying to is exactly the kind of ignorant jackass I want to see stopped from driving.

[brush]

54. Heehee! I don't want him near a co-piloted car either

If they ever perfect actual driverless cars, maybe.

[Logical]

79. :-) nt

[Odin2005]

11. "other transport methods" in big cities, maybe.

Unless you are in a large city those of us who can't drive are effectively 2nd-class citizens. Self-driving cars will be a god-send to those of us in Flyover Country who can't drive.

[Alex4Martinez]

15. Corporations want it. They're fighting transit, want perpetual private mobility.

The automobile era is over, traffic is incurable, they know that building more lanes doesn't work so now they want to increase capacity by cramming more cars into the lanes we have.

Transit would be so much better, rail transit and streetcars and transit oriented development could lead to very few people owning cars.

They hate that model, need to respond with technologies that will force people to continue being in debt.

And the government helps them with studies and research grants.

It's bullshit.

[brush]

16. I agree that better mass transit development should be the way to go

[Albertoo]

9. My spouse is scared to death of self-driving cars

Which I find funny, considering a human is more fallible than an AI system.

[brush]

13. The Tesla driverless car just killed it's rider.

[Meldread]

57. That's incorrect. Tesla's car is not driverless.

You are required to remain alert with your hands on the wheel at all times. If you take your hands off the wheel, the car notifies you that you need your hands on the wheel.

Your statement that it is a driverless car is like saying cruse control makes a car driverless. It's designed as an assisted aid (with the eventual goal of becoming driverless), making it effectively no different than cruse control.

[brush]

58. Who needs it then if it's just re-named cruise control?

[Meldread]

59. Everyone.

It's designed to assist in driving, and if used properly, it will reduce car accidents. It is a step toward TESLA developing a fully driverless vehicle. Google is effectively already there, as their technology has driven over one million miles--more than the average person drives in a lifetime--and has only had one very minor accident. There are over six million car accidents each year, with 30% of those being fatal. The average person will be in one significant car accident roughly every 20 years.

Your statement about 'who needs it' is ridiculous. It is like saying who needs seatbelts, when seatbelts don't prevent deaths in every car accident. No, seatbelts don't, but they significantly reduce car accident deaths--along with airbags. This autopilot technology by TESLA will further reduce car accidents, and thus motor vehicle injuries and fatalities.

[brush]

60. My statement is "ridiculous" when it's only a transitional step to the real thing?

I think your statement is ridiculous. Why would anyone need a step towards a fully driverless car if Google is already there?

[Meldread]

65. Because TELSA does not own Google's technology.

Google has been perfecting their technology for over a decade now.

At the moment Tesla is able to collect data and improve the technology, which is still in beta testing, but is considered safe enough for actual road usage. This gives them an advantage over Google, in that they are collecting data from actual real world usage, whereas Google has limited itself to more controlled settings. For example, their self-driving vehicles are not extensively tested in heavy rain or snow.

This is a middle ground until the technology has been extensively tested and proven more safe. It is roughly on par with airplane autopilot. Airplanes are mostly run by computers now-a-days, but pilots are still used in case of technology failure or an emergency. Tesla's car is similar. We have so few plane crashes now, as compared to the past, due to airplane autopilot technology largely saving us from human error--which is what causes the majority of plane crashes, and obviously, the majority of car crashes.

[brush]

69. My question remains. Why would anyone buy a car with technology that's not ready to market?

The fact that that one just crashed and killed the co-driver shows that it's no where near ready.

Everyone certainly doesn't need that.

[Meldread]

99. That's a ridiculous statement:

The fact that that one just crashed and killed the co-driver shows that it's no where near ready.

So, based on your argument, we should toss out autopilot in planes. After all, sometimes technology isn't perfect, it doesn't work as expected, and it fails, that's why the pilots are there. It doesn't matter that autopilot in planes saves thousands upon thousands of lives each year by eliminating and reducing human error. We should just scrap it, because it isn't perfect.

Why don't you just speak honestly and lay your cards out on the table. This has nothing to do with the one guy who died. That's just the talking point. You are opposed to the technology wholesale, and would be so even if it was a perfect driverless car that could NEVER experience a technological failure. It could be a car so safe that people regularly put their young toddlers in it and let it drive them to school, and you would still be opposed to it.

Maybe you hate the technology because you have a Luddite-mindset, or maybe you hate it because you fear the changes it will bring both economically and socially. I don't know your motives. However, at least be honest that you are opposed to the technology on principle, and drop this transparent charade that it somehow has to do with the recent Tesla accident. That is just so obviously false that it insults the intelligence of everyone here.

Your concern trolling is not amusing.

[brush]

104. I don't hate technology. Wouldn't buy that car with what amounts to cruise control 2.0 though

I'll keep driving until fully developed, exhautively tested and ready for market actual driverless cars are available.

I understand Google is further along on that than Tesla and their co-driver cars.

[Meldread]

105. Well, I'm glad you are not opposed to the technology at least.

You are fully entitled to wait. Tesla is really just one such company to begin rolling out this type of technology. I believe there are a couple of other manufacturers that have more limited capabilities, such as self-parking cars. However, over the coming years this type of technology will begin to be common in all new cars.

Already the technology exceeded human abilities at driving. As it does continue to develop, continue to be tested, and becomes even more safe I have little doubt car insurance companies will begin pricing fully manual cars into their policies--those who don't adopt will likely begin to be charged more. This would make sense, because of the higher rates of accidents. I predict that eventually states will begin mandating that all new cars sold be self-driving, for the same reasons that the states (and the country as a whole) began mandating seat belts and airbags.

This technology really is going to save thousands upon thousands of lives. People regularly over estimate just how bad drivers we are as human beings. Once the vehicles begin to have the ability to communicate with one another, traffic accidents between cars will likely become so shockingly rare it would be on the news the same way plane crashes make the news.

[brush]

107. All that's fine. Bad habit though of calling people's reasonable statements "ridiculous".

[Meldread]

108. That's because it is ridiculous.

You are setting an unrealistically high standard, and by high standard you are demanding perfection. You even used the word perfect yourself in response to someone else:

Think I'll keep driving until they perfect an actual driverless car.

Your standard is that no matter how good the technology gets, no matter how much better it is than human drivers, no matter how many lives it might save you will resist it until it is perfect. Of course, nothing ever created is perfect and flawless. Technology fails sometimes. That's life. Things break, things make mistakes. People make mistakes. The standard shouldn't be perfection, but rather who makes more mistakes and puts us at higher risk: people or the technology. The technology is going to win that battle every time. All people--including yourself--overestimate your capabilities as a driver. This is the reason we have so many horrible car accidents, many of which are fatal.

You are not refusing to get on a plane because the autopilot does most of the work, and there may be times that the actual pilot has to take over. You are not demanding that seat belts and airbags be taken out of cars because they do not have a 100% success rate, and even sometimes directly and unintentionally cause the death of the people they are trying to save. You are not setting the same standard for other technologies that already exist to fill the same role. That is why I called your statement ridiculous.

[brush]

109. It's ridiculous to expect technology to work? Gotcha.

[Meldread]

110. That's a gross simplification of what I said, and I will note that you did not deny the accusation.

[Albertoo]

75. After millions of miles. Better track record than humans.

[brush]

83. Link pls to show Tesla test car has logged millions of miles.

[Glassunion]

95. Link.

http://www.latimes.com/business/technology/la-fi-hy-tesla-google-20160701-snap-story.html

"Tesla drivers already had logged some 130 million miles using the feature before a fatal crash in Florida in May made it the subject of a preliminary federal inquiry made public on Thursday."

[brush]

96. Thank you. Guess they better try something else then since it still failed.

[Glassunion]

97. Depends on how you measure success.

Human operated vehicles kill 46% more people per mile, and cause 100% more injuries than this semi-autonomous vehicle. There has been only one fatality and no reported injuries in 130 million miles driven by their semi-autonomous system.

I would consider a 46% reduction in fatalities a small success.

[Albertoo]

100. It hasn't failed

It's safer to be in a Tesla car than in a car driven solely by a human.

Just the facts.

[brush]

103. It's sounds like nothing more than cruise control 2.0 since you still have to co-drive it.

I wouldn't go so far as to say it's safer.

Maybe we can ask the driver who was . . . oh, wait. We can't. He was killed.

Think I'll keep driving until they perfect an actual driverless car. I understand Google is further along with that than Teala.

[Albertoo]

115. You asked for the link: the stats prove it's safer with the Tesla system than without

[FrodosPet]

17. Self driving cars do certain tasks better

In a relatively sane, controlled world without obstructions, blind spots, etc, they can do a good job.

The trouble is when the world gets dark and/or noisy.

And a lot of the issues slowing them down are non-technical. They will have to prove themselves MUCH safer than humans before people will trust them en masse.

Many people will NOT want to give up driving. And financially, it would take years, decades, before existing cars can be replaced.

They are definitely coming, and they can be a liberating technology when they do. But it is going to be a slog to market availability and success, not a quick dash.

[Meldread]

41. Self-Driving Cars do not need to be perfect. They just need to be marginally better than you.

Human drivers are so horrifically bad that technology, at its current level, can easily surpass them. Google's self-driving vehicle has driven over one million miles, which is more than the average person drives in their entire lifetime, and has only had one very minor accident.

For comparison, the average driver is in one car accident roughly every 20 years. There are over six million car accidents per year in the United States, 30% of which are fatal.

Self-Driving cars do not need to be perfect. They just need to be better than that. Over time it will continue to improve, like all technology, and the cost of car insurance will take care of the rest.

[Kaleva]

66. Self driving cars may keep a check on human population growth.

[matt819]

73. Factor in other stuff

Not original. Just read a sci-fi book that included time travel and self-driving cars.

How to deal with human intervention in a self-driving car. Who has control?
Liability - carmaker, software maker, human operator
Software upgrades - when done, effect on operation of vehicle
Interaction with other self-driving cars
Identification of self-driving vs. human driven cars so that other drivers on the road know what's around them

[hunter]

89. People like to imagine they know what they are doing most of the time.

But most of the time we're running on autopilot that's not much different than a self-driving car.

The self driving car has the advantage that it doesn't get distracted, doesn't get sleepy, doesn't drink alcohol or use other mind altering drugs.

To be successful, a self-driving car has to be statistically less dangerous than a human driver.

That's a pretty low standard.

Humans are terrible drivers; even "good" drivers have accidents.

I think automobile use and ownership ought to be discouraged.

We ought to be building a society where most people find automobiles unnecessary in their daily lives.

[Rex]

106. Hard act to follow, need something on par with the human brain. nt

nt

[Humanist_Activist]

113. No you don't, the human brain is too complicated for the tasks at hand...

slowing down reaction times and becoming too distracted by its own stuff.

A computer never gets tired, never gets distracted, never loses itself in its own thoughts, doesn't get angry, doesn't get depressed, etc. That alone certainly puts them above humans in at least potentially being better than us as drivers.