How do self-driving cars ...

[twobjshelbys]

Riddle me this batman.

How do self-driving cars deal with events that require "judgement"?

4-way stops (with the SDC going all directions, turning right, straight, turning left - each has different rules)

Flashing red stop lights (different than a 4-way stop sign), ie, a power failure has all sides blinking red

and my big one:

Left turns at light controlled intersections with and without turn lanes and with and without arrows, and without any traffic control devices at all. In fact the left turn conditions are really highly variable with medians and without. In fact, there are a few left turns we use regularly that defy logic as they are near a freeway interchange.

It seems to me that there might be some issues at busy intersections...

I just have trouble thinking that all of the software implementations are going to get it right.

[68fastback]

I think you're right.

Probably some situations will be dependent on the car-car and car-infrastructure communications architecture/protocols (already being tested at a municipal level in parts of Dearborn, not just test track, but that's just Ford's implementation -- cross-brand standards work is far from complete) ...even the but I'm sure there will be 'learning' situations -lol ...not unlike some of the aberations ARPANET traffic experienced in the early days ...also it's one thing to have an adverse failure syndrome flush all network data to one node (i.e. oblivion) ...a whole 'nother thing when it's people in cars being sucked into a similar failure syndrome :lol:

[twobjshelbys]

I think intersections in residential and off-main-road areas will get clogged up. What are you going to do, sit in the car until there isn't anyone there at 2AM? Your car will sit there with that little spinny thing on its screen waiting for divine intervention.

I think it will be like the Matrix and there will be scavenger vehicles like garbage trucks that come and clean up the trash at the backed up intersections every couple of hours. I'll start a company and get a patent on the self-driving garbage scows. I'll make mine look like Pacman

[Tommy Gun]

:StealingIdea:

[The Bone]

I have noticed intersections where there round d rounds and you turn right and go around and take the street you want. It gets traffic going in one direction. I have seen them everywhere. It's a pain in the ass for us drivers who drive cars. So no left turns just right turns.
Self driving cars will cause more problem than they salve. I could sure mess with one of those. LOL

[68fastback]

...and one has to wonder how long it will take to hack 'network' to force car routing so 'predators' can have a field-day ...like the wild west railroad robberies -lol ...I mean, you know it will happen. Heck, they've hacked TPMS on Jeeps and on Chevys they've figured out how to shut the engine off thru TPMS :rofl3:

There will be lots of learning-curve and that's not unusual for such never-before tech but it's different when people are directly involved ....I think there will remain a sizable segment of the population unwilling to toss their fate to 'trust me' programming. 80% of the Fortune 500 experiences at least one successful hack each year -- 80%!! ...and well over 95% have been successfully hacked with actual data loss so far ...including some of the best managed businesses on earth.

Anyone who thinks autonomous mobility systems will be exempt from illegal activity is simply naive. And I'm not just talking about look-at-me hackers ...you can bet Russia, Iran and NK will target it -- it's the ideal mass-hysteria cyber-war soft target. No system is immune because no system can be immune.

I'm not saying that should prevent the technology from proceeding, just that we need to err on the side of bullet-roof caution -- mobile autonomy needs to eventually be considered a National Infrastructure Asset (like military systems and, now, the power grid, I believe), an attack on which constitutes an act of war. That still won't stop nation-state bad actors nor some others, but will give pause to 'playing' with it much.

The more we divest our personal autonomy the greater the risk of being someone else's beatch.

[Alloy Dave]

All good questions, and I'm sure there are groups working on all these things.

One thing I think they will go to is a "default kill" position. In other words, if the "checksum" is not working right and some sort of error is detected, they would likely default to manual control only mode. This would be similar to a semi-truck where the air brakes fail...the system by default has the brakes in the "applied" condition, and it requires air pressure to allow the vehicle to move at all.

This way, if some error/hacker was detected, IMO the smartest thing would be to have all cars revert to manual mode only.

[68fastback]

Ford (and others) will be introducing driverless cars in 2020 with no driver controls ...or so they have said is the plan in press releases. Not sure I'm buying it will happen in 2020 but the stated direction is clearly no driver controls whatsoever, so robust failure syndromes will need to be carefully evaluated.

[Alloy Dave]

Ford (and others) will be introducing driverless cars in 2020 with no driver controls ...or so they have said is the plan in press releases. Not sure I'm buying it will happen in 2020 but the stated direction is clearly no driver controls whatsoever, so robust failure syndromes will need to be carefully evaluated.

Dan do you think those will only be usable in certain areas initially? Or do you think you can drive them wherever you want? I would think they'd pick one city somewhere as a test bed for a year or two before going all across the US.

[Joe G]

Dan do you think those will only be usable in certain areas initially? Or do you think you can drive them wherever you want? I would think they'd pick one city somewhere as a test bed for a year or two before going all across the US.

:goodpost:

I vote some medium sized town in the Midwest. Maybe Columbus, Indiana? :idea:

[68fastback]

Dan do you think those will only be usable in certain areas initially? Or do you think you can drive them wherever you want? I would think they'd pick one city somewhere as a test bed for a year or two before going all across the US.

I would think it would be a gradual roll-out ...possibly even not in the US for certain trials (Ford is working closely with the government of India but don't know details). The 2020 date was US but they didn't specify where or how limited -- probably very limited initially. Uber is already using the autonomous Fusion in limited numbers in a couple of US cities but those still have driver controls and I think they are still required to at least have someone sitting behind the wheel at this point.

---

What I find a little unsettling just looking at it by the numbers... The initial moon landing was driven by a robust commercial mainframe system with custom application software with well under 1M logic points and a virtually unlimited development and testing budget with some of the brightest minds on earth working on it -- it seemingly went well but didn't really at all ...it would have failed catastrophically if 'fixes' could not have been applied on the fly by astronauts in concert with 100s of bright people in Houston frantically working on them in real-time (everthing from runaway capsule roll that almost caused astronaut blackout to distance and fuel usage estimates that were flat wrong). Robust commercial operating systems are probably closer to current state-of-the-art logic programming and are more of the scale of 5-10M logic points in software (the hardware itself is fairly robust but there's also firmware -- software baked into hardware) and we all know how bullet-proof today's systems are :shades: ...they're excellent but far from perfect.

It's believed autonomous mobility/transportation systems will be on the order of 35-60M logic points (and I'd bet it will be effectively 2-3 times that at a network level before it approaches maturity). BART on the left-coast runs on friggin rails (literally) and has extensive system and fail-safe controls yet repeated crashes (even after numerous fixes) have required mandating an engineer be put back in the cab -doh! Even if we assume the autonomous mobility hardware will be flawless (tho it never will be) that is one heck of a lot of logic, and there are many more opportunities for failure than in a 'box' system or a rail system because of the unavoidably unanticipated dynamics of a mobile networked environment. I'm not saying it's not doable, just that there will be logic dynamics, the effects of which are simply unpredictable in the real world and therefore untestable until they occur the first time. The unanticipated consequences of which will be undeniably real ...however and whenever they materialize.

In the space program there could be no real default fallback (past a certain point, except for verbal mission control and not blacked-out and astronaut judgment when blacked-out). The good news is that in an autonomous transportation system there can be fallback because the default mode can be just to avoid-and-park (won't make passengers happy, but they likely are safe). The real buggaboos will be unpredictable complex failure syndromes like ARPANET's 'drain' failure (where an IMP fails by telling all other it's the fastest path to everywhere, so it dynamically sucks the network dry). In an autonomous mobility system it would be the equivalent of every autonomous node (car/truck, etc) in a region converging on the same major GPS/intersection. I'm not saying that failure syndrome is real or will EVER happen -- it's just a conceptual example of the unforseable syndromes that invariably will occur.

We have no idea what such failure syndromes will be though deep modelling will help discover many of them, but you can only model that which the logic can reveal and conceive, not that which an unpredicted failure syndrome can produce in a the real physical world or autonomous mobility.

And none of the above includes the perverse individuals using their skill and mobile 'data centers' -lol- to intentionally subvert the proper functioning of other vehicles on the road. You know they will be out there -- the analog of those who troll wi-fi hotspots wreaking havoc on the computers of others for the sheer joy it produces in their sick minds ...a whole different problem but with potentially very serious consequences.

[Alloy Dave]

:goodpost:

I vote some medium sized town in the Midwest. Maybe Columbus, Indiana? :idea:We're too redneck for that.

[Alloy Dave]

I would think it would be a gradual roll-out ...possibly even not in the US for certain trials (Ford is working closely with the government of India but don't know details). The 2020 date was US but they didn't specify where or how limited -- probably very limited initially. Uber is already using the autonomous Fusion in limited numbers in a couple of US cities but those still have driver controls and I think they are still required to at least have someone sitting behind the wheel at this point.

---

What I find a little unsettling just looking at it by the numbers... The initial moon landing was driven by a robust commercial mainframe system with custom application software with well under 1M logic points and a virtually unlimited development and testing budget with some of the brightest minds on earth working on it -- it seemingly went well but didn't really at all ...it would have failed catastrophically if 'fixes' could not have been applied on the fly by astronauts in concert with 100s of bright people in Houston frantically working on them in real-time (everthing from runaway capsule roll that almost caused astronaut blackout to distance and fuel usage estimates that were flat wrong). Robust commercial operating systems are probably closer to current state-of-the-art logic programming and are more of the scale of 5-10M logic points in software (the hardware itself is fairly robust but there's also firmware -- software baked into hardware) and we all know how bullet-proof today's systems are :shades: ...they're excellent but far from perfect.

It's believed autonomous mobility/transportation systems will be on the order of 35-60M logic points (and I'd bet it will be effectively 2-3 times that at a network level before it approaches maturity). BART on the left-coast runs on friggin rails (literally) and has extensive system and fail-safe controls yet repeated crashes (even after numerous fixes) have required mandating an engineer be put back in the cab -doh! Even if we assume the autonomous mobility hardware will be flawless (tho it never will be) that is one heck of a lot of logic, and there are many more opportunities for failure than in a 'box' system or a rail system because of the unavoidably unanticipated dynamics of a mobile networked environment. I'm not saying it's not doable, just that there will be logic dynamics, the effects of which are simply unpredictable in the real world and therefore untestable until they occur the first time. The unanticipated consequences of which will be undeniably real ...however and whenever they materialize.

In the space program there could be no real default fallback (past a certain point, except for verbal mission control and not blacked-out and astronaut judgment when blacked-out). The good news is that in an autonomous transportation system there can be fallback because the default mode can be just to avoid-and-park (won't make passengers happy, but they likely are safe). The real buggaboos will be unpredictable complex failure syndromes like ARPANET's 'drain' failure (where an IMP fails by telling all other it's the fastest path to everywhere, so it dynamically sucks the network dry). In an autonomous mobility system it would be the equivalent of every autonomous node (car/truck, etc) in a region converging on the same major GPS/intersection. I'm not saying that failure syndrome is real or will EVER happen -- it's just a conceptual example of the unforseable syndromes that invariably will occur.

We have no idea what such failure syndromes will be though deep modelling will help discover many of them, but you can only model that which the logic can reveal and conceive, not that which an unpredicted failure syndrome can produce in a the real physical world or autonomous mobility.

And none of the above includes the perverse individuals using their skill and mobile 'data centers' -lol- to intentionally subvert the proper functioning of other vehicles on the road. You know they will be out there -- the analog of those who troll wi-fi hotspots wreaking havoc on the computers of others for the sheer joy it produces in their sick minds ...a whole different problem but with potentially very serious consequences.
I hate to say it for privacy reasons, but given all the crap that goes on with the internet (scams etc.) we may someday have to each have a static IP and some agency would have to know each and every computer's exact location at any given time so all these cheats and terrorists can be found when they do stupid things.

[Joe G]

I hate to say it for privacy reasons, but given all the crap that goes on with the internet (scams etc.) we may someday have to each have a static IP and some agency would have to know each and every computer's exact location at any given time so all these cheats and terrorists can be found when they do stupid things.

We give up our privacy way too easily the way it is, so I firmly vote NOPE!



http://www.kg4cyx.net/wp-content/uploads/2014/06/NopeMeter.png?x69270

[68fastback]

I hate to say it for privacy reasons, but given all the crap that goes on with the internet (scams etc.) we may someday have to each have a static IP and some agency would have to know each and every computer's exact location at any given time so all these cheats and terrorists can be found when they do stupid things.

That's easily thwarted with free software that can make it appear you're anywhere you want to be at any address you want to be. State actors use several levels such mis-direction today -- widely know to be available on the dark web. And there's so much more I won't post here.

There is literally no scheme that can't be broken, hacked or otherwise infiltrated -- none ...there never can be one.

Hackers can't decrypt robust large-key encryption -- it even takes governments w/supercomputers months or years (thank Bill Clinton for declassifying DES and giving it away -- ISIS routinely communicates with more security than most Americans), but hackers will use other means to lure the unwary to do it to themselves ...any system is vulnerable.

For example, even the most bullet-proof, isolated, encrypted government systems (with tightly-controlled connectivity) have been hacked via alternative means. Like fully packaged 'new' thumb drives dropped in a government sub-contractor facility parking lot -- that's how China ultimately broke in and made-off with hundreds of thousands of files of detailed military secrets including full engineering details of some of our latest and secret goodies. It's infuriating but very real.

Any smart garden-variety hacker can do similar things once they find a weakness (of any kind: system, process or human-nature). And hackers live to be the first to find and exploit those weaknesses for peer adulation and underground 'fame.'

The simple truth is that no system is un-hackable as long as humans created, manage or otherwise use it.

An autonomous mobility network would have to be high on the list of individual or nation-state hackers for havoc-inducing cyber mischief ...sad but true.

It's really not a question of if it will happen, only how, when, by whom and specifically how it's exploited.

[Alloy Dave]

I'm joining ISIS. :look:

[68fastback]

:haha:

Indiana Sons In Suds?

[Alloy Dave]

I don't know, I just want to have fun. :look: