Autopilot vs Autonomous Driving - Discuss

[OutlawX3M]

So this weekend I read through an article on NYTimes regarding a fatal crash of a Tesla in Ohio.

Tesla Faults Brakes, Not Autopilot

So Tesla defines autopilot as only one system in autonomous driving that centers around steering which is what good marketing is all about -- define a space in the market and claim you're a pioneer/leader in it. I'm just having a hard time with the way Tesla is spinning this.

It's not autpilot's fault, it's the brakes. Ok, so your steering technology wasn't the fault but the braking IS. How's that a good spin that your autonomous system? Don't you need both systems working in conjunction to have true autonomous driving that at least parallels a human driving a vehicle?

By the way, here's a good primer on Google's approach vs Tesla's approach to vehicle locating for autonomous driving.

I don't believe autonomous driving is going to replae human drivers within the next 10 years. In fact, I think it will be 20 years + before there's more miles driven autonomously than by humans in passenger vehicles. Taxis and trucking maybe....we'll see.

[atxgt3]

We're not even advanced enough to make bug free infotainment systems in our cars, there's no way fully autonomous cars are anywhere close to being ready.

I mean seriously, look at all the software you use on a daily basis and pay attention to how buggy it is. A fully autonomous car is way more complicated than most of that, and at least an order of magnitude harder than what Tesla is doing right now.

[OutlawX3M]

Originally Posted by atxgt3

I mean seriously, look at all the software you use on a daily basis and pay attention to how buggy it is. A fully autonomous car is way more complicated than most of that, and at least an order of magnitude harder than what Tesla is doing right now.

I work in the software business so I agree 1000%. Anything that's built by humans will never be 100% perfect - hence the saying, "we're only human".

The integration required for autonomous vehicles is staggering IMO. Even with advances in circuitry, robotics, sensors, et al, the computational power required to manage it all is just beyond the "by 2020...." predictions I read about.

This is why the first wave of autonomous will be in slices and it's already here my opinion. We have lane departure, collision warning, emergency braking, dsc/traction control, etc. The geeks look at this technology as a cool puzzle to piece together with some wicked coding. Great academic exercise but is it a reality?

Like you said, all you need is one bug and the consequences are dire. Tesla is trying to avoid this discussion at all costs.

[Viffermike]

Cut-and-paste from a post of mine from another related thread:

From Efthreeoh : "The term "auto pilot" comes from the aviation industry. Auto pilot works for commercial aircraft because commercial air transportation happens at very high altitudes with a low-dynamic traffic situation - i.e. little air traffic, 5-mile separation, and altitude separation. And the main point... air traffic is managed by humans and automation systems, so placement of each aircraft is monitored and controlled by a third (independent) party.

In the case of the Tesla incident, neither vehicle's placement was monitored nor controlled by a third, independent party. This is why autonomous automobile operation will never work without a vehicle traffic management system. I doubt such a system could be economically devised for ground automobile vehicular traffic. Even if it were able to happen then there is the issue of privacy rights..."

Now me:
And that's just one point of several against autonomous tech in U.S. private vehicles being a widespread reality within the next 20 years, much less five. Also consider:

1). The military still pilot drones; virtually none of them are fully automatic. Why? Because of governmental culpability for collateral damage if something goes wrong.
2). The aforementioned autopilot function in commercial aircraft -- already the most complex and systems-redundant vehicle on the planet short of a spacecraft. Yet pilots are still required to take off and land an airplane. Why? The same reasons as 1)., but the culpability falls on airlines and manufacturers instead of governments in most cases.
3). Insurance. Autonomous vehicles would shift the burden of insurance culpability -- for both property damage and bodily harm -- from operators to manufacturers. Tesla's Fatality #1, and the reaction to it, is direct evidence of this effect; most are blaming Tesla, not the operator, for the death. This is a paradigm shift in the industry that will take years upon years of political and institutional wrangling to achieve -- and for this reason, part of the purchase price of an autonomous vehicle could very well be an insurance 'premium' that will delay the technology being affordable for the masses. One thing can mitigate this: the government becoming involved, either by an aforementioned autonomous control grid or by subsidizing the insurance requirements -- and we all know how well the latter scenario flies in our country.

There are other reasons, but the above is enough. Fully self-driving cars (i.e. "autonomous") will not happen in the U.S. for a long, long time.

[OutlawX3M]

Don't overlook that in my post, "autopilot" is the term Tesla uses for their steering system and not referring to "autopilot" as you do (in aircraft).

You do make a great point that for autonomous vehicles to become a reality, there is a need for ground "control" systems. There's many projects (if you will) that are working towards that reality such as AIM: Autonomous Intersection Management.

Audi has their traffic light assistance technology too.

As for "big brother", well, we "sign away" or "suspend" our rights every day in the name of safety (think TSA) or convenience (think Google / Apple) or even incentivized pitches (Progressive's Snapshot).

But I have to remember that driving isn't a right protected by an Amendment, it's a privilege. I hate that fact but it's the reality in the USA.

[Dog Face Pony Soldier]

You'd think we'd have autonomous railways before roadways.

[Efthreeoh]

Originally Posted by Outlaw M3;20345041[B

]Don't overlook that in my post, "autopilot" is the term Tesla uses for their steering system and not referring to "autopilot" as you do (in aircraft)[/B].

You do make a great point that for autonomous vehicles to become a reality, there is a need for ground "control" systems. There's many projects (if you will) that are working towards that reality such as AIM: Autonomous Intersection Management.

Audi has their traffic light assistance technology too.

As for "big brother", well, we "sign away" or "suspend" our rights every day in the name of safety (think TSA) or convenience (think Google / Apple) or even incentivized pitches (Progressive's Snapshot).

But I have to remember that driving isn't a right protected by an Amendment, it's a privilege. I hate that fact but it's the reality in the USA.

But that's the point, the term "Autopilot" is synonymous with the operation of aircraft where the general (mis)understanding of the function of it means the vehicle operator allows the vehicle to take over control of the operation of the vehicle. With aircraft, for safety, two pilots are required to be in the cockpit. The Tesla incident obviously shows that the public (i.e. the Tesla pilot) can misconstrue what the capabilities are of the autopilot feature and lead to a catastrophic failure such as running under a tractor trailer because the system recognized it as a billboard.

Regarding big brother, I think the 4th Amendment can be interpreted as allowing free unencumbered travel without the Government knowing where you are going. Regarding air travel, the Government knows a person is travelling from one airport to another, but not their actual destination. With auto travel, for the autonomous operation of cars and the traffic control system required to allow autonomous operation (anyone see an oxymoron here?) the Government will have to know your final destination. Yes, you do "sign away" privacy to telecom companies, but they are not the Government and are not compelled to provide such information to the Feds.

[Efthreeoh]

Originally Posted by glennQNYC

You'd think we'd have autonomous railways before roadways.

When a train hits a car, it is generally the car driver's fault... The autonomous train (position controlled by the railway) crosses over the road system operated by uncontrolled individual drivers. Boom.

[OutlawX3M]

Originally Posted by Efthreeoh

But that's the point, the term "Autopilot" is synonymous with the operation of aircraft where the general (mis)understanding of the function of it means the vehicle operator allows the vehicle to take over control of the operation of the vehicle. With aircraft, for safety, two pilots are required to be in the cockpit. The Tesla incident obviously shows that the public (i.e. the Tesla pilot) can misconstrue what the capabilities are of the autopilot feature and lead to a catastrophic failure such as running under a tractor trailer because the system recognized it as a billboard.

Regarding big brother, I think the 4th Amendment can be interpreted as allowing free unencumbered travel without the Government knowing where you are going. Regarding air travel, the Government knows a person is travelling from one airport to another, but not their actual destination. With auto travel, for the autonomous operation of cars and the traffic control system required to allow autonomous operation (anyone see an oxymoron here?) the Government will have to know your final destination. Yes, you do "sign away" privacy to telecom companies, but they are not the Government and are not compelled to provide such information to the Feds.

I believe we're on the same wavelength -- autonomous cars/trucks are not as close as we're led to believe. I see it as a multi-step process that will evolve over decades because there will be setbacks such as the Ohio crash. I see the progression of this techno-wizardry as follows:

Version 1.0 - semi-autonomous
This is where Tesla and a few other companies are today. Vehicles that can for a period of time operate without human control (i.e., cruising on a freeway or in stop-n-go traffic) but not 100% of the time with the same level of safety as a human driver operating the vehicle. There's no traffic control assistance that are autonomous-aware just vehicle-based systems to mimic human control (which are limited) and human intervention is the only fall back.

Version 2.0 - semi-autonomous plus
Commuter traffic (i.e., taxis, buses) as well as commercial traffic (trucking) where the travel is within well defined routes with high quality maps & sensor systems plus traffic controls that are autonomous-aware assisting the vehicles.

For example, Uber partners with NYC to install traffic controls to assist with autonomous driving to propel the adoption of autonomous taxis. This investment can be a well planned roll-out that begins with basic routes (Penn Station to X and back) where Uber's fleet is a mixture of autonomous and human-driven vehicles.

Version 3.0 - autonomous
Nirvana for those championing autonomous vehicles. Highly developed in-vehicle systems working with "smart" traffic controls that have human oversight. That last part is the key and I'm thinking like you, the Air Traffic Control system we have in place where humans are monitor air traffic and have influence within the overall system.

One thing is certain, this will be an interesting space to watch and possibly invest in.....

[OutlawX3M]

Originally Posted by glennQNYC

You'd think we'd have autonomous railways before roadways.

Yeah, well there's one word that's keeping that from happening -- unions.

There's also a few technical issues like driving a train is much different than driving a car -- think about how much distance you need to stop a train vs a car and there's the problem. You'd have to monitor the entire railway line which becomes very cost prohibitive for long distance rail travel.

But even with that, there's many autonomous light-rail systems (AirTrain at JFK Airport is one of them) so in specific cases, this works but I don't think it will for Amtrak.

[Bbb34]

Originally Posted by Outlaw M3

I work in the software business so I agree 1000%. ..... Great academic exercise but is it a reality?

Like you said, all you need is one bug and the consequences are dire. Tesla is trying to avoid this discussion at all costs.

Completely bug free software hasn't been achieved yet, no matter what industry. And like you said, all you need is one bug and price to pay is high. And that's considering conditions are perfect and environment is "somewhat controlled". Assumptions are made that all other participants in traffic are also running completely bug free software and are somehow ensuring hardware redundancy, there are no weather related challenges, pedestrians, bikers (perhaps other drivers that are not yet forced by law to give up driving) misbehaving... etc... etc.

I think fully autonomous is still much further away than everybody would like us to believe these days.

[Red Bread]

I'll take a Commodore 64 driving a car over the people we give licenses to.

[Viffermike]

Originally Posted by Red Bread

I'll take a Commodore 64 driving a car over the people we give licenses to.

And that's yet another reason I don't see autonomous cars happening for at least another generation.

One of the big social benefits of autonomous travel theoretically is that it will enable those who would normally not be able to operate a motor vehicle -- the elderly, handicapped, etc. -- to access to a motor vehicle and become more mobile. What goes hand-in-hand with that is who may be able to use them as well who have no business in one: children, the blind, chronic DWI/DUI offenders, etc.

In my mind, anyone who can't get a license now -- and I'm of the opinion that driver's licenses are far too easy to get in the U.S. -- shouldn't be able to operate an autonomous car because they have no ability to operate one in manual mode. But clearly, autonomous vehicles are partially aimed at enabling those who can't use a vehicle manually.

This conundrum bolsters my fundamental argument regarding culpability, insurance, etc. Who would be responsible for the implications? The current system can't mitigate such a question -- therefore, the system must be adjusted. This will take many, many years and involve many of the most powerful lobbys in Washington: insurance, car manufacturers, the AARP, and so on.

Yeah. It ain't gonna happen.

[Red Bread]

Exactly why it will happen. People suck at driving. Or, more accurately, smartphone and French fry wielding Americans suck at driving. We need to take people out of driving.

Give me access to racetracks and three pedals to use. I'll take a Google car there.

[Bbb34]

Originally Posted by Red Bread

We need to take people out of driving.

Or out of software design/implementation/testing, hardware design... etc . Those same humans, with all their faults, are still the ones implementing these systems. Not to mention that some executive or consulting firm might rush things for profit, been known to happen sometimes before products are fully ready.

My point is, yes, some people aren't good drivers, but it's incredibly complex to fully implement completely autonomous solution for all conditions out there. As has been mentioned several time before, insurance companies, car manufactures and law makers will also take forever to sort things out. In scenario when you cannot blame the driver anymore, who exactly will take the responsibility for things potentially going wrong. What about security of these software systems? How do you legislate that, or those who still may want to drive their own cars, what about bicycles, motorcycles (all 3 wheel ones in the future so they can be self driven) ?

[PopsnBurbles]

Are they seriously playing with semantics here calling their automated steering system "autopilot" to simply escape the blame of their failed full autopilot system? I suppose braking isn't a part of Tesla's autopilot but the general public's sense of autopilot obviously has braking in their minds.

Shit, if a tiny company like Koenigsegg (spelling?) can admit to their failed front ABS and owned it up 100%, I'm sure Tesla sure can without doing crap like this 'technically, our autopilot didn't fail.'

[Viffermike]

Originally Posted by PopsnBurbles

Are they seriously playing with semantics here calling their automated steering system "autopilot" to simply escape the blame of their failed full autopilot system? I suppose braking isn't a part of Tesla's autopilot but the general public's sense obviously has braking in their minds.

Semantics often mean the difference between winning and losing a class-action lawsuit ... so, yes. Tesla already has a history of bending terminology to suit its needs, both legally and in its marketing.

But in the larger realm: Semantics is one of the most important aspects of effective communication, period. It's an art that far too few people appreciate, much less practice well. The hard part of the skill is contextualization. It changes over time -- sometimes subtly, sometimes suddenly -- and much of it depends on intended audience vs. actual audience.

Three great examples:
1). the Bible
2). the Quran
3). the U.S. Constitution
All are absolute semantic rat's nests. It's easy to understand why.

[OutlawX3M]

Originally Posted by PopsnBurbles

Are they seriously playing with semantics here calling their automated steering system "autopilot" to simply escape the blame of their failed full autopilot system? I suppose braking isn't a part of Tesla's autopilot but the general public's sense of autopilot obviously has braking in their minds.

Shit, if a tiny company like Koenigsegg (spelling?) can admit to their failed front ABS and owned it up 100%, I'm sure Tesla sure can without doing crap like this 'technically, our autopilot didn't fail.'

That's my point exactly! To Tesla, "autopilot" is one sub-system in their autonomous driving design so technically it didn't fail (if you follow their logic) but in reality their autonomous driving system as a whole DID fail.

“There are three types of lies -- lies, damn lies, and statistics.”
― Benjamin Disraeli

[Viffermike]

Originally Posted by PopsnBurbles

If a tiny company like Koenigsegg (spelling?) can admit to their failed front ABS and owned it up 100%, I'm sure Tesla sure can without doing crap like this 'technically, our autopilot didn't fail.'

Wrong. Koenigsegg isn't beholden to shareholders or massive future or foreign investment, and its output is basically made to order for an exceedingly elite clientele that the company has personal relationships with the majority of -- and, natch, this clientele expects the car to push operational envelopes. In a nutshell, its semantics in presenting any Koenigsegg car to a customer is minimized by an honesty of intent.

Tesla doesn't work that way. At all -- and frankly, it shouldn't because its cars are mass produced.

Bad comparison.

[OutlawX3M]

Some very good intelligent thoughts here folks. I'm liking this.

One of my main concerns around Tesla is that they aren't focused enough to succeed (opinion). They build a great EV vehicle (3 now) but they're going to struggle with Model 3 production!

They have GigaFactory to produce batteries.
They're buying SolarCity too.

Then there's Space-X and HyperLoop....

[Efthreeoh]

Originally Posted by Outlaw M3

I believe we're on the same wavelength -- autonomous cars/trucks are not as close as we're led to believe. I see it as a multi-step process that will evolve over decades because there will be setbacks such as the Ohio crash. I see the progression of this techno-wizardry as follows:

Version 1.0 - semi-autonomous
This is where Tesla and a few other companies are today. Vehicles that can for a period of time operate without human control (i.e., cruising on a freeway or in stop-n-go traffic) but not 100% of the time with the same level of safety as a human driver operating the vehicle. There's no traffic control assistance that are autonomous-aware just vehicle-based systems to mimic human control (which are limited) and human intervention is the only fall back.

Version 2.0 - semi-autonomous plus
Commuter traffic (i.e., taxis, buses) as well as commercial traffic (trucking) where the travel is within well defined routes with high quality maps & sensor systems plus traffic controls that are autonomous-aware assisting the vehicles.

For example, Uber partners with NYC to install traffic controls to assist with autonomous driving to propel the adoption of autonomous taxis. This investment can be a well planned roll-out that begins with basic routes (Penn Station to X and back) where Uber's fleet is a mixture of autonomous and human-driven vehicles.

Version 3.0 - autonomous
Nirvana for those championing autonomous vehicles. Highly developed in-vehicle systems working with "smart" traffic controls that have human oversight. That last part is the key and I'm thinking like you, the Air Traffic Control system we have in place where humans are monitor air traffic and have influence within the overall system.

One thing is certain, this will be an interesting space to watch and possibly invest in.....

Yeah, I agree, we are on the same page.

[Efthreeoh]

Originally Posted by Red Bread

Exactly why it will happen. People suck at driving. Or, more accurately, smartphone and French fry wielding Americans suck at driving. We need to take people out of driving.

Give me access to racetracks and three pedals to use. I'll take a Google car there.

I disagree. I think people are actually very good at driving. There is no system yet devised, even in the next 25 years, that can control the amount of concurrent vehicle miles driven currently executed by humans right now today, August 2, 2016.

The reason Google is involved in autonomous control of vehicles is because they realize the smart phone is fucking up the time/space continuum of safe vehicular travel, and at some point a lawyer is going to figure out how to make them culpable.