US plans $4bn for self-driving rules

[mikekle]

There are other problems that will need to be overcome aside from vehicle ownership issues. The current liability law will be one, and the other are software performance validation and verification issues, a particularly difficult example being “state explosion.” I imagine that just exploring the state space itself in a test environment will be challenging, and a contributor of end costs due to certifications needed.

The last story I watched about this, the big 3 automakers all have first gen self driving cars coming out in 2018, I believe Mercedes has one sooner, but not sure about that, they are all saying they will take responsibility for any accidents caused by this technology, that tells me they must have this fairly close to being perfected.

I would imagine the trucking industry is paying close attention or involved in this, they could save billions if they did not have to employee human drivers.

I dont think a ‘ride on demand’ structure would work too well, not in the US anyway, people here like owning their vehicle, making it their own, plus, most of the time, my car is filled with junk for work, just getting a ride to work and back would not work well for me.

Whatever they come up, it will eventually have to be mandatory for all cars, it would not work to have just some cars self driving and others being manually driven, probably be this way in the beginning, while people are getting acclimated to the idea, but for a computer program to work flawlessly, have to take out the unpredictable variables.

 

[vz71]

Humans are not the only unpredictable variables.

I highly doubt people will respond well to outlawing driving their own vehicle.
But assuming they manage to, how do we intend on dealing with pedestrian traffic, animals, and heavy weather?

 

[mikekle]

Humans are not the only unpredictable variables.

I highly doubt people will respond well to outlawing driving their own vehicle.
But assuming they manage to, how do we intend on dealing with pedestrian traffic, animals, and heavy weather?

From the stories Ive seen and read about it, I do remember this question being asked in the interview, the only hurdle automakers are still having is, other cars driving too close, computer sensors can pick up pedestrians, animals, etc coming onto the road and respond accordingly, quicker than a human.

I do agree about outlawing driving though, that will be interesting to see how they handle that, maybe like other issues, they will claim they know whats best for us and doing what they can to keep everyone safe? (sort of like they did with the healthcare thing), plus, once people find out collision insurance and maybe even liability coverage will be gone, might be more willing to get onboard with it, or maybe eventually, insurance rates, if you choose to drive manually will skyrocket, forcing most people to accept it?

This was a 60 minutes piece a couple weeks ago, they had GM, Mercedes, and Honda on the show, it was very informative, not sure if its online anywhere though, I was very surprised at the level they are at right now, Before watching, I assumed self driving was at least 10-15 yrs out, its really only about 3-5 yrs away.

 

[Exiled_Child]

These contraptions are far more workable on interstate highways than they are within cities. I’ve read that one of their weak points is sensing motion at intersections. It seems that when they come to a four way intersection, being law-abiding machines, they come to a complete and full stop, whereas most human drivers will slow way way down, but will still sort of roll through when it’s their turn. The motion sensors can’t handle this. Bicyclists, too, confuse them, because many cyclists will stop at the stop sign or light, but they’ll still be rocking backwards and forwards, and moving their feet around on the pedals as they balance themselves waiting to go. This apparently just fries the robot’s brain, and it ends up frozen as traffic just flows around it.

 

[mikekle]

These contraptions are far more workable on interstate highways than they are within cities. I’ve read that one of their weak points is sensing motion at intersections. It seems that when they come to a four way intersection, being law-abiding machines, they come to a complete and full stop, whereas most human drivers will slow way way down, but will still sort of roll through when it’s their turn. The motion sensors can’t handle this. Bicyclists, too, confuse them, because many cyclists will stop at the stop sign or light, but they’ll still be rocking backwards and forwards, and moving their feet around on the pedals as they balance themselves waiting to go. This apparently just fries the robot’s brain, and it ends up frozen as traffic just flows around it.

This is only the beginning, small things are going to pop up and cause problems, this is to be expected, but regarding full stops, traffic lights, etc. the impression I got, was, with fully self driving roads, we would not really need traffic control devices, or stop signs, you could have 1000s of cars coming and going in all different directions, if they were all computer controlled, there would be no wrecks or close calls, as the computer guidance could keep everything flowing smoothly.

The only reason we have traffic lights and stop signs now is, humans make mistakes sometimes, so its easier to have one lane of traffic going one way at a time, and the other lane completely stopped, so no one makes a mistake or doesnt see another car coming, etc…with computers fully controlling everything, they will not make those simple mistakes.

It will really be more about acclimating people to the new ways imo, for the technology, doesnt sound like its that difficult, I mean, we trust computers to do a lot of things today and we just expect them to work right everytime, but it took time to reach this level of trust to.

 

[sps49]

The last story I watched about this, the big 3 automakers all have first gen self driving cars coming out in 2018, I believe Mercedes has one sooner, but not sure about that, they are all saying they will take responsibility for any accidents caused by this technology, that tells me they must have this fairly close to being perfected.

I would imagine the trucking industry is paying close attention or involved in this, they could save billions if they did not have to employee human drivers.

I dont think a ‘ride on demand’ structure would work too well, not in the US anyway, people here like owning their vehicle, making it their own, plus, most of the time, my car is filled with junk for work, just getting a ride to work and back would not work well for me.

Whatever they come up, it will eventually have to be mandatory for all cars, it would not work to have just some cars self driving and others being manually driven, probably be this way in the beginning, while people are getting acclimated to the idea, but for a computer program to work flawlessly, have to take out the unpredictable variables.

Anyone may say that they are going to produce a self driving car in 2016, but we do not yet know what that looks like; the federal government and states are struggling to define what it is as we speak.

They will have to do more than remove the “unpredictable variables,” they will have to demonstrate that autonomous vehicles can operate in similar environments that piloted vehicles operate in today. Constraining reality to fit the capabilities of the proposed autonomous performance envelope is not a solution.

 

[mikekle]

Anyone may say that they are going to produce a self driving car in 2016, but we do not yet know what that looks like; the federal government and states are struggling to define what it is as we speak.

They will have to do more than remove the “unpredictable variables,” they will have to demonstrate that autonomous vehicles can operate in similar environments that piloted vehicles operate in today. Constraining reality to fit the capabilities of the proposed autonomous performance envelope is not a solution.

Right, this is what they have been doing in a few states the past couple years, I know CA is one of the states. I think the only brand offering this in 2016 is Mercedes, and its not fully self driving, only for expressways, Its 2018 when the other automakers will all have models out.

I have read Google and some other computer companies are heavily involved in this, there was even some mention of Google starting to make cars.

 

[sps49]

Right, this is what they have been doing in a few states the past couple years, I know CA is one of the states. I think the only brand offering this in 2016 is Mercedes, and its not fully self driving, only for expressways, Its 2018 when the other automakers will all have models out.

I have read Google and some other computer companies are heavily involved in this, there was even some mention of Google starting to make cars.

Thank you for the correction, I meant 2018.

Autonomous vehicle research has been ongoing for at least 60 years. While there have been significant advances, significant hurdles remain that must be overcome to achieve this goal, and it is not yet clear that a solution exists in the form that the problem has been defined.

 

[Patavium]

There are other problems that will need to be overcome aside from vehicle ownership issues. The current liability law will be one, and the other are software performance validation and verification issues, a particularly difficult example being “state explosion.” I imagine that just exploring the state space itself in a test environment will be challenging, and a contributor of end costs due to certifications needed.

There are certainly many challenges. Insurance companies will need to change their structure to deal with CAV, and this is changing slowly through the use of user-based insurance. There are also cyber security challenges. It is not enough to develop and deploy the technology, it also must be resilient against security threats.

Also, the US is not the only country with the capabilities of making this happen. Singapore has partnered with MIT and they have been working at deploying the vehicles for quite a while now. lta.gov.sg/ltaacademy/doc/J14Nov_p05Tan_AVnextStepSingapore.pdf

There was also deployment here: phys.org/news/2014-10-driverless-vehicles-singapore.html, where the vehicles interacted in a dense pedestrian area, although as small buggys.

Another interesting aspect is coordination between AVs, and AVs and Infrastructure.

 

[Exiled_Child]

There are certainly many challenges. Insurance companies will need to change their structure to deal with CAV, and this is changing slowly through the use of user-based insurance. There are also cyber security challenges. It is not enough to develop and deploy the technology, it also must be resilient against security threats.

Also, the US is not the only country with the capabilities of making this happen. Singapore has partnered with MIT and they have been working at deploying the vehicles for quite a while now. lta.gov.sg/ltaacademy/doc/J14Nov_p05Tan_AVnextStepSingapore.pdf

There was also deployment here: phys.org/news/2014-10-driverless-vehicles-singapore.html, where the vehicles interacted in a dense pedestrian area, although as small buggys.

Another interesting aspect is coordination between AVs, and AVs and Infrastructure.

You’re the only person I’ve ever talked to who’s directly involved with this. Do you happen to know, is the idea to make these compatible with human drivers just in the short-term, until all vehicles are self-driving? Because it seems obvious that this won’t be workable unless 100% of vehicles on the road are self-driving. Also, what are your thoughts on city buses? I’m curious, because I can’t see how they would be able to operate without human drivers there to deal with the thousand and one weird quirks that always seem to be popping up on public transit. I think even automated subway trains always have drivers sitting in the front; I know AMTRAK always does.

 

[Sarcelle]

You’re the only person I’ve ever talked to who’s directly involved with this. Do you happen to know, is the idea to make these compatible with human drivers just in the short-term, until all vehicles are self-driving? Because it seems obvious that this won’t be workable unless 100% of vehicles on the road are self-driving. Also, what are your thoughts on city buses? I’m curious, because I can’t see how they would be able to operate without human drivers there to deal with the thousand and one weird quirks that always seem to be popping up on public transit. I think even automated subway trains always have drivers sitting in the front; I know AMTRAK always does.

Talking about public transit, the light rail train in Honolulu currently being built, will be a first for the US. No human operators at all, just fully automated trains.

 

[sps49]

There are certainly many challenges. Insurance companies will need to change their structure to deal with CAV, and this is changing slowly through the use of user-based insurance. There are also cyber security challenges. It is not enough to develop and deploy the technology, it also must be resilient against security threats.

Also, the US is not the only country with the capabilities of making this happen. Singapore has partnered with MIT and they have been working at deploying the vehicles for quite a while now. lta.gov.sg/ltaacademy/doc/J14Nov_p05Tan_AVnextStepSingapore.pdf

Singapore is not exactly “making this happen.” The SAVI CARTS committee is not unique in its goal for automation. I am always suspicious when someone uses the word “holistically” when describing technology. It usually an indicator that they don’t have a good grasp on the concepts. Usually, requirements and specifications are developed, and a system engineered to be in compliance with those reqs and specs.

There was also deployment here: phys.org/news/2014-10-driverless-vehicles-singapore.html, where the vehicles interacted in a dense pedestrian area, although as small buggys.

One wonders how much of this is Singapore, and how much is MIT. I don’t doubt that good work is being performed, but the hype breeds suspicion. Emilio Frazzoli may “re-imagine,” but reality always wins. I was part of a globally important first to market high risk aerospace venture in the late 1990’s and the hype was so thick you could not see what was happening, even at the board of directors level. Always better to keep it real.

Another interesting aspect is coordination between AVs, and AVs and Infrastructure.

Now we are getting somewhere. This is the most likely incarnation of a widely deployed first gen AV system, one with a distributed and or centralized control. Not my personal preference, but I’m not the one putting up any risk either.

 

[Exiled_Child]

Talking about public transit, the light rail train in Honolulu currently being built, will be a first for the US. No human operators at all, just fully automated trains.

Don’t you think that they’ll at least have fare collectors or transit cops or someone to just deal with emergencies?

My guess is that there will always be some kind of human worker around, just in case they’re needed.

 

[Sarcelle]

Don’t you think that they’ll at least have fare collectors or transit cops or someone to just deal with emergencies?

My guess is that there will always be some kind of human worker around, just in case they’re needed.

Probably.

What I meant were operators that actually drove the train.

 

[Patavium]

Singapore is not exactly “making this happen.” The SAVI CARTS committee is not unique in its goal for automation. I am always suspicious when someone uses the word “holistically” when describing technology. It usually an indicator that they don’t have a good grasp on the concepts. Usually, requirements and specifications are developed, and a system engineered to be in compliance with those reqs and specs.

It was written for the general public, so not surprisingly it was adopted for such. In terms of actual research, there are no standards, and thus requirements and specifications vary widely. It is not simply the hardware so to speak (even hardware varies widely across the teams), but also the Artificial Intelligence approaches are very different across the teams in the CAV research area.

One wonders how much of this is Singapore, and how much is MIT. I don’t doubt that good work is being performed, but the hype breeds suspicion. Emilio Frazzoli may “re-imagine,” but reality always wins. I was part of a globally important first to market high risk aerospace venture in the late 1990’s and the hype was so thick you could not see what was happening, even at the board of directors level. Always better to keep it real.

What is your perception of this “reality” for you? Prof. Frazzoli of MIT is one of various pioneers in this area, and his AV does work, and it has seen practical experience in dense pedestrian areas, and also as mobility on demand service. CAV will happen. Singapore is leading the effort for deployment. There are test sites that are running and will run more CAV in the city. We are not talking about SciFi. We are talking about now.

In the U.S., the Federal Government (through US DOT) has just started a Smart City program, which will select various US Cities and bring CAV along with other technologies to transform the landscape of transport options available to the residents, and also freight industry. This is NOT wasted resources, but progress to make this concept a reality now in the U.S. There are similar efforts by other branches of the Federal Government such as DOE.I see this as part of the U.S. response of bringing these technologies to the civilians, and not only military, and also stay afloat given initiatives elsewhere such as Singapore.

Now we are getting somewhere. This is the most likely incarnation of a widely deployed first gen AV system, one with a distributed and or centralized control. Not my personal preference, but I’m not the one putting up any risk either.

 

[Patavium]

You’re the only person I’ve ever talked to who’s directly involved with this. Do you happen to know, is the idea to make these compatible with human drivers just in the short-term, until all vehicles are self-driving? Because it seems obvious that this won’t be workable unless 100% of vehicles on the road are self-driving. Also, what are your thoughts on city buses? I’m curious, because I can’t see how they would be able to operate without human drivers there to deal with the thousand and one weird quirks that always seem to be popping up on public transit. I think even automated subway trains always have drivers sitting in the front; I know AMTRAK always does.

Yes, it will be compatible with human drivers. CAVs will also be used for Freight. The technology does NOT requires 100% of vehicles on the road to be self-driving.

We are looking more at Personalized Transit services, such as CAV sharing vehicles. Not sure, if other are looking at CAV City Buses. Not sure what weird quirks pop up on public transit?

Automated trains has been around for while. Several countries have deployed such. Public Transit in the US, unfortunately, is quite outdated and not up really the state of the practice. You just need to look to Europe and Asia as the leaders, and even some Latin American Countries have some better examples than the US.

 

[edwest2]

Cost, liability and software failure/malfunction. The current auto industry has recall after recall for parts not working correctly. This is not impressive at all, and the cost for such a car or a retrofit? Not fully thought out and not fully tested, plus if a part fails, then what?

I’ll be impressed after the first year and 100% no failures in any part of the car.

Ed

 

[sps49]

It was written for the general public, so not surprisingly it was adopted for such. In terms of actual research, there are no standards, and thus requirements and specifications vary widely. It is not simply the hardware so to speak (even hardware varies widely across the teams), but also the Artificial Intelligence approaches are very different across the teams in the CAV research area.

Even software requires a specification and requirement. If the problem is not defined, then the costs cannot be controlled. At this point I don’t even see a FAA CAST analog in place.

What is your perception of this “reality” for you? Prof. Frazzoli of MIT is one of various pioneers in this area, and his AV does work, and it has seen practical experience in dense pedestrian areas, and also as mobility on demand service. CAV will happen. Singapore is leading the effort for deployment. There are test sites that are running and will run more CAV in the city. We are not talking about SciFi. We are talking about now.

My perception informs me that it is not ready for deployment, for reasons already articulated, as well as others.

In the U.S., the Federal Government (through US DOT) has just started a Smart City program, which will select various US Cities and bring CAV along with other technologies to transform the landscape of transport options available to the residents, and also freight industry. This is NOT wasted resources, but progress to make this concept a reality now in the U.S. There are similar efforts by other branches of the Federal Government such as DOE.I see this as part of the U.S. response of bringing these technologies to the civilians, and not only military, and also stay afloat given initiatives elsewhere such as Singapore.

The judgement on whether the resources are wasted or not will come when performance verification occurs. As far as “bringing the technology to civilians”, it’s not as easy as that might sound. The Galvin Report authors had the notion that the DoE was some kind of national technology treasure chest, ready to be bestowed upon private industry. It did not work out that way. I plan to observe events as they occur instead of writing history proactively.

 

[vz71]

Cost, liability and software failure/malfunction. The current auto industry has recall after recall for parts not working correctly. This is not impressive at all, and the cost for such a car or a retrofit? Not fully thought out and not fully tested, plus if a part fails, then what?

I’ll be impressed after the first year and 100% no failures in any part of the car.

I am sure lawyers are licking their chops just thinking about it.

Ed

Indeed.
I just drove past a red light camera that was snapping pictures of every car going through.
And this is tech that has been tested for years and supposedly is reliable.

And we want to turn over a 2 ton vehicle over to this intelligence??

 

[Exiled_Child]

This article is devestating in it’s implications, and

the map accompanying this piece has to be seen to be believed:

medium.com/basic-income/self-driving-trucks-are-going-to-hit-us-like-a-human-driven-truck-b8507d9c5961#.qjctc02fi