Of the SAE's robocar "levels" only level 4 will be meaningful, and only partly

Topic: 

It's no secret that I've been a critic of the NHTSA "levels" as a taxonomy for types of Robocars since the start. Recent changes in their use calls for some new analysis that concludes that only one of the levels is actually interesting, and only tells part of the story at that. As such, they have become even less useful as a taxonomy. Levels 2 and 3 are unsafe, and Level 5 is remote future technology. Level 4 is the only interesting one and there is thus no taxonomy.

Unfortunately, they have just been encoded into law, which is very much the wrong direction.

NHTSA and SAE both created a similar set of levels, and they were so similar that NHTSA declared they would just defer to the SAE's system. Nothing wrong with that, but the core flaws are not addressed by this. Far better, their regulations declared that the levels were just part of the story, and they put extra emphasis on what they called the "operating domain" -- namely what locations, road types and road conditions the vehicle operates in.

The levels focus entirely on the question of how much human supervision a vehicle needs. This is an important issue, but the levels treated it like the only issue, and it may not even be the most important. My other main criticism was that the levels, by being numbered, imply a progression for the technology. That progression is far from certain and in fact almost certainly wrong. SAE updated its levels to say that they are not intended to imply a progression, but as long as they are numbers this is how people read them.

Today I will go further. All but level 4 are uninteresting. Some may never exist, or exist only temporarily. They will be at best footnotes of history, not core elements of a taxonomy.

Level 4 is what I would call a vehicle capable of "unmanned" operation -- driving with nobody inside. This enables most of the interesting applications of robocars.

Here's why the other levels are less interesting:

Levels 0 and 1 -- Manual or ADAS-improved

Levels 0 and 1 refer to existing technology. We don't really need new terms for our old cars. Level 2 perhaps best described as a more advanced version of level 1 and that transition has already taken place.

Level 2 -- Supervised Autopilot

Supervised autopilots are real. This is what Tesla sells, and many others have similar offerings. They are working in one of two ways. The first is the intended way, with full time supervision. This is little more than a more advanced cruise control, and may not even be as relaxing.

The second way is what we've seen happen with Tesla -- a car that needs supervision, but is so good at driving that supervisors get complacent and stop supervising. They want a full self-driving car but don't have it, so they pretend they do. Many are now saying that this makes the idea of supervised autopilot too dangerous to deploy. The better you make it, the more likely it can lull people into bad activity.

Update: One day after I wrote this, it was revealed that NHTSA shut down comma.ai's efforts to build an aftermarket autopilot citing these concerns, among others.

Level 3 -- Standby driver

This level is really a variation of Level 4, but the vehicle needs the ability to call upon a driver who is not paying attention and get them to take control with 10 to 60 seconds of advance warning. Many people don't think this can be done safely. When Google experimented with it in 2013, they concluded it was not safe, and decided to take the steering wheel entirely out of their experimental vehicles.

Even if Level 3 is a real thing, it will be short lived as people seek an unmanned capable vehicle. And Level 4 vehicles will offer controls for special use, even if they don't permit a transition while moving.

Level 5 -- Drive absolutely everywhere

SAE, unlike NHTSA's first proposal, did want to make it clear that an unmanned capable (Level 4) vehicle would only operate in certain places or situations. So they added level 5 to make it clear that level 4 was limited in domain. That's good, but the reality is that a vehicle that can truly drive everywhere is not on anybody's plan. It probably requires AI that matches human beings.

Consider this situation in which I've been driven. In the African bush on a game safari, we spot a leopard crossing the road. So the guide drives the car off-road (on private land) running over young trees, over rocks, down into wet and dry streambeds to follow the leopard. Great fun, but this is unlikely to be an ability there is ever market demand to develop. Likewise, there are lots of small off-road tracks that are used by only one person. There is no economic incentive for a company to solve this problem any time soon.

Someday we might see cars that can do these things under the high-level control a human, but they are not going to do them on their own, unmanned. As such SAE level 5 is academic, and serves only to remind us that level 4 does not mean everywhere.

Levels vs. Cul-de-sacs

The levels are not a progression. I will contend in fact that even to the extent that levels 2, 3/4 and 5 exist, they are quite probably entirely different technologies.

Level 2 is being done with ADAS technologies. They are designed to have a driver in the loop. Their designs in many case do not have a path to the reliability level needed for unmanned, which is orders of magnitude higher. It is not just a difference of degree, it is one of kind.

Level 3 is related to level 4, in particular because a level 3 car is expected to be able to handle non-response from its driver, and safely stop or pull off the road. It can be viewed as a sucky version of a level 4 system. (It's also not that different -- see below.)

Level 5, as indicated, probably requires technologies that are more like artificial general intelligence than they are like a driving system.

As such the levels are not levels. There is no path between any of the levels and the one above it, except in the case of 3/4.

Level 4

This leaves Level 4 as the only one worth working on long term, the only one with talking about. The others are just there to create a contrast. NHTSA realizes this and gave the name ODD (Operational Design Domain) to refer to the real area of research, namely what roads and situations the vehicles can handle.

The distinction between 4 and 3 is also not as big as you might expect. Google removed the steering wheel from their prototype to set a high bar for themselves, but they actually left one in for use in testing and development. In reality, even the future's unmanned cars will feature some way in which a human can control them, for use during breakdowns, special situations, and moving the cars outside of their service areas (operational domains.) Even if the transition from autodrive to human drive is unsafe at speed, it will still be safe if the car pulls over and activates the controls for a licenced driver.

As such, the only distinction of a "level 3" car is it hopes to be able to do that transition while moving, on short but not urgent notice. A pretty minor distinction to be a core element of a taxonomy.

If Level 4 is the only interesting one, my recommendation is to drop the levels from our taxonomy, and focus the taxonomy instead on the classes of roads and conditions the vehicle can handle. It can be a given that outside of those operating domains, other forms of operation might be used, but that does not bear much on the actual problem.

I say we just identify a vehicle capable of unmanned or unsupervised operation as a self-driving car or robocar, and then get to work on the real taxonomy of problems.

Comments

If the robocar cannot drive to everywhere, can it drives only in the city or some of rural where has been installed the proper V2I?
If so, robocar could not replace the traditional car by its restricted usage, I guess.

It is not necessary to replace the car with one vehicle. The world will consist of self-driving taxis which take you where they can drive, regular Taxi/Uber style vehicles which have a human take you where they can't go, cars for hire by the minute for you to drive yourself places it's too expensive to use an Uber, and more. All these can replace a car.

But it is not necessary to replace all the cars, certainly not in the early days. Did anybody tell you otherwise?

Of course, we don't need to replace all traditional cars to robocars.
Actually we can't afford it, I think.
I hope robocars can drive everywhere by itself without V2I or something like that.
Or else people have to keep their traditional cars or lent a car to get some place without autonomous infra.
I think your definition on level 4 could match my needs.

Once robocars are ready to be autonomous, only robocars can be operating in one area. The change must be very fast, that is the reason I am talking about areas. It will be not possible to have enough robocars in few days for several cities or countries. So once there are enough robocars for one area, that area will have only robocars. The main problem RC faces is if they will make many fatal accidents. The main problem are not this accidents , but whom to make guilty: the owner, the maker, the seller, the council regulator, the software programmers, and so on. If there are only RC in one area, the possibilities than a RC run over a row of children or a blind man, etc, will be very few. If some fatalities occurs, will be a lot less than today so the accident will be analyzed, then see what must be changed,(software, camera, street surface, pizza advertisement, blind man stick , children behavior). But people will be not angry with RC, just say before was worst. But with normal cars running, ALWAYS will be accidents involving RC and cars, like happens today with the autonomous prototypes or Tesla style vehicles . In this cases, of course, the man driver (specially his lawyers), will blame the RC, and try to find a mistake in any part of the RC, software or hardware either. So RC will be obliged to change and change every time an accident occurs, to calm the public an authorities. Besides that, all the efforts and money can be put to make the things easy and safer for RC and the people interacting with them. On the other hands, with normal cars still running in the area, still will be necessary spend money in signs, rules, etc, for man drivers; signs that besides must be easily understanding by RC, but really RC do not need that signs, plus do not need policeman, radars, cameras, than check speed, seat belts, drivers permits, etc. Besides that, who will risk to take a car, driving for a driver that may be is drunk (even you), do not know really which street to take, which speed is safer , that is worried about to lose his job, that can be your partner or just a paid professional, it does not matter; driving a car that is not 100% sure is in good conditions, plus is a type of car that is involved in the 99.99% of the accidents that caused 50.000 deaths in the last year, having the possibility to take a RC, 100% safe, that will take the faster route according to the traffic at that time, or now the best panoramic route, if you prefer, and you do not need to know how to drive, etc?. Everybody or family in the area will not need to own a RC, even no one need to own a RC in the area, but everybody in the area will have the possibility to use a RC, either if he lives there or just is a visitor to the area. How to organize is theme of other block, but to have only robocars in one area, will be less than the money people in the area spend today, to have public transportation , plus taxis, plus some of them owning a car, and besides anybody can run the RC, not need driving permit. *and insurance will be near zero with RC only*

I believe a quick switch is simply not a possible option, and most people agree, so you would have to come up with better justification for your assertion it would happening. In fact, I would say the slow switch is already fact, not speculation.

This will be a short answer, then will come the total answer. I post many comments in this section, they will help to start to understand the whole idea. Do you remember the fight between VHS and Sony Beta & Video 8?. Do you remember Iridium mobile phone project?. What happened and could happened with the analog HDTV?. Those 3 things can help to understand my idea too. Is not exactly the same, but what would happened with airplanes if they were always using the car routes instead airports?. The RC is not a car with some improvements,is something so different and besides is
accompanied by robothings that will change a lot our life. So , of course we must test the RC in our normal environment of today, even deploy some of them in limited areas and amounts, but the total deploy and in big quantities, cannot do in the way like: LA has 50 % of RC, Paris 20 %, Tokyo only 10 % but increasing 2% every week, etc. It must be done area by area, 100%. I wrote before : Any system that would need in some moment a driver takes control of the car, would not work.1) Will be not safer than man driving. 2) will be more dangerous like happened in Tesla accident. 3)The drivers will became less experts because normally not driving, so they will be less prepare than now to handle the situation. 4) will kill the best of RC, that anybody, any age, any physical condition can ride alone. So I dismiss the idea of autopilot, as a final , universal product, after the period of testing it, to get information for a better RC, and that is going to finish soon. So just let talk only about 100% autonomous RC. In the base of the idea, is the concepts of use, share, utility, security, that will change the concepts of property, profit. Is not something new. When you buy time-sharing , you do not buy a flat, just the right of use something for a while in some part. So the total substitution of cars by RC in one area must be done under that concepts of use, share, etc. The need of an area is because, not a whole city or country, will be ready in many aspects, to support RC, may be Singapur . So if we can have 20, 30, 50 thousands of RC out of the factories everyday; instead to have 10 new RC everyday in 5.000 places around the world, making accidents or just trouble, in a lot of places, better you put all of them in an area that is already prepare to receive them, and in the meantime you prepare an area for tomorrow and so on. If it takes 1 min to see the Gioconda, you give 43200 tickets to 43200 people interested, each ticket with day and time, min by min, and everybody will have his min to see, without stress or wasting time.Some will be the first, some will wait (at home) 1 month, but all of them will see it in not more than 1 month . On the other hand, people will be queuing for many days, fighting, people with money will pay a lot to somebody that has a good place in the queue. But any way, somebody will be the first , somebody will be waiting (at the queue) one month, and anyway all of them will see it in not more than 1 month. For the same result, which way is better?. So if we can produce 100.000.000 RC every year, we can deploy around 30.000 everyday in 10 areas, in 10 years will be 1.000.000.000 in all the world, may be enough. Yes, you guess, not everybody, not every city, not every country will have the money, or the skills or the infrastructure, for RC. The infrastructure can be done, the skills can be learn, the money... Who needs money, for what?, to pay the robots that do not need to eat, to play soccer, to sleep?, To buy the energy, from whom?, the sun, the wind, the rain?... To buy the natural resources, who made them?, I will pay to him, give me his name, address, mobile phone. I told at the beginning, now we have the tools that allow us to think in a different way. I will write more about that point later, but this is the robocar blog, I will continue with the issue. The main thing is not HOW to deploy only RC in one area, the main thing is that few RC in one city, will cost more money, made a lot of trouble, could not be used getting all the RC advantages,may be will be regulated, ruled in a way that will cut many of RC advantages. What we do today? Build highways, routes, cars, put signals, lights, that make impossible to ride a horse in that conditions, so we banned the horses. I want to banned the normal car before it causes more trouble and fatalities than today. Not in the whole world from one day to another, possible but not desirable. The RC will help, but if we do totally area by area.Later will explain more

While I welcome comments here, if you want folks to read them, please put at least 5 to 10 times as many paragraph breaks in your items, and trim them down to be more concise, if you can.

STARTING POINTS
1) Robocars is just one of the robothings that will change a lot our life. 2) Looking about the death toll from Tobacco, HIV, wars and terrorism, traffic accidents, it sounds that we pay poor attention to cars, compared with the attention we pay to the other things. 3) Attending point 2, it seems that long time ago, cars would be banned or heavily regulated . 4) Cars are heavily interacting with our actual life, so now we can substitute them, and quickly banned them. 5) If we do not do the things in the proper way, we will lose many of the advantages of RC, plus may be the things will become worst. 6) We must no t forget that RC and robothings will be deploying at the same time, so one will help the other to make the things better, but at the same time one will be blame for the trouble than the others make. 7) RC will do that some people lose their jobs, but other robothings will make it faster and in a more visible way. 8) There is a risk that all RT, including RC, will make social unrest, with people dying trying to destroy the factories where robot works, with the result that nobody will have the robot produced things , neither people will recover the lost jobs. 9) We must start to pay attention to concepts like love, share, use, positive things, voluntary work, solidarity, help, peace, safe, ecological, self -sustained. 10) We must not think more about concepts like hate, countries, borders, property(social or individual), war, army, profit. 11) Since this is the RC blog lets go to that point, something can be interpolated to other RT.
ROBOCARS FULL AUTONOMUS ARE READY OR WILL BE SOON. IS THE SAME
For sure RC are ready to run full autonomous in mostly of the city streets , sure any highway and all routes except some very bad countryside ones. In some areas RC will need to be tested, in some mountains routes, may be , some special city areas. Some very very and rare areas like jungle, etc; well, we still have some years in advance to have enough RC to go there, so let it for the future. RC will have not more problems , or be more dangerous, with pedestrians, bicycles, children, skaters, than today cars. So at least, if not less, will not make more fatalities than cars (for sure will be less, if not zero for a time).

ONLY ROBOCARS . WHY?
For sure the main problem RC will have will be with the man driving vehicles. Motorbikes go to the theme parks, just NOW. (A motorbike is not a vehicle, is a gun, and not always a gun kill somebody, that’s all). So pedestrian, bicycles, can be crossing or using the streets like today, with less danger from RC, than today. (then we will see how to avoid a child running follow a cat or a ball)
The traffic with normal cars mixed, will be more complicated for RC (as today is more complicated to the every car). Normal car will continue making accidents against other normal cars as today (no reason that it will not occur). So, may be, will be less accidents just according the decreasing amount of NC in the area. But must be added (1)the accidents that normal cars will make against RC, because the RC behavior, even correct, will be different to the man driver behavior, so it will confuse the man driver. Plus (2)the accidents than RC makes because unexpected action of man drivers. Will explain 1 and 2 . (1)As an example, I do not know exactly, but many of the Goggle car accidents were because a car kicked GC from the back (man driver car fault), why?. May be a yellow light turn on, as correct action GC stops. BUT, because GC nows exactly the distance to the stop line, plus can reacts and stops faster than a man, GC started to stop a little after a man driver would start to stop, so the back driver though that the GC will continue and cross with the yellow/red light (illegally, but normal man driver behavior), and continue at the same speed. When GC finally stopped, it was too late for the following car to stop. I do not know if it really happened but similar things can occur between RC and NC. (2) Many illegally actions than a man driver do, we know in advance looking his face, even eyes. Can RC do that? Other things we know just because experience (This car ,stop where it must not be, means that it will make and U-turn, where is not allow), so you take an action may be not legal, but knowing what the car in the front will do, and knowing that the car on the back will know what you will do, so nothing bad happens, even everybody makes an illegal action. Yes, it can be 20 different reasons, that the RC knows, why the first car stop, so the RC took the safe, legal action for the 20 possible reasons, and cause an accident because none of the other man driving suppose what will RC did. So RC will be involved in accidents that RC are not guilty, but happened because normal cars are mixed with RC.
Other bad things about the mixed is the limit speed. RC will drive always at the top of limit speed, steadily. But normal cars will run slower, and normally over the limit in 10, 20 ,30 or more Kms, depending the street, place, hour, or where the police or radar control are, and drivers know. It can make more accidents, because at some hours, in some places ALL vehicles run at the same, fast, over limit speed, in a safe manner, and a RC at a low, legal speed, can causes accidents. This samples and others will make RC run slower than the way RC can do if only RC are running.
The main conclusion is that RC mixed with normal cars will lose many of their advantages,and make it more difficult, expensive, and less desirable, safer, faster, the total deployment of RC in one area.
On the other hand, once RC start to match the amount of NC, and even be more than NC, things will start to be more difficult for NC., like:1) less parking places , gas stations ,because less NC. More not allow park here, because parked cars makes less place to running cars. More and more shops just with stop and drop/pick up space (no parking lot), RC passengers do not need a car waiting at the door. 2) Less body and motor repair shops, less or 0 driving schools, business that are going to finish (just guess which others). 3 ) people will not want the Council waste money in radars, traffic policeman, new traffic boards, traffic lights, for few NC that will vanish one day (that is an important reason, the faster RC deploys , the better). 4 )man driven taxis and man driven buses will be more and more non profitable, because RC was a lot better service , and cheaper. 5)may be the most important “I prefer to wait 2 hours for a safe RC than go on a man driving taxi or go with you on a NC that has xx% possibilities to have an accident.”
I wrote about some problems that the mixed between RC and NC, can occur. Besides because that problems may be regulators will cut some of the possibilities of RC, and ask for more and more test and hardware and software to avoid the problems that will not exists if only RC run. I cut with this arguments now and in the second part will explain how to deploy area by area.

WHAT HAPPENS IN A SLOW SWITCH
Just think about 100, 200, 300 cities in the world, where easily 80% or more of the city area can be already ready for RC. Every month we deploy 3% of the RC each city needs, so may be 3% of NC are withdrawing. That means than in 3 years will be 100% of RC in all those cities and 0 NC ( have been enough time to proportionate the areas of each city that was not suitable for RC. In the first year ALL cities will suffer the problems because mixed RC and NC. Plus the stress because not enough RC for what the people wants. Can imagine other problems because the acceptance or not of RC or NC. After 1 year will start the problems that we see before for the NC, the last 6 months, could be a nightmare for the NC owners , you can imagine.
After 3 years you have those 100, 200, 300 cities with 100% RC but after 3 years of troubles.
If we do totally city by city, in 1, 2, 3 weeks each (depending the city), in 3 years all will be 100% RC with no transition problems. Even the experience of the first city will help to do the things better in the second and so on.

Step 1. R&D on Robocar in the lab & test field.
Step 2. R&D on Robocar in test fields & normal roads.
Step 3. test & adjust robocar in normal roads.
Step 4. adjustment & merchandize
Step 5. Roll out the new robo car
Step 6. drive it at the normal road or certain road where robocar only or not.
Step 7. spread of the relevant infrastructure through the whole land(if possible)
Step 8. spread the territory of robocar
Step 9. people switch their cars to robocars from the traditional cars
Step 10. come out of the traditional car ristricted zone or city

This is my set for the robocar & robothings.

Your point is not that different to me. I guess both of us take picture the same or similar future.

But I think we are standing at the step 2.
And I hope that robocar & robo logistics vehicles can go an errand in my city mixing with normal things(traditional car, bycicle, pedestrian, blind, kid, elder, etc).
It is step 6, I think.

Robocar & robothings could give us buch of benefits such as safety, money saving(if robocar sharing is in popular & its cost is quite reasonable), diminish of parking lots, invitation of much green than traditional parking places....

At the first time, robocar is running in the robocar only places, how can it learn better coping methods from the natural humanic mistakes?
Robocar only zone would become just a large laboratory, I think.

Why don't you Come out to the road, enjoy some mistakes & error.

Thanks Brad & Alejandro
This place give me quite a good study.

If level 5 = go anywhere, maybe today's Uber flying car announcement is level 5? Why bother dealing with the corner cases that level 5 cars would need to deal with if you can just hop over them with a flying car?

So I am not sure if flying cars -- and Uber did not announce one, only the plan to use ones others will build -- fit it any taxonomy of road vehicles. They could fit in a taxonomy of vehicles.

I agree about to focus just in level 4. Robocars are hitting the road at the same moment that a lot of robo-anything are popping up around us. Life will change a lot and at a speed that humans still cannot manage. We must not think about countries, just about communities. We cannot think more about property (private or state is the same), just rule about the right of use. We must prepare to share , to move to one community to another, to love each other , to be solidary.
Coming back to the topic, level 4 robocars must be running in determinated areas, where only robocars , pedestrian and bicycles, can be. No other type of vehicle can be in the area. The areas can be connected between others robocar areas or no robocars areas by any media. People will choose to live in a robocar area or not. Same thing will be for the roboanything, but will occur that will be areas with a lot of robothings, areas with near nothing, areas 50 - 50 %. people need time to learn and adapted, so then they can change from one area to another. Will elaborate later.

I agree that people need time to learn & adapt rococar or robo things.
I think that the robocar & non-robocar could be mixed in nature mod.
Seperating zone could cause another social & entrepreneurial cost. At the end, that cost would be turned over to customer.
Robo car & robo things such as autonomous logistics robot will be tested and operated in our neighborhood society, I think.
We have to walk a long way.

I agree that people need time to learn & adapt robocar or robo things. I agree, but changes today are faster than human adaption, so we must use the possibilities of the faster and big changes to help the human adaption. So we cannot do things in the traditional historical way.
I think that the robocar & non-robocar could be mixed in nature mod. It cannot be . The Tesla accident is an example. (I wrote more below). The main problem is that man driving cars will cause accidents even to Rcars. So people will blame RC or ask to ban man driving. Only robocars full autonomous will cause very few accidents that people will tolerate.
Separating zone could cause another social & entrepreneurial cost. At the end, that cost would be turned over to customer.
Not to change by zones will cause a big social....etc cost. Will be difficult to own a normal car where there is no gas stations, no traffic signs. The money must be used for the changes necessary for RC. Where will you park your car? Parking places will became very expensive, robocars do not need. What are you doing with your cash money ?, No ATM, no banks , no shops where to buy with cash. The faster the change, the cheaper and the less social problem. You cannot have 1.000.000.000 RC from one day to another to change in the whole world , but can have 100.000.000 in one year, so in 10 years all the world with RC. So everyday day change an area of about 30.000 vehicles, and everyday prepare the other areas for the change. Everybody will have time to prepare, to dismiss the old cars area by area (the still good ones can be used to substitute the very old ones in other areas, so not necessary to produce old cars and always will be customers for the new robocar).Same things can be done for any robothing, like mobiles,electronic money, robowaitress, robocleaner. Yes the things must be organizing in a different way. We could not do it before, but now we have the robothings, that can do many things for us, so if we use them properly , we can achieve what we could not do before. Just love, think , organize. Yes, this bis a theme for another block, but this is the way. In less than 20 years many jobs will be not exist, half of the people will have not work to do. What are we going to do? burn the robots?, we cannot they are making possible many things we already have. So changing area by area will be the way, in the meantime it will be necessary to support the life of the people in the areas, where today is difficult to survive. There are enough military budget, soldiers, trucks , etc all, around the world to do this.

Robo car & robo things such as autonomous logistics robot will be tested and operated in our neighborhood society, I think. That is what is happened today, in few years and already now, the test are finished or will be finished soon.

Good insights, Thanks.
I agree with you in many aspects.
The different point of our view is that when is the proper time to RC & Non-RC would mixed on the same lane.
Of course RC must be tested at some test bed before it comes out to the Market, road.
20 or 30 years after the road might be full of RC, but before that we'd better to think about the mixture use of those car at the same lane, same road. You mentioned Tesla case. Unfortunately the Tesla model was not the RC. It's man's fault, I think.
If RC could not approach whole the land, separating zone(RC only or Non RC only) might cause inefficiencies in many aspects.
So I told that we have a long way to walk.
We have to invent, test, operate, agree, adapt & revise on RC BEFORE it rolls out to the Market.
When it come to the market, it has to be mixed with the original one.
I think the meaning of RC is protect any manmade accident even at the normal road with Non-RC.

Dear Hunt. Is nice to talk with you, it makes us to think about our ideas to make it understandable for the other people. I will be back in about 24 hours, need time to sit to type. Alex. Thank you Brad for this very good site.

As a city-sized land developer, We, actually my company has a plan to restrict normal car within our entire project site(app. 5,200acres).
But I think it's a very radical and ambiguous future plan.
Robocar will not be ready in couple of years, right?
In that aspect, I hope that traditional car and robocar could share road with C-ITS or something.
We can give some good information & service to the traditional car driver and at the same time robo car can use those information for its self driving with something like C-ITS, I think.
In my opinion, robocar is rather far to touch at the every market. It will take more time to show up to us.
Until it comes out to us, we have to give more functions to our traditional car. We call it level 2 or something?
People will adapt to what they need to need.
It would not take more time to adapt robo car than I-phone & android's case.

What is important & meaningful variable for people is touchable tech & acceptable risk on robo car.

There is no perfect one, except we all die someday. lol.
Good day.

Another part of the answer I agree with Brad. The RoboCar must be full autonomous. A 1% supervised RoboCar will be a 50 % or less useful RC, and for sure a lot more dangerous than a full RC. What happened in the Tesla accident? A must be supervised car was not supervised and crashed to a not autonomous car (that may be was doing something that a RC would not did). The use of RC, the rules to build RC, the places where RC must ride, must be done to AVOID accidents not to find who was guilty. Once full autonomous RC would be running , everybody will want to use them (not just to own them). The actual car, cannot be used for everybody (always somebody must drive), the full RC can be used for anybody that can walk or move in some way (may be with a robowheelchair) to the proximity of a RC. Besides that, a RC can come to pick up you even if you are the only person in 100 km around and you have no idea how o where you must drive. And because man driving or man supervised cars will be making accidents , even against full RC, people will soon asking for a rule that "forbid man driving vehicles in this town". The utility of RC will quickly override the proud of "I own this car". Since the RC cannot be driven, nobody will need the pleasant to run in a MB or Ferrari, if you want,buy one and ride in the future designated places for man driving vehicles. Because everybody will need to ride a RC, different ways than ownership or Uber style taxi companies must appears. Not new things , just club users that live in the same area, or work in the same company or play tennis together. Even every city can have RC free of use for independent users. In other part I wrote that we must start with robocars and many robothings by areas. Because the interaction between the actual things and the new robothings will be caothic, dangerous, and really impossible. It must write later more about, but just think in a few samples: A city or just an area where there are no man driving taxis, because it is not profitable, because RCtaxi., there are no gasoline stations or even not EV charging points, because not enough independent owned vehicles. The city do not want to waste citizens tax money in lights, signs ,etc for drivers, that day by day will be less and less. Any type of shop will accept only electronic money (credit, debit card, mobile transfer). No ATMs, no banks (money transfer between internet connected devices). Can a person without a mobile, some kind of electronic money, that do not want to take RC live in the area? Can somebody without that things, but driving a car risks to go there?. Of course is just not to isolated the area,just organise the things that will permit this type of people live or go there in the meantime they learn about the new things. That is the reason I wrote before that the things must be introduce area by area all together and very quickly. And the way area by area means available to everybody in the area, not just about who can pay and who cannot, like the public education, that mostly every country has.

In the 1960s people extrapolated from a few successes in chess and chatbots to believe that full human-replacing AI was just a couple of years away. I feel like we're doing that now in autonomous vehicles. Making a software agent that can drive exactly as well as a human in all the situations that humans drive seems to me the wrong approach. (We certainly don't want cars to drive exactly like humans!) Better to take the 90% of driving that is quite easy to automate and automate it. Figure out what is causing those 10% situations and just do something else entirely with that. A great example is your idea of rating roads and driving environments. That makes a lot more sense. Imagine a freeway rated "you can actually go to sleep here".

As I said, the level 5 is an aspirational science fiction concept, and the level 4, which is a robocar with a limited domain, is the only project of importance.

Given this frame of mind, it would seem like Otto and Comma.ai and Tesla are actually on the right track. And Google is on the wrong one. Google wants to drive around Mountainview, which is nice and, as you say, aspirational, but it seems like the real benefits now could be pushed on the interstates. By certifying certain roads or lanes as largely free from things that driving AI will find overly complicated, it would seem we could get some exit to exit autonomous driving _now_. Even if the freeways had to have special markings or locating beacons/references, how could it not be worth it?

But Google still claims the forefront of publicity about autonomous cars when I know they will never make a driving AI that will comprehend what I do on my bicycle.

A car limited to interstates would not be a meaningful level 4 -- this is in fact the primary reason people started talking about the level 3 idea, which Google played with, the standby supervision.

That car can't take you door to door, so it provides only modest benefit. All the benefits of taxi service, self-refueling, self-parking are gone.

Google built a highway car years ago and abandoned that for reasons they have outlined.

Call it a meaninful 3.5 then. If you could send big rigs down the interstate with no humans on them (they pick up/drop off drivers at the exits for the in town last mile) or let people driving on long hauls sleep then I think that would be a profound change that would spur all kinds of concomitant progress in autonomous vehicle technology. And as far as I know, (as you point out with Google) that technology is ready to roll.

Or I'm not understanding the subtleties of the levels here. I'm proposing a car that can be 100% autonomous on 10% of the roads or 90% of the distance for long haul driving. The rest of the time the human driver would need to be 100% driving.

I would argue the only relevant question is: is an L3 car with OK software and crappy human supervision after than "just human"?

If the answer is affirmative (which Tesla is claiming for its L2+ autopilot), then it is a moral impherative to deploy it as fast and as widely as possible.

In fact, one may argue that under such circumstances, insisting on L4 is immoral.

Do at the end of the day, the discussion boils down to numbers: will we save more lives deploying L3 today, or wait for L4 tomorrow.

And until Tesla publishes concrete data, we can only speculates.

Tesla's plan is indeed to get to a fully unmanned capable vehicle -- they are effectively advertising it already.

Their plan is to get there by taking a supervised car and learning from millions of supervised miles. My estimate has been that this is the wrong path, better to go directly for the real target as Google has. However, over time I have reduced my confidence in this opinion, though I would not say I have given up on it. I rated it as low chance to work before, now I rate it slightly higher.

However, I am in favour of different approaches being tried, as long as they are not reckless, with the goal of getting to the best technology as quickly as we can.

The question of "decent autopilot and poor supervision" is a very difficult one. I suspect what that means is that while the average performance for that hypothetical vehicle is good, for the really bad supervisor (ie. a guy watching a movie) it is an unacceptable level of safety.

NHTSA is saying to no longer depend on the driver, which may cut off this route.

Understand that TESLA's system as it sits (data on new system is yet to come) is pretty poor. It does not see things like a truck crossing the road in front of it. There are many other things it doesn't see. I don't think we are quite ready for "It can't see a truck crossing in front of it, but that only happens once in 200M miles while the driver is not looking, so the safety is high enough."

I agree with Brad. The RoboCar must be full autonomous. A 1% supervised RoboCar will be a 50 % or less useful RC, and for sure a lot more dangerous than a full RC. What happened in the Tesla accident? A must be supervised car was not supervised and crashed to a not autonomous car (that may be was doing something that a RC would not did). The use of RC, the rules to build RC, the places where RC must ride, must be done to AVOID accidents not to find who was guilty. Once full autonomous RC would be running , everybody will want to use them (not just to own them). The actual car, cannot be used for everybody (always somebody must drive), the full RC can be used for anybody that can walk or move in some way (may be with a robowheelchair) to the proximity of a RC. Besides that, a RC can come to pick up you even if you are the only person in 100 km around and you have no idea how o where you must drive. And because man driving or man supervised cars will be making accidents , even against full RC, people will soon asking for a rule that "forbid man driving vehicles in this town". The utility of RC will quickly override the proud of "I own this car". Since the RC cannot be driven, nobody will need the pleasant to run in a MB or Ferrari, if you want,buy one and ride in the future designated places for man driving vehicles. Because everybody will need to ride a RC, different ways than ownership or Uber style taxi companies must appears. Not new things , just club users that live in the same area, or work in the same company or play tennis together. Even every city can have RC free of use for independent users. In other part I wrote that we must start with robocars and many robothings by areas. Because the interaction between the actual things and the new robothings will be caothic, dangerous, and really impossible. It must write later more about, but just think in a few xamples: A city or just an area where there are no man driving taxis, because it is not profitable, because RCtaxi., there are no gasoline stations or even not EV charging points, because not enough independent owned vehicles. The city do not want to waste citizens tax money in lights, signs ,etc for drivers, that day by day will be less and less. Any type of shop will accept only electronic money (credit, debit card, mobile transfer). No ATMs, no banks (money transfer between internet connected devices). Can a person without a mobile, some kind of electronic money, that do not want to take RC live in the area? Can somebody without that things, but driving a car risks to go there?. Of course is just not to isolated the area,just organise the things that will permit this type of people live or go there in the meantime they learn about the new things. That is the reason I wrote before that the things must be introduce area by area all together and very quickly. And the way area by area means available to everybody in the area, not just about who can pay and who cannot, like the public education, that mostly every country has.

Wasn't the original DARPA Challenge basically an off road autonomous vehicle that had to navigate rough terrain?

From Wiki: "Vehicles in the 2005 race passed through three narrow tunnels and negotiated more than 100 sharp left and right turns. The race concluded through Beer Bottle Pass, a winding mountain pass with a sheer drop-off on one side and a rock face on the other."

Yes, even in 2003 they knew that "level 4" was the main interesting project.

In that case it was because the military wants a way to move cargo in war zones without putting warfighters at risk. For them, only unmanned operation will do.

I always wondered about that military cargo use case. Couldn't they just dronify the trucks and drive them from a safe location? With real people remotely at the wheel?

If you could assure the data networks you could. In the sky they are more confident about that. On the ground it's not too hard to overwhelm signals, even if you had complete coverage.

As there might not be mapped roads for military vehicles to work upon. Closer to the front, there may be nothing but bomb holes and blown out bridges and the like. In places like Afghanistan that mountain pass might just be a clearing between rocks. Is this not similar to the level 5 situation?

But every driving situation means every driving situation. The military ATV might handle that but will it handle highway driving, or the toll booths on the golden gate bridge or everything else anybody can name? There just isn't a market to do that the hard way, not yet.

A vehicle that can do every driving situation needs to understand the world at the level humans do. Which is coming some day, but for other reasons, not driving.

Brad, regarding level 3 - well 60 seconds warning doesn't work. But what about 10 minutes warning ? say for example for a truck driven automatically on the highway, and than, some distance before entering the city(a few miles), the driver resumes control ?

Like Uber/Otto is trying to do ?

I would suspect that will be safe. It doesn't match the typical vision most people had for level 3, which includes unscheduled handoffs. That's why I called level 3 "standby" supervision. A person is not actively supervising but standing by. Not sleeping, as Otto hopes.

Since I am calling for abandonment of levels based taxonomy, it's not very important to pick the nits on this, but I would view a system of scheduled handoff more an example of operating domains.

You could consider a car which lets the driver sleep but is fully capable of pulling over to a safe spot and stopping, then waking the driver. Problem is, you are not technically supposed to pull off the the side of the highway except in emergency situations.

I envision the autonomous driver being able to depart from a station much or exactly like a freeway rest stop and arrive at another. Can you think of more boring driving from a human perspecitve (throw in traffic)? For such a trip the driver should be able to be completely off-line. I like your term stand-by. If there is a situation where the car has to pull over and wake up the stand-by driver (or call in a technician for unmanned trucking) I think (with the right kind of road) we can assume that it would indeed be an emergency.

If it's a rare thing, it could be counted as an emergency.

I don't really fit a sleeping driver as one who is standing by, but there are various definitions of the term. There are many variations possible. The one that people discuss from a safety standpoint is the one where the car demands you take over while moving, with 10 to 30 seconds of notice. This is what most people were thinking about when they wanted to create a "level 3" concept. Again, I think this is a rare mode, a possibly unsafe one, and not really a deserving part of a taxonomy.

Any system that would need in some moment a driver takes control of the car, would not work.1) Will be not safer than man driving. 2) will be more dangerous like happened in Tesla accident. 3)The drivers will became less experts because normally not driving, so they will be less prepare than now to handle the situation. 4) will kill the best of RC, that anybody, any age, any physical condition can ride alone.

I think Uber has less than 10 years. It is running in some areas. It is running with some small problems in some areas. In some areas had been regulated more or less. It was forbidden in some areas. But I think in everyplace Uber was fight, sometimes very aggressivily , by taxidrivers. It is a lost battle for drivers. In few years, and already now, Will be not Uber drivers, just Uber robotaxis, and then Uber may be will finish, robocar clubs will appear, and even public robocar transportation. That is the speed of the change, but the robocars and the robothings will continue growing.
The real good use of the robothings, will be when we change the idea of the right to property (state or private is the same), for the idea of the right of use. Once we understand that we do not need to own everything we want but instead we can use everything we need, we do not need to have countries , just communities. We do not need borders,army, just a small piece of the earth where to live, that we do not need to own, just to use. And robothings will help to achieve the goal. Let the robots grow, Just think how to better use them, not for profit, just to make our life better.

Add new comment