The Autonomous Cars News, Development, and Discussion Thread

[lrt's friend]

Originally Posted by M II A II R II K

Here’s the real nightmare scenario for self-driving cars

https://www.vox.com/energy-and-envir...ising-business

Advertising has a limit as a source of revenue. We already know this as Internet advertising is killing traditional media. We cannot have an economy based on advertising alone. I cannot see advertising alone being able to pay for our entire transportation network.

Also, the basis of the argument is faulty. While competition tends to lower prices, limited competition has the effect of increasing prices as we move towards monopoly and pseudo price collusion. There are many examples of this. While it may be possible to implement other software to compete, which is the basis of Uber and Lyft, it is much more difficult and costly to manage fleets of millions of AVs. Uber and Lyft are software companies, while the cost and risk of ownership is with individual car owners. It is an enormous leap to move these companies to own the vehicles as well. The cost of these companies assuming those risks will be borne by the customer.

The pressure will be on to assume control of transportation and remove individual car owners for safety reasons. At that point, convenience pricing especially during peak hours will increase rapidly to generate sufficient profit. A perfect example of this is Highway 407 in Toronto that was privatized and tolls were increased dramatically to maximize profit. The highway ceased to perform its original purpose to reduce congestion on the parallel and public Highway 401.

[Hatman]

The great Luke Starkenburg has filmed a trip on an autonomous bus in Germany:
• Video Link

[sundah]

I've just discovered this forum which is unfortunate, as it could have been of considerable help in research I have been undertaking. Actually this has been an update of research I made in the early 1990s on how automated vehicles on automated highways could alter urban transportation.

The outcome of this later research can be seen at:
https://freedominu.com.au/learningzo...ntvehicles.php

Essentially, this research and the prevailing thrust of the forum are in accord. If I had been aware of it earlier I may have made more of the prospect of ultra-light vehicles for commuting, and I had not foreseen the likelihood of much longer commutes as vehicles became more lounge-like, or the possibility of the 'last-mile' of trips to a CBD being by public transit.

However I believe I have something to add, as indicated by the following abstract of the paper and one of its illustrations. There is a similar drawing of a New York type CBD.

"Automation will increase the capacity of freeways three or more times and reduce costs by a similar factor. Twin-deck, automated, light-vehicle-only tunnels can increase capacities nine times and reduce travel costs to about half those of a conventional urban freeway: and this with few NIMBYs, residents or property owners to mollify. All personal motorised trips to CBDs having transportation requirements like those of New York could be provided in ten tunnels not much larger than a conventional two lane tunnel. Those like London or Paris would require seven. Shared-autonomous-vehicles will greatly facilitate such possibilities by reducing the need for for parking, even negating the need in inner city areas. Costs will be significantly lower than conventional public transportation and should be met by users (no subsidies). These are powerful possibilities, yet the extent they will be realised will depend on the particulars of particular cities and the distance of a particular time."

"The drawing depicts a quarter of a London or Paris type CBD, showing tunnelled automated highway requirements if all personal vehicular travel were by automated cars and other light vehicles. It is close to scale although the number and sizes of the buildings is surmised.

A tunnelled AHS loop has six levels of four lanes (three for capacity and another to enable the necessary weaving and merging). One of the spurs to CBD stations is shown. Privately owned vehicles would return along the spur, park themselves an wait to be summoned. Shared vehicles would go where ever needed to pick up other passengers.

The number and size of the parking buildings (drawn in this case in undisguised functionalist manner) is based on the assumption that 10 percent of workers would use private vehicles, while non-commuters would use 70 percent. It could be both possible and desirable to use existing streets to distribute and collect passengers, or some combination of surface and below ground facilities.

[sundah]

[sundah]

Image on www.pbase.co/sunduh

[Hatman]

Pictures are difficult to post in this forum. You need to know the image's address, then post that between the sets of [img] brackets. If you are using chrome, right click on the image you want to share and select 'copy image address'.

I agree with your premise that segregated roads will gain much more capacity through autonomous vehicles than regular surface streets. This should be intuitive, just like more metro trains can fit through subway tunnels per hour than light rail trains running in a shared-traffic lane per hour. If you give one system complete control of its infrastructure, it will be able to optimize to great results.

So should we build tunnels for cars the same way we built tunnels for subway trains? Elon Musk's Boring Company seems to think so (or at least they did before he announced a refocusing on pedestrian uses).

I just don't think the economics work out. Tunnels are very expensive not just to build but to maintain. Anyone can dig a hole in the ground, that part is relatively easy. The hard part is keeping the hole open; ground water and shifting ground materials cause the tunnel walls to deteriorate over time, and constant inspections and replacements over many miles concrete lining is not a cheap task. So, if you're going to build a tunnel you should want to push as many customers through it as possible in order just to pay for the darn thing in a reasonable way. This means passenger density, and no matter how efficient autonomous cars can be, they will have a hard time beating the passenger density of an autonomous subway train.

That's just my opinion though, and your link looks fascinating. Thanks for posting, and I hope you keep coming back!

[dubu]

There's many good things about subways. Like lighter trains and there's not turns all the time. The heavy lightrail trains and turns wear down the rails faster. It's good to have on ground rail, preferably sorta straight and subways. Like half n half. That's my thoughts too. Though I think it's sorta self explanatory.

[llamaorama]

Something I haven't heard discussed but I think will impact traffic is autonomous vehicles being used for deliveries. That is going to add a lot more trips. If Amazon suddenly wanted to be like a pizza place and offer 30 minute delivery for basic items from local warehouses, they wouldn't be using a truck on an efficient traveling salesman route, they would need to employ single vehicles with only one or two orders on board in a point to point network. If every person didn't get out much and only drives to work and back every day also orders something every other day, it would add 33% more trips to our roads. Presumably the status quo is people may grocery shop once a week during off peak hours, and they eat out once or twice a week also during peak hours and prefer locations that are easy to get to. With autonomous delivery, some people are going to eat take out every single night from that cool place that's on the other side of the city and order a random roll of toilet paper using their alexa. They may stay at home but they still contribute to traffic volumes.

Autonomous vehicles are going to hurt transit but in denser cities I think they are going to clog up streets. Have you ever dropped off someone at the airport and it took forever because there were so many cars pulling in and out of the road by the terminal, creating congestion? EVERY SINGLE DOWNTOWN STREET is going to be exactly like that when autonomous cars come out.

Realistically what cities should do is charge road fees but make public streets free to pedestrians and cyclists and make transit cheaper. That will shift people to alternative modes when roads start to hit capacity. My only fear is that with street-level tolling the conservative suburbs will just privatize public rights of way and those private road companies will see pedestrians as freeloaders who slow down paying customers in cars.

[M II A II R II K]

4 Things For Transit Agencies to Remember in a World of Driverless Car Hype

http://transitcenter.org/2018/05/03/...less-car-hype/

.....

- The threat that automated vehicles pose to public transit is a political threat. Pundits, politicians, and activists are now widely speculating that self-driving cars might “kill transit as we know it.” While some of these pronouncements are an honest attempt to account for the prospect of AVs in transportation planning, others are made in bad-faith – with real world consequences. Nashville’s transit referendum recently went down in flames, aided in at least some part by rhetoric from City Councilors proclaiming that Davidson County residents shouldn’t vote to invest in transit because “driverless buses will be ready in 12 months.”

- Even if some trips in autonomous vehicles are shared (and sharing won’t happen unless cities incentivize it) transit will always offer more capacity. Many current debates that try to cast Uber-type services as “on demand transit” obscure the reality that they are actually 21st Century taxis. That matters, because dense cities can’t function well without affordable ways to move scores of people through the congested, bottlenecked networks in and around them. --- Automakers, investors, and developers of autonomous technology are increasingly warning that the rollout of automated vehicles will be slow and piecemeal.

- Transit agencies are already facing pressure to enter into partnerships with transportation network companies and launch “microtransit” pilots; autonomous vehicles may be next (in fact, small driverless shuttles are already being tested in Houston, Buffalo, and Atlanta). But testing new technology for global motor vehicle markets for its own sake is a job for universities, research agencies, and private R&D; transit agencies that experiment need clear theories about what new services can accomplish – like improved access, increased ridership, or safer streets – and a plan to measure results against those goals.

- Cities are already facing an explosion of new transportation technologies. In particular, the exponential growth of transportation network companies has had an impact on cities, adding a new choice for riders who can afford it, and adding substantially to city traffic congestion, which slows down buses. It’s no coincidence that cities and agencies on the front line of the onslaught of TNC traffic—like San Francisco, Seattle, and now New York City—have recognized the importance of making comprehensive efforts to make local bus service faster, more frequent and reliable, and more legible for riders. Bus network redesigns (including bus stop balancing), transit-only lanes, transit signal priority, and all-door boarding need to be high on the agenda for transit agencies.

.....

[sundah]

I had hoped to respond to Hatman's 14th May post much before this, but family matters which involved spending time in another city became more urgent. Also, this is my third attempt, presumably because the other two times out. So this time I'll avoid interruptions (lunches etc.) and make a long post into two shorter ones.

I had calculated that the cost of automated tunnels for light vehicles, on a passenger/mile basis, would be half that of a conventional urban freeway. This was based on USDOT, USBEA and APTA data, plus findings of the International Tunnelling Association. But the feasibility of such tunnels was doubted, as their on-going costs are significantly higher than those for at-grade or elevated facilities.

It's true these costs are higher, but not to the extent that would make then unviable. For instance, the World Road Association puts the cost for the operation, maintenance and heavy repairs as between 20 to 29 percent of all costs over a thirty year period. So the automated tunnels would still have a major cost advantage, and still come with minimal opposition from NIMBYs, property owners or residents.

But obviously only some automated highways will be in tunnels, and while they can be bored in most ground conditions, there will be certain cities with an unfavourable geology. Tunnels have most relevance in the cities of European or developing countries, those without existing, extensive, urban freeway systems. Automated highways in the U.S. will mostly come from converting freeways at minimal cost, albeit with considerable transitional difficulties, Tunnelled spurs are likely, eventually, in some cities to provide better vehicular access to CBDs, but it can be expected that priority for automated highway systems be given to better metropolitan, congestion-free, mobility.

[sundah]

Hatman also indicated that the higher capacities achieved by urban rail assured that its tunnelling would remain more justified than that for automated light vehicles. Certainly rail's capacities will remain much higher, as up to about 72,000 passengers/hour is possible, while that of an automated lane would only be about 12,000 (at 1.67 people per vehicle).

Somewhat coincidentally, the capacity of a six lane automated tunnel would also be 72,000, although, of course, the six lane tunnel would be larger. A comparison of tunnel profiles showed that the cross-sectional are or the six lane automated tunnel would be about 2.8 times greater than that for a track of rail. This would be the primary determinant in their comparative cost, however, as the six lane tunnel requires two decks, the difference is more likely to be closer to 3.5.

Even so, it's far more likely the automated tunnels would pay their way. Based on the data from the USDOT, USBEA and APTA, in 2014 motorists on average paid 29.8 cents/passenger/mile and were subsidised 3.5 c/p/m by the three levels of government. In comparison, riders of public transportation paid 25,6 c/p/m and were subsidised 82.5 c/p/m.

Again the above sources show that despite their higher capacity capabilities, largely because of dead-heading at peak hours and the need to maintain services at times when there are few passengers, the average occupancy of buses was only 10.66 passengers, while that for rail carriages being 35.75.

Whats more, the sources showed that freeway lanes were carrying 2.88 times the number of people than a track of rail. So a six lane automated tunnel would have 17 times the number of users to pay for its 3.5 times greater cost? (Admittedly, this is taking this figuring a a bit far, as many factors are involved, but it's an indication.)

Such reasoning has little relevance for cities already having substantial rail systems, like New York's subway. They are durable and will continue to serve, particularly commuters for many decades. Also, there is no certainty as to when automated highway systems will eventuate. If they had the priority of a national defensive system, the would probably be here now, at least in some embryonic form. But they are not and the conversion of freeways will have to wait until a significant majority of vehicles are manufactured or retro-fitted with the necessary intelligence. Given that the average life of a car is about ten years this could take twenty years or more. The Chinese might feel no need to wait so long, and start building automated highways from the outset.

[M II A II R II K]

AUTONOMOUS VEHICLES MIGHT DRIVE CITIES TO FINANCIAL RUIN

https://www.wired.com/story/autonomo...financial-ruin

.....

- The advent of driverless cars will likely mean that municipalities will have to make do with much, much less. Driverless cars, left to their own devices, will be fundamentally predatory: taking a lot, giving little, and shifting burdens to beleaguered local governments. It would be a good idea to slam on the brakes while cities work through their priorities. Otherwise, we risk creating municipalities that are utterly incapable of assisting almost anyone with anything—a series of sprawling relics where American cities used to be.

- Many cities balance their budgets using money brought in by cars: gas taxes, vehicle registration fees, traffic tickets, and billions of dollars in parking revenue. But driverless cars don't need these things: Many will be electric, will never get a ticket, and can circle the block endlessly rather than park. Because these sources account for somewhere between 15 and 50 percent of city transportation revenue in America, as autonomous vehicles become more common, huge deficits are ahead.

- Cities know this: They're beginning to look at fees that could be charged for accessing pickup and dropoff zones, taxes for empty seats, fees for parking fleets of cars, and other creative assessments that might make up the difference. But many states, urged on by auto manufacturers, won't let cities take these steps. Several have already acted to block local policies regulating self-driving cars. --- This loss of city revenue comes at a harrowing time. Thousands of local public entities are already struggling financially following the Great Recession.

- It will take great power and great leadership to head off this grim future. Here's an idea, from France: There, the government charges 3 percent on the total gross salaries of all employees of companies with more than 11 employees, and the proceeds fund a local transport authority. (The tax is levied on the employer not the employee, and in return, employees receive subsidized or free travel on public transport.) --- Vice president of market development at Keolis, said that the Bordeaux transit authority charges a flat fee of about $50 per month for unlimited access to all forms of transit (trams, trains, buses, bikes, ferries, park and ride).

- The hard-boiled US crowd listening to him audibly gasped at that figure. Ridership is way up, the authority has brought many more buses into service, and it is recovering far more of its expenditures than any comparable US entity. --- It's all just money. We have it; we just need to allocate it better. That will mean viewing public transit as a crucial element of well-being in America. And, in the meantime, we need to press Pause on aggressive plans to deploy driverless cars in cities across the United States.

.....

[lrt's friend]

If autonomous cars are always on the move, are we not wasting an enormous amount of energy. Also, are we not also creating additional congestion as these vehicles circle the streets?

The comment about a monthly charge for unlimited use of transit is interesting. I know that our city has already implemented this for college and university students. It is still a big step to move to the general population, but I can understand why there have been positive results.

[jmecklenborg]

Originally Posted by lrt's friend

Uber and Lyft are software companies, while the cost and risk of ownership is with individual car owners. It is an enormous leap to move these companies to own the vehicles as well. The cost of these companies assuming those risks will be borne by the customer.

The pressure will be on to assume control of transportation and remove individual car owners for safety reasons. At that point, convenience pricing especially during peak hours will increase rapidly to generate sufficient profit. A perfect example of this is Highway 407 in Toronto that was privatized and tolls were increased dramatically to maximize profit. The highway ceased to perform its original purpose to reduce congestion on the parallel and public Highway 401.

Great points. Most companies with their own delivery fleets lease the trucks, they don't own them. Hell, the big car rental companies might not even own their own vehicles (I don't know).

The thing is that leasing companies can repair delivery trucks for years and years because nobody really cares about semi trucks, box trucks, and trailers. But people don't want to pay to ride in an old car. They want to be in a newer car.

So if everyone is ridesharing in driverless cars, there will be a 2+ tiered system well beyond Uber X and Uber Black. The poor will be puttering around in cars with blue smoke coming out the back and torn up upholstery.

Also, all of the hoopla over driverless trucks completely forgets that delivery truck drivers often unload and stock the stores and other places that they deliver to. A gas station cashier doesn't go out to the Budweiser truck with a 2-wheeler. So if a truck needs a guy to unload it, there isn't any cost savings to having a computer drive him over instead of him drive the thing himself.

[M II A II R II K]

• Video Link

[jmecklenborg]

More industry propaganda promoting driverless cars as safer than ordinary driving and a solution to traffic congestion.

We're 10+ years into hearing that these cars are going to take over in two years, but it never happens. We still haven't seen one travel by itself from New York to LA. Let's see one travel from Wall St. up to Harlem and back with a passenger who doesn't get frustrated by how cautious the car is.

These things are facing two huge problems -- they're going to be too cautious in city traffic. People are going to get incredibly impatient. Second, riding in a car with no physical controls at highway speeds is going to be terrifying.

Next time you're riding in your car at 70mph, take your hands off the wheel and feet off the pedals and imagine that they no longer physically exist. All the sudden things get really scary. It's like going out for an innocent nature hike and then getting lost in the woods. All of the sudden every sound is scary.

[Hatman]

I figure this is worth posting here:

• Video Link

This is what a Tesla is able to 'see' with its forward-facing camera. It is cool that it is already able to recognize so many objects with just its vision system - no expensive lidar present. Green is all the area it considers to be drivable terrain, and boxes identify other important objects.

It's also worth noting that this is from Tesla's older Version 8 software, and that a newer and even more impressive version is being released now in Version 9:

• Video Link

[mousquet]

^ It seems it only missed a pedestrian over here at 1:15, while obviously, not a single mistake is allowed.
It is quite impressive and exciting otherwise. It sure stomachs a whole load of real-time data.

Their system is apparently growing mature, but beyond software engineering, it probably needs some faster (badly costly) embedded hardware to run it, so it would be perfectly safe.

Guess it's only a matter of years from now.

[SIGSEGV]

Autonomous cars don't just need to contend with normal conditions, they must contend with potential adversarial inputs.

[mousquet]

^ That's no problem in theory. Any kind of situation can be handled so long as designers thought of it.

It should even go way further. These vehicles will be provided with some latest "machine learning" tech in the end.
In other words, machines embedded in there are likely to enhance their outputs by data analysis, on their own, so to speak.
The more data they'll be collecting, the more accurate their responses are likely to be.
They will drive much better and safer than any IndyCar champion.

The possibilities are just about endless, provided that the onboard hardware is fast enough to run this type of things.