I made a little Tweet Storm Saturday morning on an issue that’s been on my mind about driverless cars and the City:

Here’s the link to the very good video produced by the Ryerson City Building Institute in Ontario, Canada: https://www.youtube.com/watch?v=R1B9z8ituS8&feature=youtu.be

There are of course many other issues, not the least of which being this technology’s utility for Transit services. But interestingly as a result of my tweets I was sent this link from the US Highway Admin on the very subject of aviation standards versus road standards. Because, let’s face it, the standards are wildly different: 38,000 people were killed directly by auto-dependency last year in the US, that’s just in crashes, that doesn’t include those dying of respiratory diseases, or from the way driving makes people fat and sad, also leading to earlier death from the diseases of inactivity.

I have an additional thought too. At what point will the near perfect safety performance of driverless cars lead to human driving becoming illegal? I suspect this is an almost inevitable consequence of this technology. Likely to start in certain areas then be extended. Perhaps what Google et al are ultimately doing with Autonomous Vehicles will lead to a redefinition of the conceptual link between cars and freedom in American culture?

Share this

55 comments

  1. That’s exactly my thoughts. People seem to think that autonomous cars will only have to be as good as human drivers, whereas in fact they will need to be perfect. Can you imagine the lawsuits in the US if a Ford driverless car, say, killed someone? I agree that the only possible way of achieving such a perfect result is through the vehicle being ridiculously slow and cautious in the presence of any potentially ambiguous hazards.

  2. We know they have to stop so we will all just pull out in front of them. Moral hazard. So yes the only way it will work is separate areas for drivers and machines.

  3. This is exactly right. And what’s more, everyone who steps out of an autonomous vehicle becomes a pedestrian who won’t tolerate being forcibly separated from vehicles. If I can trust an autonomous vehicle to avoid colliding with me, then I’ll step out into traffic. End results is (hopefully) city and town centres effectively becoming no-go zones for driverless vehicles.

    I have been banging on for some time amongst my profession about the lack of thinking about pedestrians and cyclists in what passes for transport and traffic management planning around autonomous vehicles. Lots of excitement about intersections working like fish schools meeting, but nary a ped or cyclist in the mix. I suspect the outcome will be as Patrick suggests – a two-tier arrangement with motorways and big arterials facilitating autonomous vehicles, and cities and towns with highly pedestrianised spaces. But what to do with the autonomous vehicles when they get off the motorway?

    And what does the crash-avoidance capability mean for buses, potentially a big winner from autonomous technology (most of the cost is in the driver – autonomous buses could mean massively cheaper fares or a big boost in service) – do they get forced out of centres too?

  4. I don’t generally agree with this line of thought. Especially not in NZ. An AV in theory can react instantly, unlike a human which needs at least 2 sec to react. 2 seconds is an eternity. AV’s will also be driving to conditions and under the speed limit. AVs wont get angry, drunk, distracted, tired, old, suicidal. AV’s will be very unlikely to cause accidents because of all those reasons, furthermore they are also far more likely to reduce a crash to a near miss when the other driver is at fault. This alone will help to reduce road deaths. Especially by bringing speeds down everywhere.

    How would you tell the difference if it is an AV or not? I’m not pulling out in front of a car, risking my life on the chance it will stop for me. That is just stupid and with automatic video recorded evidence of the entire event, im sure the police could use it to fine someone straight away. Even in the US lawsuits would be difficult if the other party is clearly at fault based on video evidence.

    I certainly won’t program my AV to take me through shared spaces else I’ll be stuck there forever. But if some ped jumps out and gets hit in this country, they are at fault and can’t sue anyone anyway.

    But this stuff is still a decade or three away before it becomes common place.

    1. Agree with nonsense. Last time i checked my reaction time was approximately 200ms, or 0.2s. Most people would be in that range.

      1. I think what Ari means is that conservatively it can take some people 2s to perceive something, decide what to do, and then take that action. Particularly if it is something novel or unexpected. Basic human factors theory for road design.

  5. AVs will be significantly slower because they won’t take any risk. I remember reading a post here before that explained better than I can. But basically we can drive so fast only because subconsciously we accept the possibility that we might die or kill somebody any time we step in a car. Of course computers can’t do that and rightly so because it’s not their life on the line

  6. Yes I agree that the biggest initial improvement will be the motorways.

    The form factor of vehicles will change to single commuter and minivans, these will replace buses and train Public Transport. The train network will revert to moving cargo. The other major development will be Mobility as a Service (MaaS) which will be an AI controlled transportation network that works out the most efficient way to move people from A to B. The AI Transportation network will redesign our roadways to make them more efficient and safer (AI roadways separated from human roadways and footpaths) for the transportation of people and goods. The AI Transportation network will have access to massive amounts of information to determine new ways in which to move people and will therefore provide planning advice to Transport Officials .

    People will no longer need to own a vehicle and will only pay for the service to be moved from A to B. Less vehicles will be required and auto manufacturers will become transportation service providers providing a range of services from luxury to basic. Less vehicles means that private and public car parking space can be reallocated and more roadway allocated to active (walking, running, cycling) modes of transport which in many cases can be separated from AI roadways.

    Realistically this is all still 25-30 years away as it will take another 10-15 years before all new vehicles have AI mode, sensors added to roadways, a vehicle form factor shift and another 10-15 years for the entire fleet to turn over to the new AI fleet. People will accept over time that an AI driver/network is safer than Human drivers, there will be 20-30 years of data backing this up, legislation in place and the whole fear of accident/liability will be minimised.

    This all seems Asmovian now but the technology is well under development. However we still need to come up with smart ideas over the next 10 years to keep our cities moving. Most of the current road building plans are too little too late and only wastes time/money and creates more congestion.

    1. Or we could keep the trains, and have our driverless cars bring people to the station. Depending on if having one train moving 1000 people will be cheaper than having driverless cars & vans move all these people.

      By the way, “projectmrcrocar”, what does that mean?

      1. Exactly roeland, that is indeed the much more likely outcome, but the techno-fasntascists will never stop predicting the end of rail transit, it seems. Always the next new thing will achieve this.

        It isn’t entirely clear what the motivation is; for some it’s a confusion about idealised political states; trains are somehow communist, and therefore evil, but for others i think it simply reflects their psychological in problems dealing with other people. A deep compulsion for some, especially it seems people with a technological bent, to work out a way to move through the city entirely in their own space, separate from others….

        1. I agree that the form of autos will change to narrow shape. Side by side seating is too wasteful. It’s similar to the reason why people don’t often have side seats on motorcycles or bicycles even though they’re available. The cost of suburban public transit trains and buses is far more expensive and serve far fewer places than a fleet of 100% electric, single-width, roll-cage protected, weather protected, standard door, standard window cars. My deep compulsion is on time preservation. The space of a narrow car is more shared than current cars. With windows down, they’re more like bicycles.

      2. AI (273%) mode plus a form factor change to single commuter (100%) and 2-6 people minivans (200%) will increase the peak capacity of existing roads by a factor 7.5-11. The motorway lane capacity increases from 3,000 car/hr to 22,500-33,000 car/hr.

        If an AI vehicle can pick you up and deliver you to your destination in less time or cost than the train then people will not use the train.

        1. How is swapping from one car to another (the latter driverless) for specific trips (esp. at peak time), going to change anything. Like-for-like?

          And isn’t this option – outsourcing your drive – already available via Uber and Taxis? Difference?

          Finally, presumably on all routes driverless cars will be competing with other driverless cars and manually driven ones, meaning congestion. Meaning a dedicated RTN would be quicker…no?

          Honest questions, BTW.

          1. Of course you’re right KLK, the RTN will always better where ever there is a concentration of destinations like city centres, those bot-cars will still be tripping over each other; traffic congestion, y’all. Toa’s math is an abstraction.

          2. Just reiterating that people will not own cars, instead they will pay for a service to be transported from A to B in either a single commuter or 2-6 person shared minivan AI.

            1. The current form factor is inefficient therefore a change to single commuter cars (100%) which takes half the road space and shared minivans (200%) will increase the road space efficiency when compared to the current single commuters in up to minivan sized vehicles

            2. Humans are inefficient and use far more road space than required. An AI car increases capacity by 273% as per the below ieee study.
            http://spectrum.ieee.org/automaton/robotics/artificial-intelligence/intelligent-cars-could-boost-highway-capacity-by-273

            Therefore motorway lane capacity increases from 3,000 car/hr to 22,500-33,000 car/hr and ends congestion for the next 100-200 years (assuming current population growth and excluding any significant increase of working from home).

          3. Erm, a motorway lane capacity is at best 2,000 vehicles an hour. Even if you can increase that 273% then you still only get 5,460. About half the capacity of a conventional busway lane.

            Here’s a though for you Toa, have you considered what automation technology might do to the capacity of a bus or rail lane?

          4. And I’m just reiterating that what you think is ‘certain’ is anything but. No sign of your skinny cars being taken up anywhere in number despite congestion already being huge. Oh there’s 200 million e-bikes in China right now, and how many motor bikes and scooters too…. Your certainties aren’t convincing.

          5. https://en.wikipedia.org/wiki/Lane
            ….the range can be as low as 1000 passenger cars / hour to as high as 4800 passenger cars /hour but mostly falls between 1500 to 2400 passenger cars / hour

            The factor of 7.5-11 is correct depending upon number of single commuter and shared minivans AI.

            As mentioned above Auto-manufacturers will become transport service providers (many are already investing in carshare) so will provide almost door to door service for less cost and time than trains or buses. Many people just will not use trains or buses anymore.

            If your destination is the CBD then only minivans AI will be allowed in/out during peak times.

          6. Driverless cars going door to door will never be cheaper than driverless buses. That’s physically impossible. Just remember that any technology that makes cars better makes buses even better, even sooner.

          7. An AI bus will never go door to door as it is too operationally expensive, inefficient and in some cases not physically possible. Large fleets of AI minivans and AI single commuter vehicles will replace buses.

          8. Toa – you say that “people will not own cars” – but unless you are proposing a law and banning the ownership of cars, people will still want to own, drive, and use their own cars. Cars are not just means of transport, they are personalized pieces of space. You can hang fluffy dice in your window, supercharge your engine, paint your car any colour you like, and eat food however you want in your pajamas. None of that is gonna happen in a auto bot you hire by the minute.

          9. I could definitely see a scenario where there is less late night PT than there is currently, however the core rapid transit lines are likely to be as important as ever.

          10. “If your destination is the CBD then only minivans AI will be allowed in/out during peak times”

            So you are just duplicating what current and future rail lines and busways can already do? And do it much more efficiently?

            And you’ll be doing it competing with all other driverless vans in limited space?

            On a related matter, your posts on this over the last year seem to have gone from proposing driverless versions of cars in their typical form, to driverless cars of a reduced width, to now driverless vans carrying 2-6 people (the latter just being a small bus, so again, replication?).

            If the new bus network has stops in easy walking distance of our neighborhoods, those buses delivering people to train stations which take us directly to a destination like the CBD, then it would make these vans redundant. And that’s before we talk of people having easy access to rail stations.

            If we don’t want to or can’t take a bus/train, OK. But then we can already outsource the driving to a an Uber or Taxi driver. Its them that would fear driverless, I guess.

          11. Just need to read through the earlier posts for the background. The throughput/capacity of motorways will increase by a factor of 7.5-11(3,000 car/hr to 22,500-33,000 car/hr) ending congestion for the next 150-200 years. AI single commuter cars and minivans will replace buses and trains and be the new future Public Transport.

            The key of this is the massive increase in throughput/capacity that AI and new form factors create ie 22,500-33,000 cars/hr (ie car/hr=passengers/hr) per motorway lane.

            By comparison a single existing Auckland train line moves 8 trains per hour with a full capacity (seated/standing) 373 x 8 = 2,984 passengers/hour per train line.

          12. From 8 to 24 trains per hour?

            That’s one train every 150 seconds on the Southern line, allowing for signalling, freight trains, multiple line merging/diverging, stopping at stations and the movement of up to 750 passengers?

            VCs SkyTrain which is a fully automated dedicated RTN cannot do 150 seconds?

          13. That’s the upper limit in the CRL with upgraded signalling. So that is indeed the capacity. It won’t happen till it’s needed, the plan on opening is for 18 tph at the peaks (a train every 3:20 mins) with 21 tph through Otahuhu hub (sub 3 mins). No doubt this requires the completion of the Third Main for freight [in black below]. Given that full 6-car trains are already being experienced at the peaks, the demand by 2022 will require this kind of pattern at the least, the latent demand for decent Rapid Transit is very high in AKL we are discovering:

          14. Ok I see that post CRL Britomart will have 18 trains per hour in both directions.

            However I checked the current timetable for Britomart for all four lines between 8-9am weekdays. I counted 21 trains arriving and 20 leaving per hour and is 41 trains per hour in both directions.

            This is a halving of throughput to the CBD?

          15. Toa, no it’s a doubling; 18 trains per hour each way; rising to 24 each way over time. It’d be pretty daft to spend $2b to halve capacity.

            30-48 trains every hour with up to 750-1000 (when jam-packed) people delivered to and through the city on each one, is a human resource that no number of skinny or bot cars can match. Especially at anything like the spatial cost, which, of course for underground rail, is zero. It will be this resource, plus those arriving by ferry, bike, and foot, that will enable the city to thrive and the streets to function at all for buses, deliveries, emergency vehicles, oh and a few cars….

            You may your bot car to the suburban station; but you’ll arrive in the dynamic city on Rapid Transit.

          16. Ok so from 41tph as in the Brittomart terminus, down to 36tph when the CRL opens in 2021 and then increasing to 48tph by 2041. The peak 2041 hourly passenger throughput of the 3 CBD stations will be 48 x 1000 = 48,000 passengers/hr using all 4 feeder lines in both directions. Unsure whether 700-1000 people will be able to board/exit every 150 seconds in 2041 as the VC SkyTrain RTN is about half the number of people 580 (Wiki) in 162 seconds.

            A single motorway lane with AI single commuter and shared passenger minivans will move 22,500-33,000 passengers/hr in one direction. 6 lanes (3 in both directions) = 135,000 – 198,000 passengers/hr which is the end of motorway congestion for the next 150-200 years.

            If in 2041 an AI vehicle can pick you up from home and deliver you to your place of work (or close to it in the CBD) in comfort, at low cost and without congestion then why would you use the train?

            Have you read the below 2012 IEEE article as that explains AI vehicle capacity increases.
            http://spectrum.ieee.org/automaton/robotics/artificial-intelligence/intelligent-cars-could-boost-highway-capacity-by-273
            My calculations include the road space efficiencies of a small AI single commuter vehicle and the adoption of carshared AI Minivans via Mobility as a Service (MaaS) operators.

          17. NO! Currently up to 20 trains into Britomart and those same 20 out again. Post CRL there will be up to 48 trains in and 48 trains out. This is the power of through-routing the dead-end.

            Additionally these are real already existing and currently and increasingly well used vehicles, not notional fantasy ones dependent on a whole lot of ifs and buts, and magic math.

            In 2022/3 when the CRL opens there will be 36 tph operating through the CRL. It is problematic that it is going to take this long because the need is already there, we have been so held up by ill informed thinking at many levels of transport provision. And daydreaming about AVs while ignoring spatial inefficiencies for road vehicles is simply the latest version of this.

          18. The 48tph seems very theoretical especially since these will be shared lines with multiple merging/diverging rail lines in close proximity. Also there will be up to 1000 people boarding/exiting in 150 seconds at closely spaced CBD stations. So some scepticism over the CRL performing better than the VC SkyTrain which is a dedicated high speed RTN.

            I note that some of the tph boxes indicate peak direction only. Therefore I suspect the other boxes are total tph with 18tph total (not each direction) at the three CBD stations making a lot more sense as it will provide adequate headway/block signalling for the various merging/diverging rail lines. Does the 2021-2041 Post CRL map have supporting documentation as I really want to know how those throughput calculations were derived, the headway and block signalling assumptions used?

    2. “The train network will revert to moving cargo”

      Why? Trains in Europe, Japan, China are already moving people around their respective countries, many hundreds at a time, at speeds well in excess of 300kmh. In another decade that will be closer to 400kmh.

      There is no remote possibility that cars or buses, autonomous or otherwise, will have that capability within our lifetimes, if ever.

      It should also be noted that this rail technology has a long term safety record that road traffic is currently incapable of matching.

  7. I recall reading an article on some Congressional hearing in the US of A on this very matter. The congressman questioning the manufacturer (Chris Urmson) was very concerned that a driverless vehicle had almost(!) killed someone.
    My immediate thought was: WTF!! Cars with drivers kill some 38,000 people in his country every year and here he was concerned about one – ONE, almost killing.

  8. Accept the first premise (2-5) but not the second (6-7).

    Fully autonomous vehicles will soon be safe enough to put into chaotic unprotected environments. If not by 2020, by 2025.

    Until that time they will be highly restricted in those environments, or kept from them completely.

    1. Well I think you miss the point of 6 and 7. Not that they won’t work just fine, in fact i agree they will have to be super safe; and in doing so they will function, but just won’t get anywhere much. As they will have to be extremely conservative and in ‘chaotic unprotected environments’ that will mean crawling along at best, and grinding to halt most of the time. It’s not that the tech won’t be good; it’ll be too good to get anywhere.

      1. Totally agree. The reason most people don’t cross mid-block on a busy road is because they fear getting hit. Imagine knowing with 100% certainty that the 1300kg machine hurtling along at 50Kph would stop. There’s no way people could be made to queue at ped-crossings and wait for the ‘green man’.

  9. I’d also considered the possibility of this ‘transit hacking’, either being able to confidently walk onto a road or groups of people herding driverless cars just for the fun of it. The pessimistic side of me thinks that rather than driverless cars ushering in a pedestrian and cyclist utopia, that any behaviour that interferes with the smooth operation of the roads will be discouraged through physical barriers. or criminalised.

    1. Yes this will likely happen, especially in the US, but surely that will be impractical on every city street. Me and a friend will be able to render, say, High St, totally vehicle free, simply by playing silly buggers at the entrance… come on; this is gonna happen, especially in a general context of more and more bot-things bossing us about…. or perhaps you’re totally right and this is how a police-state of machines happens; every bot-car photographing, hunting down, and prosecuting humans for simple trivialities like stepping off a footpath… perhaps supported by police-drones hovering above every street corner…..oh happy days.

      Ok that’s my pitch for a dystopian mini-series; any producers out there….?

      1. Well the other possibly dystopia outcome is that due to all those cars, the city will be so harsh and thoroughly unpleasant for humans, that travelling outside a car becomes unthinkable.

        Oh wait a minute …

      2. Eventually after a few years of the police state, the people start a revolution, banning the autonomous vehicles from the city centres. Public transport is allowed to remain on specified routes. The empty streets are turned into broad pedestrian malls with cycle ways throughout, allowing people to move freely on foot. At intersections, pedestrians have priority,but it’s not much of an issue as the autonomous public transport vehicles are all timed to minimise the disruption at intersections,generally resulting in only a few seconds delay for pedestrians and cyclists. For the businesses, delivery vehicles and bots automatically replenish supplies overnight when there are only a few people around….

    2. This already happens in the weekends with intoxicated people slowing down traffic in the CBD. Most people will stop, drive slowly past, call the police or have a few stern words.

  10. “38,000 people were killed directly by auto-dependency”

    I challenge you to produce a single crash-investigation report confirming that was the cause of at least 1 death.

    You broke the BS detector, the needle flew right off the scale….

  11. I’ll drive my own car until I’m either mentally or physically incapable.

    Some may suggest that’s already occured ahem…

  12. It occurred to me recently that the safety benefits of autonomous cars are nearly all achievable without the computer actually replacing the human driver. Audible lane departure warnings require only a camera. Add GPS and a basic map and it’s only a small step to a car that will not allow its human driver to exceed the local speed limit, or take a bend too fast, or cross the centre line (unless there is a vehicle in front to overtake, and enough space to complete the overtaking manoeuvre safely), or leave the road at high speed, or fail to give way at an intersection, or just “not see” a motorcyclist or pedestrian, or make any of the other simple mistakes that cause the vast majority of road deaths and injuries. So perhaps that’s the way it’ll play out: long before the legal and regulatory issues raised by computer-driven cars are sorted out, the technology will have largely solved the safety problem while still keeping a human at the wheel.

Leave a Reply