Saturday, 12 March 2016

Monotonous Autonomous

The media is buzzing with the prospect of "driverless" cars being let loose on UK roads after it has been reported that the Chancellor will making an announcement in next week's budget.

Personally, I am bored and worried about the whole thing in equal measures. Bored, because I think we have far more serious things to worry about with transport in the UK and worried, because the whole idea is being pushed by the big interests which continue to stifle change, especially in urban areas.

Proponents of the technology talk about how it will make roads safer for all users. Don't get me wrong, vehicle technology has done wonders for occupant safety over the years and we now have things like assisted braking, collision detection and anti-skid detection. Heck, some cars even park themselves nowadays. I have no doubt that in an autonomous-utopian future cars will be safer for the people inside and outside the vehicles because on average and over time, computers will perform better than humans. Even with our cleverest cars now, the humans in control still mange to screw things up.

The real reason for the official push and the massive investment by companies is that of the need to make and sell stuff as without people spending money, the current economic model would collapse. They are also a way of cutting costs and improving the bottom line. For example, there will be tests of autonomous HGVs on the M6 which will see platoons of lorries running closely spaced, with the lead truck under human control. The following lorries will have drivers (for safety as it is a trial), but the idea will essentially squeeze lorries into a smaller length of road and in theory, use the road space for efficiently. The other benefit is being closer together, the following lorries will slipstream each other and so reduce fuel consumption. 

The end game with this of course would be lorries not needing drivers at all and thus cutting out costly humans. A haulage depot could send its fleet out in the evening and customers could arrive with their goods waiting to be offloaded - hell, in a modern warehouse set up, lorries could be self-docking and robots could unload them which means even less staff are needed. Industrial robots have been here for decades, so it's not pie in the sky. Don't forget beyond road transport, aircraft and container ships have been piloted by computers for a long time!

Beyond the big trucks, there will always be a need for local deliveries and it is no coincidence that research into autonomous vehicles and drones takes place to cover the last mile. There is a real possibility that one could order something on a website and the widget turns up having never been touched by human hands!

For personal transport, the future is varied and by this I mean split between those who can afford cars and those who can't afford them or don't want them. For those that can, one will be able to jump in the car and enjoy a chauffeured journey to work with time to read, catch up with emails or nap (more than we see now at any rate!); those without will have booked a car with an app - taxi drivers will be a thing of the past as observed by the People's Cycling Front of South Gloucestershire.

For motorways and trunk roads, again, there will be similar benefits as with lorries, one can join a platoon of cars and just relax as one is whisked along. Imagine not having the fatigue of driving long distances! Actually, who even needs to learn to drive any more - yay cars for everyone (who can afford them). 

Of course, there are flaws in the plan and that is the ones about road capacity and people not in cars. On motorways and trunk roads, things are already pretty safe because the road layouts are enshrined in standards which have been made mandatory. There are no people walking and cycling (banned on motorways and not provided for on trunk roads) and so control over vehicles will be relatively simple. The vehicles would talk to the road control computers and so speeds would be adjusted in real time with the system anticipating hold-ups ahead; in short, the automation would actually increase capacity and improve safety. 

Of course we know that creating capacity gives more opportunities for induced demand, so the long-term flaw must be that we will end up with a strategic road system which will end up just as full as it was before.

The thing which bothers me most is what happens away from the carefully controlled conditions of the strategic road system. Cities, rural lanes, residential streets. We are told that autonomous vehicles will be safer around people walking and cycling than humans. That might be true, but it depends on the model the cars are running. For example, if autonomous vehicles are programmed to pass people cycling with a 2 metre gap, then overtaking opportunities will be hampered in town (and perhaps that would be a very good thing) but the people in the car will still get frustrated and may take control back (I'm assuming there would be an override of course) and so the safety argument goes. 

What about people crossing the road? With traffic signals, the vehicles can talk to the signal controller and know when the lights will change and slow down in good time. What about zebra crossings or people just crossing - will the sensors ever be good enough to cope? There have been thought experiments about autonomous vehicles sacrificing the occupants to save other people or the vehicles being able to run over one person to avoid a crowd. In reality, the forthcoming test vehicles will be very cautious because the "powers that be" wouldn't want carnage to be the output of the experiments - although there have been issues!

Perhaps the future will be more complicated. Like cruise control, one gets on the motorway and then the car takes over, but touch the pedals or the wheel and the car cedes control to the driver. In town, I can't see the technology taking over because of the complicated way the urban environment works. If it is pushed, then expect to see people walking and cycling having a bit of fun. 

Unless vehicles go 100% autonomous and with no way for people to take control, then the idea is already a busted flush. The promised mobility for non-drivers won't be possible, so apart from out-of-town logistics and clever motorway cruise control, the technology won't improve our towns and cities. We will still have the problems of congestion and car storage and until we tackle that, we are heading up the business as usual dead end.


  1. Clark in Vancouver12 March 2016 at 18:12

    I predict that there will be an even louder call from people to protect their neighbourhoods from automobiles if they are out there driving around.

    I find it funny that they intend to run long strings of freight trucks in a row down a highway. Why don't they just use a train?

    Apparently someone with a drivers license would still be required to be in one. This means that currently disenfranchised people who cannot get a drivers license will still need alternatives. Autonomous vehicles will not be useful for them.

    This whole thing is just a way to sell cars to an oversaturated market.

    1. Don't get me wrong, cars have a use and a place, but they are too numerous in cities. Freight should be on rail, but platoons of lorries to out of town depots with E-vans and cargobikes for the "last mile" might be a helpful model.

    2. Trains are actually far easier to automate than cars are. Vancouver already has the skytrain, running for decades with the only deaths due to suicides and people stepping off the platforms. Of course freight needs more people to ensure that things are loaded safely, and other related things like that, but even computers can do most of that work. Humans provide the imagination, the computers do the algorithms.

      Freight lines in the UK do need to be improved. More electrification, double tracking in some places, railway crossings could be fully protected with full road barriers and more grade separation, automation, faster speeds, even for freight trains, perhaps something like 160 km/h as opposed to perhaps 97 on the motorway for trucks, the latter of which gets in congestion. Passenger trains of course could go far faster too even with automation (heck even planes go almost the speed of sound with automatic planes, given that they pretty much just need to go in a straight line at a given altitude, trains are like that, being on tracks), in the realm of 250-350 km/h. It would be far more efficient for the UK. Vans and LGVs are good ways to do the last mile especially if automatic LGVs can do that. I even heard something called the cargo hopper, used in the 30 km/h zone in the historic commercial and residential city centre of Utrecht, as a way to make deliveries easier in the narrow streets and be less of an obstruction, and even at just 30 km/h, a danger, than vans.

      If these things were electric models, then they would be even more efficient, 80% efficiency from an electric motor than perhaps 20% for a petrol engine, don't know what it is for a diesel engine, and that we can generate that power cleanly and renewably if we wanted to.

    3. Oh yes, there are plenty of technologies out there and I think different things will fit different places.

  2. one thing that concerns me is what happens when the driver of an automated vehicle leaves the motorway, especially at the end of a long journey. A lot already have issues adjusting just to a lower speed. After a couple of hours of facebooking or answering work emails, I can see a percentage simply forgetting they need to drive again.

    It is just going to be a trial, so I reckon we won't see significant numbers until after 2020. company car and commercial fleets turn over fairly rapidly, so probably at least 5 to 6 years away.

    But you're right, M-ways and trunk DCs are not where the issues are. UK needs to urgently tackle the urban environment and also the twisty country roads, both from a safety, but a pollution perspective. We simply need fewer cars being driven less.

    1. It is no coincidence that the £15bn for road building is being spent on the strategic network, along with managed motorways and so on. This is about building the infra for the autonomous vehicles.

  3. It is important to note that this cannot change the reason for building cycle paths due to subjective safety. Even if you know that an autocar is not going to crush you due to a very carefully controlled computer system, you still don't like having metal at perhaps 113 km/h, probably a higher speed limit due to such automation (perhaps 127 km/h, or if we're feeling really adventurous maybe 150 km/h or more).

    In the world of UK highway control, the DfT and the counties still need to be building separate pathways on everything except the really low volume 60 km/h roads, and also probably except for most dual carriageways due to how unpleasant it would be to cycle even next to them, let alone on them, due to the noise and speed of vehicles. As a byproduct the dual carriageways could often easily be rebuilt with hard shoulders and turned into full motorways or rebuilt as autowegen like the Dutch do it if it is a particularly high quality roadway with a design speed of 100 km/h, at least 1 lane per direction, at least something of a hard shoulder and limited, preferably no, at grade access, and at least a cable or guardrail divide between the two directions.

    1. It's a good point. Most drivers don't crash most of the time, but having vehicles passing close is never going to feel safe.

    2. Also while it would be more efficient, it still would need petrol or diesel fuel, unless it was an electric vehicle, and whatever feeds the power is renewable and clean, like wind or solar (I don't count nuclear, I think that despite the fact that a small amount of mass can be converted into an incredibly large amount of energy, the risks are too great, the UK does have a serious problem with flooding, and continues to ignore the Dutch solutions to this).

      Also, sitting in a car even when you don't directly control it (of course you tell it where to go and any preferable routes, like scenic or fast) is about as healthy as sitting in a car that is under your control, except that you don't crash nearly as often. It is a slow form of cardio problems and other elements of a sedentary lifestyles.

  4. I do find it interesting to see this negative perspective of self-driving cars from a rational point of view.

    My social bubble is mostly young tech-y people, most of whom don't have a driver's licence and don't want one, and we've been keeping one eye on the progress with a thought of "Finally, take the angry/impatient/fallible element out!" And of course, if a car is fully autonomous, no-one needs to own one; just have a few car-club vehicles stored in each neighbourhood, and they can come when you call; less parking taking up valuable road space.

    Most of the opposition to this view is from hysterical newspaper columnists crying either (a) robots are taking over the world, panic! or (b) the technology is not good enough, and can't improve. The first seems ridiculous and the second seems defeatist. It won't be fixed in 2020, but the autonomous-utopian future may be here in a hundred years.

    Your blog, however, appears to be a continuation of the debate, do we need to change behaviour or the environment? Since changing human nature is impossible (no matter how much advertising and training we throw at ourselves), the obvious solution is to remove the human! And I think self-driving cars really will, eventually, improve things behaviour-wise.

    So the debate is not "Self-driving cars: good or bad?" It is: Are they good enough? Not unless the utopia comes with fixes for air pollution, transport poverty, noise pollution, obesity - and congestion.

    1. As I have stated, I can understand the attraction for the carefully controlled conditions of motorways, but hit a country lane or a busy city centre and yes, it is going to take a long time for the software to catch up. Perhaps there is a hybrid future; autonomy to edge of town and then active travel or transit in town.

    2. Are you sufficiently expert in AV technology to make such a judgment?

    3. No, I'm not. But as a professional engineer I keep up with developments and with AVs, the talk is always of a "few year's from now". I think we have a long way to go yet for the places off the strategic road network.

  5. "And of course, if a car is fully autonomous, no-one needs to own one; just have a few car-club vehicles stored in each neighbourhood, and they can come when you call; less parking taking up valuable road space".

    I'm afraid I never understand this argument. By that logic I don't need to own a car right now as I'm sat at home writing this response. Nor do I need to own one for the eight hours a day when I'm asleep or for those hours I'm sat at my desk at work. Just like many people. And yet I do own a car. Just like many people. Completely illogical, but that's humans for you! I just see autonomous vehicles as glorified taxis - and I don't rely on them for my transport needs. How is whether a car is or is not autonomous going to change this? (I guarantee they'll be a lot slower than the average private hire as well ;-) ).

    Andy R.

    1. Car clubs already exist but they certainly haven't resulted in people abandoning their own cars in droves. The car club companies seem to be doing OK financially which means there are enough people who want to use their service and some people will have given up private cars but not enough to make much difference to the street scape.

      As you say, I think many people aren't rational about their decision to own a car and it is more of a habit (or an addiction...) And once people have a car, then it is "I drive because I have a car" rather than a rational analysis of the best mode of transport for a particular journey. The economics of driving means there isn't an obvious cost to each journey given you'd need to do sums on depreciation and spread costs for service, VED and fuel across many journeys.

      One difference between an automated car club compared to current car clubs is there would be greater possible supply of vehicles for a booking because at the moment with conventional car clubs, you are limited to car club locations within an easy journey from home.

      One thing that could favour autonomous cars would be greater charges for conventional cars, and I'm thinking of both parking and road pricing. Even in congested areas like London, residents parking is ridiculously cheap for what is in effect, storage of private property on public land. And road pricing, where every journey has a direct cost, may also challenge the driving habit.

    2. From what I know, clubs are only viable in dense city areas!

  6. Not disputing that car clubs and taxis haven't displaced ownership yet!
    But if your taxi is very cheap, because there's no driver's wages to pay, or your car club car can be outside your front door, no walking required, no chance of other locals having booked the only one in walking distance - that changes the convenience calculation.

    I'm reporting the point of view of young urban people for whom a car is seen as a hassle, not a status symbol. If we can get enough of the benefits without owning one, that's a win!

    And separately, a friend pointed out that, on the fear of legal changes with make the urban realm worse for pedestrians and cyclists - tech companies aren't car companies. By which they meant, tech companies are competing on their tech being good. If they start campaigning to make the problem easier to solve, they will be mercilessly taken apart in public by rivals for not being capable enough. Their entire marketing success depends on proving themselves to be the most competent. Google etc have no interest in making the problem easy enough that Volkswagen etc can solve it!

    1. There is huge attraction for the convenience of the car without the hassle (and cost) of ownership. I think the tech companies are the ones to watch as they will put the traditional car makers out of business I reckon!

  7. One thing that seems to be evident in many discussions about 'driverless' or 'autonomous' vehicles is exactly what is being discussed, how it will look and act, and what aspects of it are end goals versus interim solutions.

    I read an article once comparing self-flying planes with self-lifting (?) lifts. On the surface, the two are incredibly similar in that, under normal operation, a human is not in direct control. However on a plane, there are people required to be ready to assume full control at any moment, and can do so completely; in a lift, although a manual mode is normally provided, there is no more control afforded than a big, red stop button.

    I currently prefer the lift-style option, as it seems the bigger win in terms of potential safety, especially given the level of 'professionalism' of the average punter behind the wheel. And as another commenter pointed out, it's a tougher problem, one that the big boys working on have no self-interest in simplifying. Indeed, it looks very much like where google is headed with their 'friendly' looking, 100% autonomous vehicle, despite their driver-overrideable test Lexuses (Lexii?).

    However, until it becomes clear through legislation or overwhelming public discourse exactly which type we are talking about, I think it's important to consider both possibilities, and include mention of the other to avoid confusion (and potentially talking at cross purposes).

    Essentially, I agree with the points you make in your article, but only if the driver can take the wheel. The landscape changes drastically if the massive metal spike is replaced with a massive red button labelled 'do not push'.

    1. Really good points. I suppose the other thing is that lifts (and railways), they operate in a closed environment and so the fail safe is to just stop. This is of course harder to apply in environments with external variables as you state.

  8. I think that one crucial question missing from the discourse is that of agency. The one thing that cars do provide very reliably indeed is a sense of multiplied agency. The marketers know this well - 100% of car adverts emphasise agency over all else.
    So today’s cars flatter the driver by enhancing his or her sense of agency. People say they love their cars (and when they do, they are merely over-reciprocating the flattery their cars lavish on them). This is what makes motoring addictive. Agency, flattery, love. Oh, and sunk costs.
    So, autonomous vehicles will rob the motorist of agency. If we think about them like that, then they can’t really be regarded as cars the way we understand them today. People will not love them, for they will not flatter their users the way cars today do. The market will not, cannot demonstrate mass demand across all segments. Why would you want to own or use an intelligent machine which doesn’t love you, when you can own a dumb machine which does? [Maybe, if driverless cars can be programmed with a slavishly servile personality which unconditionally loves the user, like a puppy - imprinted like a duckling, but which avoids the uncanny valley, my argument fails. But then, if the vehicle is autonomous, and has a personality, but is owned, then it is not just slavish, it is an *actual* slave. Freedom for the Autonomous Vehicles! I’m Cartacus!]
    Urbanites will take to the Greenwich-model JonnyCabs well enough. It’ll be like Uber but without having to interact with a working-class person, so they’ll love it. But other business models outside urban hotspots for these vehicles are hard to see working anytime soon. What we are far more likely to see is the iterative roll-out of guardian-angel technology. And, at some point, these vehicles will be all-but fully autonomous. But, onrushing singularity aside, that’s several iterations away - a generation yet? Probably. Driverless cars have, like fusion power, always been a generation away.