Ask Hackaday: Why Do Self Driving Cars Keep Causing Traffic Jams?

Despite what some people might tell you, self-driving cars aren’t really on the market yet. Instead, there’s a small handful of startups and big tech companies that are rapidly developing prototypes of this technology. These vehicles are furiously testing in various cities around the world.

In fact, depending on where you live, you might have noticed them out and about. Not least because many of them keep causing traffic jams, much to the frustration of their fellow road users. Let’s dive in and look at what’s going wrong.

Ground Zero: San Francisco

Waymo has been operating a fleet of fully-driverless vehicles in cities including Phoenix and San Francisco. Credit: Waymo

Two of the biggest companies in the self-driving space are Cruise and Waymo, tied to General Motors and Google’s head company Alphabet, respectively. Both operate prototype fleets in San Francisco, with the city offering a perfect training ground for autonomous vehicles. The busy metropolis is no cakewalk, with plenty of tight streets, busy traffic, and various types of public transport to contend with.

It’s no coincidence, then, that San Francisco has become ground zero for stories of self-driving calamities out on the roads. December saw a Cruise vehicle stop at a red light, later refusing to move for a full thirteen minutes. January saw a Waymo car block an intersection at rush hour, much to the ire of other road users. Others have blocked buses, and gaggles of cars have lost their way in deep fog.

Emergency responders have come into conflict with the vehicles, too. In concerning scenes, a driverless Cruise vehicle attempted to drive through a zone where firefighters were battling a blaze downtown. The vehicle nearly ran over fire hoses, only coming to a stop after one firefighter smashed the car’s windscreen. It’s by no means an isolated incident either. In a more recent incident, public safety officials resorted to lighting flares and shouting before eventually getting an autonomous vehicle to halt on an emergency scene. Meanwhile, a Waymo vehicle got caught up near a parade, eventually responding to hand signals from a police officer to get out of the way.

Error: Road Not Found

In some of these cases, the cause of the problems is obvious. For example, in heavy fog, a driverless car will obviously lose visibility of the road around it, and it may decide driving further is too dangerous. Alternatively, rain, dirt, or snow may cover a sensor, completely blinding the car. However, unlike a human, it may not have the skills required to safely find a place to stop. To avoid doing harm, the cars may simply come to a halt where they stand, unfortunately blocking roads in the process.

Cruise has spoken openly about the challenges of running self-driving systems in inclement weather. The company already implements air-blow systems and wipers to keep cameras and sensors clean even in wet and dirty conditionsCredit: Cruise

Other issues can be more technical in nature. A cellular connectivity issue may frustrate a driverless car’s navigation, and strip it of up-to-the-minute data on traffic conditions and road closures. There’s also the potential for groupthink causing problems. Such a glitch might send a bunch of driverless cars the wrong way down a street that’s temporarily one-way, for example. It would only take a couple driverless cars to follow each other and the situation would quickly become unrecoverable without human rescue. Humans are readily able to adapt and find their way out of confusing situations. They’ll readily find a way around roadworks, temporary blockages, and vehicles stranded in the roadway.

Driverless cars, on the other hand, are still learning to drive under normal conditions. They’ve barely got that down yet, let alone dealing with the odd random situations that occur all the time out on public roads. If in doubt, they simply cry for help and come to a stop. Untangling these messes can take quite some time for the humans watching on the back end, which simply isn’t good enough.

Indeed, it’s this behaviour that is at the root of many of these incidents. In the event that the world is too complex to understand, or a sensor fails, or something bad happens, driverless cars typically decide to just stop. While this behavior minimises the risk of a dangerous collision or incident, it nonetheless completely frustrates every other user of the road network. In some cases, the cars have been able to get going again after some period of time. In others, teams of support staff have had to deploy to the area to recover the vehicles. In the latter case, it’s perhaps useful that many of the prototype driverless cars retain regular steering wheels and pedals, allowing humans to drive them away if all else fails.

The Ask

We’d love to know your insight on this problem. How can driverless cars become rugged individualists that can handle the rough and tumble of everyday city driving? Failing that, is there a better way they could respond to problems that doesn’t involve simply sitting on the brakes in the middle of a busy street? Sound off with your best ideas, such that the world’s finest self-driving developers can crib from the comment section and get this problem sorted toot-sweet!

133 thoughts on “Ask Hackaday: Why Do Self Driving Cars Keep Causing Traffic Jams?

  1. It’s a perfect storm of engineering, or rather the lack of cohesion of their convergence into a simpatico of various subdisciplines aimed at robust implementation. Literally, “the devil’s in the details”!

  2. Invest the same money in building trains. Intercity and local. Move the bulk of people by train, leaving fewer people on the roads, therefore reducing road traffic and road accidents.

    1. The economics gets in the way.

      The more stops you add, the slower it gets because trains can’t accelerate or stop quickly. Then, if you want it to run at all hours of the day, it will run mostly empty, so the routes and times are limited such that you get people congestion on the train to increase the average ridership per car.

      Plus as with all public transportation, when it’s not full with daily commuters, the local mental ward outpatients and addicts use them to ride around town. That’s the main reason why I stopped riding the bus.

      1. maybe you should visit the Netherlands and take a random train. it just works. OK there are sometimes accidents and delays, but not on the same route on the same time every day like with cars. we have a unified card to travel by train, bus, tram or metro throughout the whole country. check in, hop on a train and check out when changing transport type or at your destination. that simple. and of course, a nice route app and website helps a lot. we have 9292.nl to help you from any place in the Netherlands to any other place, from door to door.

        1. Our city has been looking at rebuilding the local train loop, but it would take tearing up so many neighborhoods that it can’t be done anymore. Space is at premium.

        2. The USA is 237 times bigger than the Netherlands. We currently have 140,000 miles rail. Compared to 2,000 miles of rail in the Netherlands, I should hope your rail system works well. My home state of Virginia has about 6,000 miles of railroad tracks and there is only about 8 major cities you can go to in the state. The blue ridge mountains prevents rails going west so you have to go north into Maryland or South to North Carolina to get to the western half of the state. Not an Apples to Apples comparison.

          1. Europe is similar in size and population to the US. It has an excellent and extensive rail network for local and long distance transport. What’s your point?

          2. “Europe” is concentrated around the Germany-France-Italy axis in population density with cities and municipalities closely spaced. The rest is more sparsely populated. The US has large population centers dotted all over the place and spread apart more, with a big “flyover country” in the middle” where nobody lives.

          3. Even your “more sparsely populated” parts of Europe tend have to pretty good rail networks compared to the USA…
            And in many ways the “flyover country” is an easy problem for the rail network to optimise as you don’t have to build in nearly as much cross linking or stations as there is nobody out there needing the rail service. You just need to build a reasonably fast rail link and it will be very competitive with flying, if not better in time taken and cost.

            If you have good rail links in the target destinations as well that is really great at making rail travel convenient, but also helps the airports for your “flyover” – with Heathrow being a pretty good example direct lines from the airport to the railway hubs that go everywhere else remotely nearby (as you wouldn’t fly into Heathrow over a closer airport to your goal). But if you can take a 200+mph, quieter, comfortable train instead of the aircraft and probably end up closer to the target (airports tend to be pushed way way out of the cities they serve as they are noisy etc) the only good reason not to is aviation fuel is too heavily subsidised….

          4. Andrew: ‘Europe is similar in size and population to the US’.

            ^^This is what Europeans actually believe. They should get out more.

            750 / 4 is not ‘similar’ to 330 / 4 (millions cancelled, using freedom units for area). It’s the same order of magnitude.

            Throw out Alaska and the result still isn’t ‘similar’. Closer, but justify leaving Scandinavia in when you remove Alaska.
            To be ‘fair’ you need to include the populated parts of Canada, they are just a state.

          5. Also, there is a massive discrepancy in population centers. Our population is north-eastern seaboard, the gulf coast, and the west coast. Between them are vast tracts of nothing. Who wants to run the infrastructure out there when the cost will only be on the heads of the rail companies. Diesel-electric is not considered high speed, and the cost of a ticket on air in the US is cheap enough to utilize it instead. 52 bucks for me to fly from Florida to Newark. It’s not clean, but its also not even 3 hours in the sky.

        3. Just a quick reminder:

          Area of the Netherlands: 41.5 k (km^2)
          Area of Texas: 695.5 k (km^2)

          41 American states are bigger that them. They might be able to take Maryland in a fight.

          The single American city of BostiYorkAdephia is much bigger.

          If I advised the Netherlands to run their ‘semi autonomous area’ the same way as the USA, I’d be called an idiot with no clue and they’d be right. Guess what?

      2. Electric trains can accelerate quickly. How about instead of having a train make every stop we have train cars which travel express from a single stop to the central terminal?

        1. Small ones can. There used to be such track buses around here too, but they were uneconomical and the tracks were torn down for housing development decades ago. Roads can handle the same duty, and you need roads anyways, so the tracks were superfluous.

          1. Ironically, you can put a two-way street and a bicycle lane in the place of a conventional railroad because of how much guard space you need for the embankment, overhead catenary and guard fencing. Also, level crossings are discouraged for safety reasons, so any intersection would need to be an under/overpass.

          2. “big” electric trains can acellerate fast as well. In the Netherlands most trains are EMU’s (Electric multiple units) and especially for the “local” trains (they still travel long distances, just stop at every station) and they usually have about every other axle powered, with a power/weight ratio of 11kW/ton being common.
            These trains usually consist of 3-6-car units, but can be coupled together to form 200m long trains capable 1400 people in one go.
            Oh, and these trains run most of the network, 7 days a week, at least every half hour from 5:30 till 0:30 (starting later in the weekend) and on some trajectories up to 6 times per hour. This in addition to the also 2-6 intercity trains per hour.

        2. Infrastructure and scheduling would be the issue – you’d need a lot of rails to manage that. It would be like trying to build a road directly from every possible departure point to every possible destination. It’s also not friendly to the people to live midway between.

          Roads and rails both solve the problem in equivalent ways – highways, expressways and surface streets for roads, and intercity, regional rail, and subway/light rail/streetcar for rail.

      3. Oh no! Rail can never work! Let’s give up!

        And you’re showing your prejudices. How else will mental patients and drug addicts get around? You surely don’t expect them to drive!

        You could work on mental health and transportation at the same time. Win win!

        1. I don’t think it’s so much that rail doesn’t work or that I have a problem with it but the sheer effort it would require to make it to where every city has access to it throughout the continental US would be monumental to the point of unreasonable. Would I love to take a train and not play amateur dodgecars with all the morons around me who don’t deserve a license? Absolutely. Will I hold my breath for it to come to my doorstep before I die? Let’s try! *inhale*….

          The problem is not trains. It’s who makes it happen and how it gets implemented. The US also currently has a massive housing shortage in several areas – and when it comes down to it, especially in some Southern US states where a high property tax makes up for the lack of income tax, you are asking the state and local governments to drop one income stream (housing, which earns property taxes) for an expense (building rail infrastructure, which will require money/taxes). It’s definitely a chicken/egg scenario which (I assume) most European countries don’t need to manage. Then there’s telling people who can’t afford housing in the US right now “oh no, we’ve can’t make housing cheaper, we have trains to build.” That’s a lead balloon waiting to plummet before it flies. No amount of hope and good intention circumvents putting food on the table.

          Maybe rich people do it, right? Elon Musk is a billionaire who couldn’t manage to get a Hyperloop (yes, it’s a different type of train but good grief, he could have just pointed in a direction and paid to build something functional) built in any reasonable amount of speed despite having more than enough money to set on fire for the shiny object that distracted him at the time. Then again, this guy sold flamethrowers and managed to burn down Twitter without a single one of them.

          Then there’s the pesky issue of what company would fund all of this. All that metal and rail and tech costs money. Sadly, companies have this thing called a “profit line” that if it doesn’t reach, then your venture capital investors get their knickers in a twist. Does that suck that some rich people won’t skip on that yacht and throw their bonus towards infrastructure (not to mention maintenance, etc.)? Absolutely. Can one person change it by bitching on an internet forum about it? You be the judge.

          Maybe Warren Buffett can convince the other billionaires to buy train companies too. Of course – he’s sitting on a crap ton of money too – and I don’t see BNSF (the railroad that Berkshire owns) spitting out new railways everywhere. And as someone who lives near a major grain distribution area, some of the current rail systems are *surprise!* poorly maintained.

          And if you wonder “why can’t the government just do it?”, then I am going to make a guess and say you’re not familiar with US politics. Just imagine a dumpster fire…. and that’s it. Just imagine a dumpster fire.

          Oh and on that subject of “prejudices”. “How else will mental patients get around?” With a car or bike I assume. You see, “mental patients” like veterans, schizophrenics or anxiety sufferers are still entitled to the same rights as everyone else as long as they’re not screwing around with or hurting anyone else (not that this is always honored here in the US, to be fair). I’m technically what you would call a “mental patient” – and I can drive a car just fine.

          1. “Maybe rich people do it, right? Elon Musk is a billionaire who couldn’t manage to get a Hyperloop”
            That was literally a thing he made up to halt the building of high speed rail in California. He knew HS rail would impact the EV market in Cali. So he came up with some BS to push and prevent HS rail. It worked. HS rail was stopped. Hyperloop served it’s function and died.
            You picked the worse example.
            To pay for rail just get billionaires and their corporations to actually pay their fair way in tax. Problem solved.
            Bad politics is bad politics and is a different argument. Trains can, did, and would work for the US.

      4. > The more stops you add, the slower it gets because trains can’t accelerate or stop quickly.

        London Underground seems to work very effectively despite having stops sometimes only a couple of 100s of yards apart.

        Horses for courses – you have smaller rapid trains (or trams) with lots of stops in the dense centre and offload people onto faster longer distance trains at hub stations. It’s not rocket science, it’s been working quite well for about 200 years here.

    2. I think we must first engineer the self-driving road. The DLR (Dockland’s Light Railway) in London is a good example of a self-driving road where there is little chance of a vehicle encountering an unexpected obstacle. Most of the safety, comes from physical separation, people can’t cross the road because they can’t access the road.
      However, I still think there is scope for autonomous vehicles on local roads, but I doubt they will be more tin snails slowly moving along predefines routes. We can’t rely on the robot driver to always react in sensible ways (after all human ones often do daft things), so our road needs to be intelligent and be able to see the local as well as the regional set of conditions.
      In short, we should be reinventing the railway, not the car.

    3. That’s the good thing, you can invest _your_ money anyway you please. In the meantime they’ll take your advice regarding _their_ money ‘under advisement’.
      Hint: “I’ll take that under advisement’ means the same thing as ‘bless your heart’.

      I fully expect ‘self driving’ cars to ultimately fail at anything more complicated then lane following on freeways (they’ll get there eventually), but it’s not my money. Not like they’re spending it subsidizing bums like the government.

  3. I’m surprised that anyone is surprised by these problems.

    First it was always a far reach to assume that the autonomous car challenge could be met without altering the environment too – like embedded sensors and signals, and central traffic control.

    Second, it should have been obvious that in order to be safe enough and to avoid liability claims, a self-driving car will have to drive like uncle Albert on a Sunday afternoon – obeying all laws, not speeding, not bumpertailing, and executing all moves with an abundance of caution. Which means that all the risk-taking human drivers around it will of course hate it.

    The more you look at self-driving cars, the better public transit, trains and bikes look. The problems with cars is that there are too many in the city.

          1. Bro you dismissed trains with “Long as we can keep them on the tracks” why should anyone take your assessment of what does and doesn’t work seriously?

      1. “Bikes? Death traps anywhere near vehicles.”
        Bullshit, it just needs some infrastructure changes and adjustment of driver attitude and training. Maybe with US drivers, who seem to be incapable of looking anywhere but 6 feet in front of their own vehicle.

    1. >The problems with cars is that there are too many in the city.

      On the contrary. There’s just not enough road, or, there’s too many people crammed in the city for the amount of transport infrastructure it can offer.

      1. I guess I still suffer from the delusion that urban infrastructure should support all people, and not just cater to one selfish and inefficient means of transportation.

          1. Not a false dilemma by any means. There are any number of cities in North America that are horrible for anyone with no car. Mississauga, ON, is one of the worst examples I have personal experience of – for decades, if you tried to get around it sans car you would regularly face the prospect of walking literally miles out of your way to reach a point where you could safely cross a six-lane road, in order to reach a destination that was directly opposite your starting point. This was at a time when the city was almost entirely made up of single-family homes – the classic “bedroom suburb”.

        1. By allowing cars into the city, you support people visiting from places where only cars are practical. If we could plan out a city in advance, we could try things like having walkable islands surrounded by better-flowing streets. But instead we cram some things together and then a supporting area sprawls out around that with everything that was left out plus a few other things besides, and it ends up failing to optimize how likely it is someone can actually get where they need to go. (E.G. no grocery stores within a certain radius, or they’re on the other side of the interstate).

          1. From the early 20th century, city planners have used the “if you build it more will come” excuse to not spend money on infrastructure, because every time they expand the road network a little bit, it gets congested again from the latent demand.

            Corollary to that, traffic jams are reduced by reducing roads, not adding them, because if you increase the capacity somewhere, you have to increase it everywhere, whereas if you constrict traffic somewhere, traffic elsewhere drops as well because more people start to avoid driving.

            When viewed as an isolated problem, by people who don’t place any value in cars because they themselves don’t have to drive, it’s easy to conclude that the solution to the traffic problem is simply to have no traffic.

          2. I find that explanation imperfect; I think often the problem is with what specifically you’re adding or removing and how it interferes with allowing people to move continuously without turning or slowing down.

            As an example, it should be reasonable to assert that if a rural divided highway crosses a lot of less commonly used roads or even driveways, then the users of the larger road are disrupted less if they do not have to change lanes or stop for people who need to turn. If you have a full width shoulder, perhaps even painted as a right turn lane, there will be less disruption to traffic from those using the smaller roads than if the highway used that space for an additional lane.

            Other kinds of limited access / reducing the number of awkward connections between roads are good in other situations applicable to cities, and are better than adding lanes to every road without thinking about it. There’s a variety of things you can play with to try and make thru traffic stick to using fewer roads even though we didn’t properly lay things out to make that easier and faster. But getting rid of roads completely just shuts the gates to anyone who might live nearby a city but outside it.

            Related; the sentiment has often been that a residential grid of thru streets allows people to split between more roads, easing congestion. I think that only sometimes happens, and mostly the number of intersections and lack of ability to have more efficient forms of intersection are often problematic. A larger grid surrounding and bypassing most of those seems more reasonable. I’ve seen some evidence that cities have closed more streets to thru traffic and made some of them one-way when they were originally neither.

          3. There is a nifty thing in Europe called a Park and Ride.
            A huge carpark outside of the city. Usually just off a major road like a highway. That then in turn has bus, tram, metro or train links to the city centre.

      2. The mayor of Paris, France does not agree with you and the city is transitioning from car dominance to more varied ways of transport. A few streets are already converted form thoroughfares to destination streets with half the tarmac specified for cars. the rest is for bikes and or buses. they even have a very dense grid of public bikes, so you don’t even have to own your own.

        but your comment reflects the reality that change is difficult. and yes, car drivers in Paris have to learn that there are other people on the street as well. Here in the Netherlands we grew up with bikes and scooters around cars in the city, so we are alert.

        1. From the examples I’ve seen, streets and areas that were “reclaimed” for pedestrians tend to see less business and experience an economic decline because the locals who hang around there aren’t enough – the businesses depend on people who drive into the city to spend their money there. It depends on whether there’s plenty of curbside parking (or a parking garage) right next to the destination street – otherwise many of the customers just drive elsewhere and won’t even attempt to go downtown because there’s too much congestion and not enough parking.

          Our city built underground parking and put elevator shafts up to the street level above to get around this problem, but that’s not really reducing the cars or the roads.

          1. Before the underground parking was built is because the city lost two big multi-story walk-in department stores right in the middle of the best commercial district. They were already surrounded by promenades with poor access by car to begin with.

          2. Most cities I’ve ever seen have none, or at best one parking space outside a business in nearly all cases, as cities tend to be really darn old and built upon and around centuries old foundations with the exceptions being the relatively modern invention of business park and things like IKEA, Car dealerships etc that build in the carparking space as a core part of their existence.

            So usually the people have to walk in to the business anyway, taking away cars on those roads makes no difference to them at all. Unless it is no longer possible to get into comfortable walking distance via other means.

          3. Oh also lots of cities have been loosing their old gems of highstreet commerce, as online shopping takes over – nothing to do with traffic rule changes. Simply why spend £50+ more on the product and have to spend an hour or two going around the store when you can pick something from an online catalogue and have it delivered? The answer most folks come up with is that there is no reason to do so…

          4. > none, or at best one parking space outside a business in nearly all cases

            There’s one or few parking spots immediately in front, but every street has a few spots that are distributed all around, so there’s bound to be one around the corner somewhere nearby – unless they’ve blocked the road and made it into a promenade or replaced the parking with bicycle lanes.

            The problems start when the population grows and the traffic increases, but the city refuses to re-structure and widen the streets for the increased traffic. Then they start arguing with the “If you build more, more will come” fallacy to avoid spending money on it. Here it got so bad that you had to walk all across from the other side of town to find parking, miles away, until they finally got the point.

          5. > when you can pick something from an online catalogue and have it delivered?

            Two problems: 1) 3-5 working days, 2) they deliver to a package automat that is somewhere downtown, or you pay extra for the courier.

          6. >> the businesses depend on people who drive into the city to spend their money there

            People do not inherently “drive to” locations, they travel to them. The mode is completely unimportant to the businesses in question, with the possible exception of automotive-related businesses.

          7. Exactly, people outside of an urban area don’t have these other options – they essentially have to drive, especially if they’re going to go and buy a physical object to carry home. So they *drive into* a city as you said.

          8. >People do not inherently “drive to” locations

            Often in cities, the people who live within the city proper are 1:5 to the people who commute in from elsewhere. There’s multiple communities 20-40 miles away that don’t have the shops and services available, so they drive to the big city to get it.

  4. Trains, which are restricted to the paths they can take and do not have to contend with dozens or hundreds of points of information at all times, would not have this problem. I don’t think that self-driving cars will ever be the solution to our transportation issues. But rich dudes want to pretend they’re innovating solutions for humanity because making wild promises and burning through venture capital money is easy, while the actual solutions (governments investing in infrastructure for the people they govern) is boring and difficult.

        1. Even a small train has much less traction than rubber wheels against asphalt. The entire point of trains is low friction with metal against metal.

          A freight train takes about 2-3 km to stop. A short passenger train can do it in 700 meters. Metro or subway trains, about 100-200 meters. A modern car traveling at highway speed can stop within 50-70 meters. Trams get more or less the same, but only because they travel at much slower speeds, and where trams go skidding 50 meters in an emergency stop is often too much – the wayward pedestrian is already long dead..

          1. For what I found on a quick search, the standard warning distance for local passenger trains around Europe is 400m, up to 6000 meters for longer and heavier trains.

      1. Not really a problem once you understand the concept of rails. Keep pedestrians, cars, wildlife away from them and there is little reason to emergency stop a train. It’s not like it’s going to suddenly swerve onto the footpath.

    1. Are you just the same guy trolling this conversation with trains?

      Trains are not a solution for city transit, they are expenssive and fixed for massive mass transportation. Busses are much better and are more flexible to changes. For longer travels, trains are fine.

      Sure self-driving cars are way out there in the future, but they will come eventually. Personally i will not be waiting a train or bus to get me to work and other places nearby and waste hours of my day. I will drive a car until they literally slaughter me for being an infidel. As for self-driving cars, level 5 or no thanks.

      The amount of self-chaining, unbelievable.

      1. Trains are not always giant high speed intercity with the standard gauge or larger track spacing, lots of smaller lighter railway systems out there, and trams that bring the self guided and remotely powered train to a road…

        I’m also not sure self-driving cars are that certain to come – the world is changing, folks are becoming more aware they need to be active, the need to travel is going down as mail order and work from home is more common, the cost of moving a huge metal box with 1.01 people in them on average is likely to go up, and if it takes too long personal aircars are likely to turn up first. Those are actually in most ways easier to make self-driving, as they don’t need to be aware of all the dynamic moving targets there are at ground level. To me it seems we are just as likely to abandon cars almost entirely before self-driving gets good enough to be in general use – on the motorway and other major roads perhaps is close but on that fiddling bit at the ends of most journey…

        1. It takes a lot of power to keep a vehicle in the air, and what would be a minor problem in a ground car can be fatal in the air. Power is expensive. Expert maintenance is expensive. Throwing sand and gravel around with each takeoff and each landing is dangerous and unacceptable. Control in brisk winds close to immovable objects is probably impossible.

          Personal aircars will never be commonplace in any reasonably foreseeable future.

          1. I did say ‘if it takes too long’ – but with the challenges of dealing with pets, toddlers, potholes, bicycles, lighting and weather conditions getting sufficiently reliable self driving inside the urban area does not seem at all likely any time soon.

      2. >>Trains are not a solution for city transit
        That depends very much on your definition of “train”. For most of the world, “trains” start at streetcar size and move up from there. Different sizes suit different purposes – intercity trains to go from city to city, commuter trains to get into the city from the suburbs, subways or surface-level equivalents (e.g. Berlin’s S-Bahn or the Chicago El) to travel cross-town, and streetcars/LRT for shorter intra-city trips. Most major cities around the world readily recognize that all of these are key to a city that functions well for all.

      3. When I worked downtown, I took a train, and rode the ~1 mile at either end with a folding bike. or I took transit when the weather was bad. Easy peasy. WAAAY less hassle than driving in.

        Forcing a significant proportion of the working population to drive downtown each day is madness.

  5. I am a techno nut, have lived the last 70 years of messing with bl
    electronics, but, I think that until tramways/special pathways, are built for electric cars, they are a waste of time and space.
    We have regular cars for people who need/want/have to drive, we have Taxis, busses, trains, car services, for people who cannot drive, or do not want to drive, or own a vehicle.
    Put the ‘self’ driving cars back in the shop/museum/dump. I hope no one develops a self driving aircraft. There are automated aircraft but eyes & ears are still required.

    A

    1. Well, special pathways for electric cars are sort of already a thing, they are called trains.
      Once you start going into any sort of adaptation for self driving anything, you basically have a train immediately.

      And thus it follows that self driving trains are the best starting point, and also the solution to most travels.

      For example a bus line, split the bus into smaller units, self driving with any adaptations, and you can summon one on demand. If there’s no riders, the “busses” can stand still and wait.

      And a bus is pretty much a tram with rubber wheels, so whether there is a rail or a reserved lane(or both) is today sort of a philosophical question, guiding a cart along a predefined path is a solved question today.

      As such, the difference between a tram and a bus is whether others respect and fear and understand the situation.

    1. This is something that would work in any situation that isn’t high-speed, and the cars already know to lower its speed in the face of uncertainty. Most of the time, the person wouldn’t have to drive the vehicle, they’d just have to tell the vehicle how to interpret what it’s seeing, and give it instructions for how to handle it. Worst case, the operator would have to manually turn the vehicle around and mark a path as impassable.

    2. Remote operation would be a possible solution, but that’s taking a step away from the myth that you can have self-driving cars without needing any other external changes or services.

  6. There’s no way to say anything meaningful without knowing what they already do.

    Falling back to human drone pilots for difficult to interpret situations sounds like a no-brainer, but it’s probably not something that the AI people want to consider. Nonetheless, Mechanical Turk Fallback would solve everything except network disruption.

    Increasing the complexity of the vehicle to include, for instance, the ability to move the car sideways might allow it more latitude in getting out of the way but, again, that’s something AI people would consider “not their problem.”

    Things that they’ve probably already considered:
    – reducing the complexity of their data set. When I drive, I don’t actually know the make, model, or color of the vehicles around me. It’s unimportant compared to things like erratic speed and steering.
    – separate interpretation models for pavement markings, traffic signs, people, and bicycles.
    – a supervisor mind that tracks them from HQ, and makes recommendations based on city-wide conditions.
    – connection to the city’s emergency service dispatch to watch for things to avoid.

    Suggesting that we improve our rail infrastructure is a tone-deaf answer for political reasons. Asking AI specialists to become politicians suggests that the commenters don’t know how politicians work.

    1. “Suggesting that we improve our rail infrastructure is a tone-deaf answer for political reasons. Asking AI specialists to become politicians suggests that the commenters don’t know how politicians work.”

      or how politicians don’t work. Which is really the problem.

  7. Take a bag of old tennis balls, dump em on the road and there should be plenty of pace for pedestrians to move downtown…

    Right now me car (2017) does not even read the speed correctly, let alone judging ice, etc in winter.

    I believe in automated driving on highways and continues moving jams, but I don’t see the last mile. Some human interaction concept will be needed, allowing slow car movement. PlayStation Style maybe…
    Also for the officers.

    Greetings
    Georg

  8. I might be totally off base here, but…

    It may be that until (basically) every vehicle on a segment of road can communicate with every other, at a minimum to send/receive position, this will be a nearly intractable problem to overcome. Even then the computational challenge is likely daunting. I’m sure this thought is not very reassuring, for many reasons….

    1. The computer between my ears, with its accessories, handles driving without needing telepathy. What’s needed is the time and investment to develop the silicon technology to the same competency as humans. Be patient.

      1. There’s a big complication in the form of legal matters. We generally silently condone human drivers to break the rules of the road, in situations where it improves traffic flow at a minimal cost in safety (or sometimes even improving safety). Examples are: going across double yellow lines, temporarily blocking intersections, going over kerbs/grass/sidewalks, rolling stops, speeding, not always yielding when strictly required.

        Now, are you going to program the self driving cars to break the laws in the same way, or are you going to be following the law to the letter? In the first case, the problem is that some of the rule breaking will lead to accidents, and angry mobs demanding that the cars follow the law. In the second case, the cars will be inefficient and increase the risk of traffic jams and frustration in other drivers. It may even lead to more accident due to cars behaving in a counter intuitive way.

  9. Move into three dimensions. Separate the directions by height into planes, eight directions should be enough: N, NE, E, SE, S, etc… When you change direction you go to a different plane. Everyone else is going the same direction so no risk of collisions. Airports, power lines, etc. would be no-go zones, have to go around.

  10. It mystifies me why state and local governments still allow this experimental vehicles to use the public streets? Human drivers would’ve been likely slapped with many fines and possibly lose their licenses.

    1. It is in San Francisco. A major mayor candidate (was it?) ran on platform that all lifestyle crimes (drugs, prostitution, etc) will be decriminalized. Many CA police jurisdictions do not police traffic stops or there is at least huge public push to stop doing so for traffic violations.
      The state of policing is so hands-tied that if your house gets broken into (my friend) or you get into a traffic collision by a license-less uninsured parolee the cops refuse to show up (personal experience) then have a terrible attitude when you insist and need a report for insurance. It’s a mess here. It sucks. At least our taxes are super high though.

      1. It’s a beautiful thing watching a philosophy self destruct in real time. See also Seattle.

        Sure the loonyist of the left won’t wake (over 30 ‘have no brain’), but it will put them out of power for decades again. Just a natural generational thing.
        The kids haven’t forgotten, they never knew. Now it sucks to be them for a while, cognitive dissonance is painful, they’ll need a good cry.

        Antifa tards are no smarter than Weathertards were.

        Remember: ‘You don’t need a weatherman to see which way the wind blows’, but you can tell which way the wind was blowing at the moment of detonation by drawing a map of where all the bits of weatherman landed.

  11. i haven’t seen evidence of anyone actually doing it but i’ve been of the opinion that real self-driving will come from an iterative process of assistive technologies.

    at first, it is very primitive. like a regular car with a regular driver, and automatic emergency braking. if the car reaches a certain confidence level that you’re about to hit a concrete wall, it stops on its own. the driver doesn’t even necessarily realize the car has the feature until some day it saves their life. but as the automatic systems become more capable, drivers will pay less and less attention. instead of finnessing the accelerator, they will begin to use the pedal as a binary ON/OFF control. you already see people doing that with anti-lock braking. the vehicle will manage acceleration without losing traction, and will prevent overspeed automatically. instead of carefully judging a turning movement, they’ll just spin the wheel carelessly to the right and the car will pick the right path.

    it seems plausible to me because it’s baby steps, and it’s already happening. you don’t have to solve every problem at once. the automatic systems won’t pick up so many responsibilities until they’ve been proven to be robust. one day you’ll screw up a turn and the car will stop you from destroying your rims on the curb and you’ll go “huh, neat” or *maybe you won’t even notice it happened* but over time it’ll redefine your relationship with the car.

    of course there is still the danger of drivers overestimating the capacity of the systems and relying on them prematurely sigh

    1. > instead of finnessing the accelerator, they will begin to use the pedal as a binary ON/OFF control.

      I was once a passenger in a car with a driver who used the lane-keeping assistant to ping-pong between the shoulder and the divider while he was talking with his hands, on the phone… never again.

    2. >it seems plausible to me because it’s baby steps

      Depends on where the baby steps are going. Do they go up to 100% and above, or do you get 99.9% performance and just ignore the 0.1% of times it didn’t work and call your lawyers in to argue that it wasn’t the car’s fault?

      1. I’m thinking of one particular company whose CEO says “Our self-driving cars are safer than human drivers” (statistically) even while they keep making stupid blunders and killing people…

        1. You can bet he’s smart enough to know he’s just lying with statistics.

          Trying not to give innumerate morons a pass. They all know their lying too, just not aware of how.

      2. As human drivers are never 100% either, and frequently make stupid blunders killing people…

        Doesn’t matter what the failures are, or how they were caused (within reason) humans or the self driving systems will sometimes make them. How many accidents now are because folks depend on the ABS, or the traction control and it wigs out, fails or just this isn’t their normal car it doesn’t have the same feature level?

        I agree with Greg here the assistive techs will very likely over time become more and more like a self driving car, and may well get there without anybody trying to bundle the whole system into a one point of failure self driving module.

        1. It’s the ability to choose whether you want to risk yourself, versus having risks imposed on you by a system that was deemed “good enough” by bean-counting businessmen and lawyers.

        2. Always remember that the average risk of accidents in traffic involves a small minority of drivers who are drunk, old, or reckless, or driving already broken vehicles out of neglect, causing the majority of accidents.

          When the self-driving car is judged “good enough” by statistics, meaning it performs marginally better than the human average, most people would face greater risk of injury stepping into such a vehicle.

      1. A few years ago, I went backwards down a hill in a car with traction control that would not let the wheels slip forwards, but would let them roll backwards to prevent said slip.

  12. someone here on HaD suggested what I thought was a great idea, give all police and DOT personnel a special app on their phone and when they put it in front of the car’s camera (or BT connect, whatever) the car will follow exactly 3′ behind the phone to wherever it is guided.

    Maybe even have a device in the car that would allow the passenger the same functionality, though that’s pretty ripe for malfeasance.

  13. I don’t get the point. These are too few examples. Are there overall statistics? Some glitches will always be. Some failures are normal. They aren’t even that dangerous. I thought I was going to read a scientific approach, not a buch of events.

  14. lol I grew up in the bay area. And currently live in a city that is consistently voted worst drivers in America.
    That description of self driving cars messing up is so tame compared to what real meat bag a-holes can and often do that I can’t take it seriously. It isn’t uncommon to see multiple fatality crashes in one drive up the I-5 corridor. Bring on the self driving cars, please! Obviously one guys super bias though.

  15. Self driving under normal conditions is hard. Many variables to take into account. Many tradeoffs between speed, comfort, safety, traffic flow, energy efficiency etc.
    Self driving under difficult or confusing conditions would require a general intelligence that can understand human language and behavior and can recognize danger (wild animals, enemy’s army, natural disasters, etc). I’m not comfortable with combining something like chat-GPT with delf driving cars.
    I’m all for cars assisting humans in driving better. But in a way that won’t make us rely on them too much. People should be required to be trained with these features in drivers ed.
    If a self driving car blocks traffic they should be giving out huge fines to the owners.

  16. Autonomous cars are part of the business model of “transportation as a service.” It will allow corporate owners to do away with those pesky uber and lyft drivers.

    Client: Allexxa, why are we stopped?
    Allexxa: We are not stopped. We are updating. There is an additional charge to travel without ads.

  17. For me this just goes to show just how well humans do with driving with just 2 main sensors. We arent perfect and our insessent need to breed and make more of us means more and more road users (thus increasing the likelyhood of accidents), but it is staggering how many sensors a self driving car needs to do only a fraction of what a human can do with just the Mk1 eyeball.

  18. I think part of the autonomous vehicle puzzle comes from the idea that the cars must be, as you put it, “rugged individualists.” Human drivers in congested places are not that.

    We pick up all sorts of cues from other drivers. We catch their eyes, and a subtle not from one tells the other to go ahead. We guess the intent of pedestrians from their body language. In parking lots, we look at the front wheel direction and reverse lights of other cars in attempts to discern their drivers’ intent and stay clear of them. We use hand signals and light signals to communicate with other drivers. We roll down our windows and speak to public safety people and other people.

    There’s a lot of inter-driver communication unaccounted for by autonomous systems. That, and the ability to act on it, are part of what’s missing.

    Don’t get me started on “autopilot” systems and inclement weather. When autonomous vehicles designed in Detroit Michigan rather than Palo Alto CA hit the roads that situation will necessarily improve.

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.