By on December 10, 2012
YouTube Preview Image

The notion of the crossroads as a sacred place is older than Robert Johnson, older than the blues, older than the automobile or the Romans who laid the first large-scale transportation network. The crossroads contains possibilities, changes, choices. It is where the gods live and the demons dwell. Soon, that may be almost literally true.

A recent Virginia Tech study suggests that the best way for driverless vehicles to handle intersections may be to hand over their decision trees to an “intersection controller” 200 meters or so before they actually reach said intersection. The “intersection controller” will possess local intelligence and will adjust the speed and behavior of incoming vehicles.

… the intersection controller governs the vehicles within 200 meters from the intersection. The vehicles report their physical characteristics, such as power, mass, speed, location, and acceleration.

“The aim of giving complete authority to the controller is to overcome any selfish behavior by an autonomous vehicle and benefit all vehicles in the intersection zone… The controller determines the optimum speed and acceleration at each time step for every vehicle within the intersection zone by processing the input data through a real-time simulator/tool.”

The system will begin testing at a Virginia Tech roundabout in the near future.

Speaking as someone who occasionally designs complex systems in his day job, I can’t decide if the potential for mayhem is made greater by allowing all the driverless cars to make their own wacky decisions or by turning over control to a system which could easily be compromised. If you have ten thousand cars going through an intersection every day, the chances are high that one of them will misbehave, but the others should be able to compensate. A malicious “intersection controller”, on the other hand, could just stack up dead bodies like cordwood until its registers overflowed.

It’s also a certainty that people will figure out how to game the intersection controller for their own benefit. One way would be to program your car to report brake failure 200 meters out. The intersection controller will cheerfully stop all other traffic for safety’s sake and you can just glide through like royalty. Some kid with a laptop can also sit under the intersection controller and generate phantom traffic. If all the traffic is encrypted, it can be recorded and replayed to cause havoc. Alternately, one could impersonate an intersection controller, pull a solid man-in-the-middle attack, and make pretty metal sculptures.

There’s something very touching about the way advocates of increased central control assume that everyone will just happily go along with whatever scheme they’ve dreamed up this week, whether it’s intersection controllers, mandatory smog inspections, or 85-mph speedometers to limit highway exuberance. There are more things in heaven and earth than are dreamt of their philosophy. The crossroads are a place of chaos as well as divinity. Ask Robert Johnson.

Get the latest TTAC e-Newsletter!

47 Comments on “At The Intersection Between Humanity And The Future, Robots Will Be In Control...”


  • avatar
    Landcrusher

    You make really good points about security. We really need to get the whole cybercrime thing sorted.

    • 0 avatar
      DeadWeight

      Yes, the benevolent government, with the “intelligence” agencies that have to scare in order to justify their existence, are working diligently and successfully on that.

      http://www.wired.com/threatlevel/2012/03/ff_nsadatacenter/

      • 0 avatar
        Landcrusher

        Yeah, that’s precisely not what I meant. If you steal 10k from a bank, you get caught, but if you steal 100k from a group of people they don’t bother to follow the money. And if they do, apparently you get a slap on the wrist. Certainly, a slap is in order for hacker pranks that are just irritating, but destruction is a whole other thing. The government just doesn’t appear to have any of this sorted out.

  • avatar
    marjanmm

    Not sure in US but there are already traffic lights optimization systems in European cities which monitor the traffic and in real time update the traffic lights in the congested zones.
    The traffic here is often so congested that anything that can enable me to get to work 10 minutes earlier is very welcome and this concept if it worked correctly would be in average very beneficial to everyone.
    Besides, a very large percentage of accidents in cities are happening at intersections, the use of a central regulator would certainly decrease the rate even if it is imperfect at first.
    Fear mongering has rarely stopped progress. Distaste of government/central control might. This post seems to be a mix of both.

    Of course not everyone would go along willingly but if a law requires such systems, everyone will have to comply albeit not happily.

    • 0 avatar
      67dodgeman

      “if a law … everyone will have to comply”

      Obviously you’re not very familiar with the southern population of the country, then. ;)

      We believe if a law can be broken, well, hold my beer and watch this!

  • avatar
    Maymar

    And yet I still trust the machines more than my fleshy compatriots. At least when something goes wrong with the machines, someone might actually try and fix it.

    Plus, if the machines ever rise, having documented support of them couldn’t hurt.

  • avatar
    stryker1

    Sure, there’s a lot of room to design autonomous-car-systems in really stupid ways. What they’re talking about here is so blue-sky far-future that it doesn’t really matter. A future where some of the cars on the road are self driving is probably not happening in the next 5 years. A future where all of the cars on the road are self-driving, which would justify and necessitate such a system is far enough off that, by the time we get there (if we get there), propositions like this one will seem perfectly reasonable.

  • avatar
    Lorenzo

    How many people read ’200 meters’ as “200 feet”? That’s 656 feet, and the average rectangular city block is 280 feet by 650 feet, intersection to intersection. In other words, your car would be under the control of at least two controllers, and as many as five, SIMULTANEOUSLY in most urban areas. Does anyone see a problem with that?

  • avatar
    67dodgeman

    Was a story, I think locally, where the emergency vehicles would beam a signal ahead and traffic lights would switch to give green lights for all emergency traffic. Someone outfitted their personal vehicle with a black market device that broadcast the same signal, and presto – daily commute with no red lights! Took multiple complaints about strange light changes, poor timing, etc. before someone investigated enough to find him.

    If it can be done, someone will do it.

    • 0 avatar
      WRohrl

      Where I live the emergency signal turns all lights to red. Problem solved, everybody theoretically stops at the light, the fire engine goes through it. Using the device doesn’t help the civilian driver…

    • 0 avatar
      Syke

      40 years ago, Erie, PA (which is built on the flat, so the street system if almost a perfect checkerboard) used a system that allowed police and emergency vehicles to switch all the lights green in the direction they were traveling (with a flashing white light explaining to the motorists why the light had suddenly changed). I rapidly learned that the fastest way across town was to get in behind an ambulance and just shadow it. No red lights until they reached their destination. And yes, every so often somebody hooked one of those up in their car.

      • 0 avatar
        d524zoom-zoom

        I have a friend/ co-worker that has a very elaborate CB system with 4 alternators, and several industrial type batteries in his Suburban that if he keys his mike near traffic lights it will make them go to flashing yellow. The next thing out of his mouth if he makes that mistake is “better not go this way for a while”….it is somewhat funny.

  • avatar
    TrailerTrash

    I am as paranoid as any ex drugged hippie.
    But let us control our fears and demons somewhat, OK?

    Look…every commercial airliner today is operating without the pilots during most of the flight/take off and landings. There are a great many electronics and back up electronics working to make the city of 270 people get from airport to airport safely.

    These new transportation machines have outpaced the human mind and skill level to function. It is “pilot error” that causes the disasters today…not the machines.

    So eventually we need to accept that we modern monkeys might love to drive, but we cannot drive as well as the machines we design. We fail.
    We age. We are not all equally skilled to begin with.

    To suggest these systems will pile crushed car upon crushed car is nieve…and a bit misleading. They will be multi level backups just like in most modern systems.

    The ego has to go….and eventually we will be transported around by systems….not as individuals.

    • 0 avatar
      ClutchCarGo

      Exactly. We have such deep fear of own creations that we put logic aside. In the described scenario, the individual autonomous cars would still have their own sensors and algorithms that would refuse to follow orders from a central controller that had been commandeered and was directing the vehicle into what could be locally determined as in harm’s way. At most, all of the autonomous cars would report the anomaly and revert to their own decision making. As to hacking your own vehicle to game the central control, autonomous cars will certainly have unique IDs that are reported to the controllers passed. If your car repeatedly reports failures you can expect either a visit from the authorities or to have your car remotely disabled until you show that it has been repaired. If you want to fret about autonomous cars and central controllers, fret about what will be passed to Traffic Control and who gets to review the info.

    • 0 avatar
      corntrollio

      Incidentally, planes also get more local control when you get closer to where you might run into other planes, such as near an airport or a busy metro area, for example New York TRACON, NorCal TRACON, SoCal TRACON, etc. and then, of course, the tower.

      As someone else also mentioned, you can also already hack traffic lights, but most people don’t.

    • 0 avatar
      Landcrusher

      Where do you guys get these myths about airlines and planes? For anyone who read this bit, disregard all the parts that have to do with aviation. Nothing to see, move along.

      • 0 avatar
        corntrollio

        What errors do you find in my statement, sir?

      • 0 avatar
        el scotto

        +1 The autopilot is turned on/off by the pilot and I don’t think combat take offs/landings are done by software. Lots and lots of human control; the pilot is responsible for the plane.

      • 0 avatar
        Landcrusher

        Actually, without reading TTs post first, nothing. But if you read it in the context it’s in it gives the impression ATC is controlling planes from the ground. ATC has no more control than a cop in an intersection. Less if you are in pistol range of the cop.

        Although auto pilots capable of landing an airliner unassisted were first installed in the seventies, most of them still don’t have them. The model that did have them wasn’t a big seller. Odds of your airliner landing itself today are slim, and I don’t think any of them takeoff that way even today. Landing and takeoff are mostly hands on, it’s cross country where George (the auto pilot) does most of the work.

        Also, our ATC, being an uncared for, unsexy, uninteresting bureaucracy is so insanely backwards that they are at least 20 years behind on automation they should have. The first near death of general aviation at the hands of the lawyers (also in the seventies) is at the roots of the problem along with several other factors. No one likely wants to make an autopilot responsible for an airliner taking off and landing unless they can get an extra million a box for the liability insurance.

      • 0 avatar
        corntrollio

        I wasn’t suggesting that ATC controls planes at all (although TCAS is close), but rather that the concept of local control of busy intersections wasn’t uncommon in other areas. I just happened to reply there because someone was writing about aviation. The main suggestion was that it makes sense to research it for automated cars because it’s an extension of what we do already for humans.

        You’re right about ATC being backwards these days. Congress doesn’t appropriate them enough money to improve their operations. There are numerous things we could be doing to make flight operations more efficient from an ATC standpoint.

        Re: autopilot, my impression was that autopilot is mostly used for cruising, which appears to be correct. That it’d be used for takeoff/landing was news to me, but I didn’t have enough info to doubt it and marked it down as something I would research. No need to do that, I guess.

      • 0 avatar
        TrailerTrash

        Oh…please.

        My best friends in and around Chicago are pilots or air traffic control at Ohare approach.
        You don’t know what I am talking about.
        My AA friends says there is a standing joke amongst pilots The planes are so well self piloted all that is really needed is a dog.
        Ti bite the fucking hand of any pilot who tries to touch the controls.

        Yes..at final approach pilots do take over…but come on…you’d be amazed at what time immediately after take off auto goes in.
        You simply punch in coordinates given by air traffic and the damned plane does everything.

        And come very soon…more satlie control is coming.
        Get over it.

        And I ain’t taking anything away from the pilots..I am giving respect to the programing and designs of the machines they are flying.
        Hell…they are such great machines their top pilots can’t handle small planes. They aren’t nearly as forgiving.

      • 0 avatar
        TrailerTrash

        Landcrusher

        nowhere did I ever imply planes were being controlled from the ground.
        I reread my stuff…and see how I was taken wrongly. I was simply trying to explain the extent to which complex systems DO run many of the flight periods…including very close to landing and immediately after take off… Look…it is STILL complex..only different today. You are more concerned with a knob than the stick.
        We went from 3 to 2 pilots. 4 engines to 2.
        But we will stay for a long time with the capt and co pilot…as sort of pilot and back up system…just like most back up systems on the plane.
        But it is much more exlectronic and hidden.
        My train at the airport(s) are all driverless.
        Many planes are pilotless today in the miltary.
        But that doesn’t mean I am for commercial aviation as such.

        But the fear of technology is funny.

      • 0 avatar
        corntrollio

        To be more precise, it was from 2 pilots and a flight engineer to eliminating the flight engineer.

      • 0 avatar
        el scotto

        @ Trailer Trash Almost no pilot-less planes in the military. The ground guys do have the cool backpack drones. I used to work Predator. Unmanned does not mean autonomous. That’s on my resume with the dates and a one line sentence about what I did. As Forrest said, that’s about all I’m gonna say about that. Your train at the airport has a simple operation. It goes back and forth between two fixed points.

  • avatar
    mikenem

    “I fear the day that technology will surpass our human interaction. The world will have a generation of idiots.”

    - Einstein

    Today I obeyed a sign that says Do Not Block Intersection. I left a space for others to get through while I was stopped at a light. 4 people waived as they drove through and I politely waived back.

    • 0 avatar
      Freddy M

      Sounds like Einstein was having a Nostradamus moment, eerily describing the early 21st Century, what with our new fangled mobile genius phones and interwebs, and replacement of friends with Facebook groups.

      Dagnabbit.

  • avatar

    Reality time: Robocar version 1.0 will be a car that studiously, even perversely conforms to present traffic laws and conventions. Stryker1 gets that exactly right up there.

    There’s lots of ways that this could go wrong, but complexity aside, I’d point out that a creative kid could vandalize a current “dumb” traffic signal, and yet they don’t bother.

    Also, I’d expect that the default will be that even when your car gets a clearance to pass through the intersection, it won’t use that blindly. It will still have its own sensors, and a car crossing in the wrong place and the wrong time* will surely trigger “something’s wrong” mode.

    Also, note they’re testing this first at a roundabout. Roundabouts are great intersections for this kind of design, because the available level of mayhem is much lower than in a conventional intersection. Indeed, you don’t need to negotiate much: this could work like any driver approaching a roundabout, only now the driver has a seemingly-psychic ability to predict that there will be a car-sized gap in traffic at exactly the right moment, as long as approach speed from 200m out is adjusted to 12 m/s.

    *Yes, the cars could theoretically be crossing through the intersection at inhumanly tight clearances, but an implication is the robots will have a precise idea of where approaching cars should be at any moment. If your car gets to the intersection and sees a car .25 seconds ahead of its expected position, then you cancel the intersection entry and emergency brake promptly.

    • 0 avatar
      nonce

      I’d expect that the default will be that even when your car gets a clearance to pass through the intersection, it won’t use that blindly. It will still have its own sensors, and a car crossing in the wrong place and the wrong time* will surely trigger “something’s wrong” mode.

      Absolutely. Computer-driven cars can already detect people running red lights, and they don’t just T-bone into them.

      The amount of FUD in the article and the comments is silly. Even if all cars were computer-controlled, you have to assume that out of millions there would still be thousands that are broken, so cars won’t blindly trust what is being reported by their neighbors anyway.

      Cars are going to be so paranoid that the real chaos isn’t going to be someone causing crashes; it’s going to be someone broadcasting an emergency signal that makes all the cars pull over.

  • avatar
    Lynn E.

    “if a law … everyone will have to comply” Obviously you’re not very familiar with the southern population of the country, then. ;)

    Fortunately those states are leaving the union anyway. You know, the states that collect more federal money than they contribute in taxes.

    Every time I read comments here about safety or government regulations I am aware that many don’t understand population. In spite of a lot more people in the US we have been able to keep our traffic death toll about the same since the 50′s.

    This didn’t come about because of warm hearts at GM. This came about because of federal regulations that have improved our roads and forced the car companies to improve our vehicles.

    • 0 avatar
      28-Cars-Later

      “This came about because of federal regulations that have improved our roads and forced the car companies to improve our vehicles.”

      …and thus creating a nanny state.

      • 0 avatar
        smartascii

        I don’t know that this is true. One often-accepted definition of government is that its purpose is to keep citizens safe, both from outside nation-states and from each other. If the government creates a law that makes it illegal for me to kill you and punishes me if I do, well, that’s good government. If the government creates a law that makes it illegal for me to drink a fifth of rotgut and drive my multiton vehicle (and act which MAY kill you), is that different? And if the government makes a law which requires that roads be designed and built to specifications that increase safety, or that vehicles sold here are built and equipped to a standard that reduces your risk if I crash into you, is that different still?

        At what point does it move from good governance to nannyism? I suspect that it happens at the point when it stops you from doing something YOU want to do…

      • 0 avatar
        corntrollio

        “At what point does it move from good governance to nannyism?”

        At the point at which you’re not smart enough to come up with good counterargument, so you just use a two-word phrase like “nanny state” to be dismissive. See also, “butt-hurt”.

      • 0 avatar
        28-Cars-Later

        I would say when a manufacturer can’t build the product they wish the nanny line had been crossed. Now before you Nader me to death, picture this: a loose national standard of safety, emissions, size, weight etc. valid in all fifty states. No EPA, no CAFE, no NTSB, no state BS like CARB with conflicting mandates, ONE loose standard updated per decade and agreed to by all fifty states. This could happen by compromise between said agencies.

        If an automaker chooses to go far beyond the standard and become noted for something, as Volvo did with safety and Toyota the hybrid, then more power to them. But if Dodge wants to drop a V8 in every Charger standard, let them. If the Dodge doesn’t want to offer 37 airbags in their V8 Charger and only offer the minimum of two, sounds good. If you don’t like it, don’t buy it. This used to be a free country after all. The market will regulate itself and consumers have to choice to vote with their wallets what matters to them most.

        This is pretty much the opposite of what happens now, where the least common denominator is pandered too simply because its cheaper to succumb to the pressure and amortize the additional costs than to fight the mandate.

  • avatar
    sfenders

    > “If all the traffic is encrypted, it can be recorded and replayed to cause havoc.”

    If you thought of a replay attack on your own knowing little about cryptography, congrats that is pretty good. However, it only works if the people building the thing don’t know much about cryptography.

    This “intersection controller” seems a lot like the traffic lights we already have. Essentially it’s vulnerable to the same kinds of attacks. Somehow it isn’t much of a problem in real life. If the traffic lights go wrong, people react accordingly, some of the time. If the super-intelligent intersection controller goes wrong, the robotic drivers will no doubt do much better.

  • avatar
    Darkhorse

    Having worked in the software business for 30 years, I’m a bit uncomfortable with robot cars. EVERY software system, no matter how much time it spent in QC and testing, has bugs. It’s mathematically impossible to find them all. Do some research on the Airbus A300 and A320. Numerous fatal accidents due to software problems in their fly by wire systems.

    • 0 avatar
      corntrollio

      Hard to take you seriously. The Airbus A300 is not fly-by-wire. The rudder issue of flight 587 was entirely mechanical, although there’s a question of whether it was American’s fault or Airbus’ fault.

      The only A320 crash that I could find in a quick search that was initially blamed on computers was Air France 296, which was an exhibition flight at an air show, but it turns out that was pilot error too. The flight computer said the altitude was 30 feet, but the French pilot thought the instrumentation was metric and thought his altitude was 100 feet (i.e. 30 metres).

      Care to direct me to fatal accidents on A320s that were specifically due to fly-by-wire to substantiate your claim?

      For reference, the A330 actually is fly-by-wire. Air France 447 crashed because of pilot error in understanding how the instrumentation worked, but it was not due to fly-by-wire per se.

      • 0 avatar
        golden2husky

        Wasn’t there some issue with Airbus and speed sensors that tricked the plane into some kind of nosedive? IIRC there was a crash into the ocean that was caused by this, and some planes that went into nosedive conditions and were brought back to control by the pilots?

      • 0 avatar
        Landcrusher

        G2H,
        There have been many Airbus systems scares, but I am not sure they have been found to be the primary cause of any fatalities. Modern ATPs get drilled to deal with those sorts of things, and it’s likely they would be found a contributing cause by the NTSB. I truly doubt the French will ever find an Airbus failure as primary cause, until they give up socialism anyway.
        The one big South American story is just tragic. Those people were killed as a direct result of euro policies against piston aviation. Had the crew had more stick and rudder experience, it would not have happened, IMO. These days, you actually want a Canadian pilot, as they actually have to know how to fly before sitting in an airliner.
        The US pilots are getting in with very little experience as an unwitting conspiracy of government entities, NIMBYs and the FAA/airline/union block have made getting flight experience harder and harder to get here. Fifteen years ago, pilots came to the US from all over the world to learn to fly and build hours. US airlines could expect to hire candidates with triple the minimum hours or more.
        No longer.

    • 0 avatar

      The software doesn’t need to be perfect, just significantly better than the horrible human operators who needlessly cause 11 million crashes and 40,000 fatalities every year in the USA alone.

    • 0 avatar
      Neb

      When dealing with any serious industrial accident, it is very rare for just one factor to have been entirely responsible for it. This is true of the Air France south Atlantic crash. If you were a judge divvying up responsibility for that accident, the blame would be split three ways. On the airline, for insufficient pilot training and not replacing the faulty pitot tubes; Airbus, for its non-intuitive control design; and on the pilots themselves.

  • avatar
    BigMeats

    And where will all these Klaatu-Kompatible vehicles come from? Who will subsidize them for working poor or will those folks be denied private transportation? This would become a class and race issue par excellence.
    It could be Jesse Junior’s resurrection.

  • avatar
    sportyaccordy

    As goofy as the govt is, I really doubt the systems will be so unsecure that “some kid with a laptop” would be able to hack into them and wreak havoc

    By this logic why hasn’t this “kid with a laptop” hacked into any other autonomic transportation system? Post reeks of right wing paranoia

  • avatar
    Landcrusher

    Interesting comment Sporty. What specifically makes the “kid with a laptop” qualify as “right wing paranoia” rather than regular paranoia?


Back to TopLeave a Reply

You must be logged in to post a comment.

Subscribe without commenting

Recent Comments

New Car Research

Get a Free Dealer Quote

Staff

  • Authors

  • Brendan McAleer, Canada
  • Marcelo De Vasconcellos, Brazil
  • Matthias Gasnier, Australia
  • Tycho de Feyter, China
  • W. Christian 'Mental' Ward, Abu Dhabi
  • Mark Stevenson, Canada
  • Faisal Ali Khan, India