Robo-safety?

An article in the Detroit News describes advances in robo-car technology.

And this raises the question: what kinds of car-bike crashes could this technology prevent?

As suggested in the article: rear-end collisions and drift-out-of-lane collisions. Also perhaps right hooks, if a vehicle has sensors and robo-control to inhibit turning across the path of an overtaking vehicle. In all of these collisions, the bicycle or other vehicle to whom a robo-car might yield is, if not in full view of the driver — so robotics would mitigate driver distraction — in full view from a sensor somewhere on the vehicle.

Dooring collisions and drive-outs from parking? Yes, if a sensor in the side-view mirror prevents the door from opening. But this doesn’t prevent the cyclist from striking a pedestrian who walks out to get into the vehicle, and also prevents a person from getting into or out of the car, also potentially presenting a hazard. The real world is inhabited by real people, not only robotic cars.

Could robotics prevent head-on and ride-out collisions? Only unreliably, because of larger distances and greater closing speeds, and because the cyclist, or a pedestrian, or another car, or a deer leaping across the road — may have been hidden around a curve, or may suddenly appear from concealment. The robo-protection might prevent a vehicle from crossing the center of the roadway, or rounding a curve too fast to yield to traffic ahead — but this only works without drastically lowering travel speed if the vehicle, person or animal to be avoided  is operating properly (e.g. oncoming cyclist is on the correct side of the roadway, pedestrian does not dart out into the street).

Left cross collisions? NOT, if the cyclist is overtaking other vehicles, or parked vehicles, close to the curb, where concealed by them.

All in all, then, the technology could:

* greatly reduce the risk collisions for cyclists and others  traveling according to the  rules of the road;

* Significantly reduce the risk of car-bike collisions for cyclists edge riding or riding in a bike lane to the right of other traffic;

* be of little or no help when cyclists are riding in a bikeway concealed behind a barrier, or crossing in a crosswalk.

The potential for robotic control to prevent collisions is, then, very similar to that of an alert driver — though with additional ability to look in more than one direction at once, and into blindspots. Why? Because the rules for efficient and safe interaction — the rules of movement — aren’t any different when robotics are in control.

And I’ll also draw a more sweeping conclusion: robotic technology, once universally and correctly implemented on motor vehicles, would very significantly improve the safety of cyclists and pedestrians, and make rumble strips, a real problem for cyclists, pointless. Many other facilities-based solutions also would become white elephants.

This is assuming that the technology reflects a real traffic mix on streets, including non-motorized travelers, rather than some fantasy vision of a motorist’s utopia.

I’m only discussing technology here which can be implemented independently on each vehicle, without its communicating with others. There have been utopian schemes described in which all vehicles would communicate and negotiate with one another, avoiding the need for traffic signals, but except on limited-access highways, those pose a very daunting challenge due to the need for instant wireless communication with wide bandwidth, and that not all moving objects can be instrumented, nor would people tolerate this on their bodies — also, that a monkey wrench could be thrown into such systems, bringing traffic to a stop, simply by pedestrians’ walking out into the street. The Atlantic magazine has an article describing such a scheme — as usual in its articles about transportation, short in attention to unintended consequences.

Now, am I guilty of the same, in my optimism?

11 Responses to Robo-safety?

  1. I fear it will quickly be found that these robotic cars are unworkable in the real world. Motorists without the technology will be able to force their way into traffic by playing chicken with the robots. They do it now, but humans are willing to play chicken back, so everyone knows what’s happening and things generally move forward. With robots in charge of some cars’ safety, I foresee traffic jams galore as robot cars are brought to a standstill by the numerous motorists who are willing to take the chance and pull too closely in front of a robocar. People will quickly find that these cars won’t be able to get anywhere because their avoidance systems will bring them to a halt every few seconds whenever there’s a traffic jam and someone thinks they can fit into a gap, or whenever a line of cars is going along at freeway speeds and someone cuts in front of a robocar, causing the robocar to apply the brakes to maintain a safe stopping distance. I just don’t see how the developers can get around these kinds of problems.

  2. Good point, Ian.

    Engineering controls are thought to work better than often fallable human ones in complex systems, except for a few nasty situations, i.e., when reality presents a situation that the engineers did not consider or provide a countermeasure to defeat.

    A good example of standard practice failing was the loss of the SSN Thresher in the sixties. The standard procedure during the type of accident that occurred was to scram (shut down) the nuclear reactor and isolate the propulsion system. Of course, in the case of the Thresher, one does not turn off power when the result is to plummet past crush depth. Humans can at times make better crisis decisions than machines when the machines are limited by the ingenuity of their programmers.

    I agree with John that robocars might well prevent several collision types where human failures are common, such as run off the road or rear end collisions. OTOH, the law of unintended consequences could indeed apply. Will drivers get even less concerned with their vehicles, leaving the possibility of an unforseen event lacking the possibility of intelligent intervention?

  3. Another potential drawback is that people may confuse regular models with robo models. If they then try to play chicken as they could safely with a robocar, many more collisions could occur between non-robot cars.

    Yes, in terms of cycling interactions, the potential is all upside, since so many cyclist-car collisions are caused by motorist inattention. In terms of car-on-car, I think the whole thing is a non-starter, unless of course ALL cars are retrofitted with the technology all at one time (which is not going to happen). As a computer gamer and modder who has been playing and modding games since the 1980s, I’ve had a lot of experience with artificial intelligence, and the one thing that is always true of it is that humans always find ways around it: this is true of AI in games and it will be true of AI in vehicles. When motorists get into a car, my experience is that they are hell-bent on saving every second they can in any way they can – for most motorists, the road is a race track. With AI cars on the road, the average non-robot-equipped motorist will try to game the system for all he’s worth, finding every flaw in the AI cars around him and exploiting it to gain an advantage.

  4. Then there’s the almost infinite potential for frustration on the part of drivers of AI cars. Just imagine how they will react when the car enforces laws such as the 3ft passing law combined with the no crossing a double yellow line law. Here in MD, the double yellow is never enforced and government have basically agreed that it is in many cases a stupid law. But Robot cars won’t see it as stupid – they will enforce the law, period. So drivers will have to sit quietly while their robocar idles along behind a cyclist because the car won’t pass any cyclist unless there’s more than 3ft, and won’t cross the double yellow even when there’s no oncoming traffic whatsoever. I foresee many robocars returned to dealerships because motorists can’t deal with the AI’s rigid interpretation of traffic laws.

    Also, motorists will become even more angry at cyclists who control the lane. Just imagine how much they’ll fume when we’re five feet away from the curb, when if we moved three or four feet farther right their car would allow them to overtake. There is a great potential in robocars for INCREASING road rage, and that will surely make cyclists less safe, because their car’s AI can’t stop the driver from throwing a boittle out of the window, or getting out of the car at the next stop light and beating the “offending” cyclist to a pulp.

  5. BTW, I didn’t mean that double yellow lines are stupid. The stupid part is that, unlike many states, MD law allows NO crossing of a double yellow line, even when the oncoming lane is empty.

  6. “* greatly reduce the risk collisions for cyclists and others traveling according to the rules of the road;”

    The risk would also greatly be reduced if human drivers were somehow required to travel according to the rules of the road.

    Honestly, I am supremely skeptical. The solution to our problems isn’t more technology and more consumption of goods. We don’t need a better and safer way for people to move between suburbs and downtown, we need cities where people don’t NEED to do that.

    Many of these robo-car evangelists support their zeal by saying that we can have more efficient roads by having safe high-speed travel. What is considered a “safe” speed is basically proportional to the safety of the vehicles. If we have ultra-safe robo-cars, then the laws will be changed to allow those cars to go faster. Our North American society is used to several thousands dying each year from motor accidents, if we can increase vehicle speeds without killing MORE people, that’s what will happen. It won’t make anyone safer, it’ll just make SOV commuting more convenient and comfortable.

  7. Have to agree with Jean-François – the promise of safety on the roads always seems to end up being a reality of more speed and and no improvement in safety. It’s transportation’s equivalent of the Jevons Paradox.

  8. I’m thinking that if there is widespread adoption of robo cars, many people would probably rather use them like zipcar – annual membership, on-demand use. If you could just have one show up where you are whenever you want and you don’t have to do any maintenance or upkeep (which I suspect will get increasingly expensive). you wouldn’t have to shovel out your car in snow, or scrape it – it would be waiting for you when you left the house, and waiting for you when you left the office. There would be a need to build storage facilities for these cars, but I think suddenly we wouldn’t have any need for on-street parking – just temporary loading zones (which means more space for cyclists and pedestrians!). Of course that’s assuming someone doesn’t have to actually be in the car at all times…

  9. I find the notion that a higher proportion of road users rigidly following the rules of the road might be negative – to be thought provoking…

  10. If everyone followed the rules, it wouldn’t be a problem at all – it would potentially improve safety immensely, and the more cars that do it, the safer things would be. I just think that the average motorist sees traffic flow as much more important than safety, and the problems come with the people who are able to exploit the software in the cars around them, which will (I contend) cause more delays in traffic flow, which will cause frustration among a group of road users, many of whom are seriously frustrated already. If I’m right, all of this may delay by years or even decades any widespread implementation of robocars.

  11. It’s hard to predict the unintended consequences of something so novel. We’ve had robotic devices controlling aircraft for decades now, but then traffic in the skies is much more orderly than on the ground. I tried to look for positive possibilities. I think that they might apply with a carefully-selected set of robotic options.

Leave a Reply

Your email address will not be published. Required fields are marked *


8 − seven =

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>