After the shock of the recent fatal crash of a Metro-North train in New York (Wikipedia) wore off I began to wonder… Google and BMW are very close to creating practical self-driving cars (see recent New Yorker article and one in Technology Review). Why aren’t trains, which operate in a much more structured environment and whose capital cost per vehicle is much higher, already self-driving?
Separately, when my American friends hear about a train crash or a building failure in China they use that as more evidence that something is fundamentally wrong with the Chinese social and political system. But when a train derails in the U.S., that is just bad luck and not a reflection on anything having to do with America or the U.S. social/political system.
And finally, though not related to trains, the above articles make me wonder what happens when most of the cars on a highway are self-driving. Each car is putting out a huge quantity of signals via laser, radar, and ultrasound and depending on getting clear returns. This seems easier when there is only one self-driving car on a given highway. What happens when signals from different cars collide?
It’s easy enough to impose a device specific signature (timing, pulse variation, etc) on the source signal, with the receiver then ignoring what doesn’t match.
A similar question – how does radar work on airliners? Busy air corridors have lots of jets relatively close (I always enjoy seeing planes crossing below) so surely receiving the radar transmissions from other planes and at much greater strength than reflections of their own transmission. What stops them being “dazzled” – timing of the pulses? And is the technique random or does each radar set modulate in a different way, eg using its “serial number” as part of the mechanism?
A lot of subway trains already ARE automated – e.g. some lines in Nuremberg and Paris.
For subways, driverless trains are a solved problem, the only reason why not all subways are automated yet is having someone on the train to do customer support and handle edge cases (e.g. a stuck door), and resistance of unions to train drivers losing their jobs.
*Why aren’t trains, which operate in a much more structured environment and whose capital cost per vehicle is much higher, already self-driving?*
They are.
The Federal Railroad Administration blocks, delays, and prohibits the use of modern standardized train control systems in the USA.
http://pedestrianobservations.wordpress.com/2013/12/04/the-metro-north-accident-and-train-control/
This is the same FRA that prohibits the use of European or Japanese safety standards and equipment and blocks real high speed rail proposals. Its quirky rules are one large factor that help to keep US passenger rail projects costing 5-10x the worldwide market prices.
Tom,
Airliners don’t have radar to help them see other planes at all. The radar on board an airliner is only for weather monitoring purposes. The only use of radar for collision avoidance is the ATC radar on the ground.
Airliners do have traffic avoidance systems that rely on transponders, which are small radios that broadcast a pulse only when they receive an interrogation from another device.
The LA subway system was designed to use driverless trains in the ’90s. That was scrapped, IIRC because they had bugs in the software being used twenty years ago.
“What happens when signals from different cars collide?”
Automotive radar systems typically use Frequency Modulated Continuous Wave (FM-CW), and as mentioned have frequency offsets inside the band and filters to ignore returns that do not match. The radar is in 77GHz range. microwaves at this frequency are extremely directional, there is high insertion loss (attenuation of signal through the air), and the receiver arrays have highly directional antennas.
I’m always amused how naive the questions Philip asks. I can’t believe you don’t know the answer!
The reason why no more train lines are automatic, is that that would mean firing all those railworkers!
In France, the only automatic lines are a few anecdotal NEW lines: between the Orly airport and the RER connection, and the metro line 14 in Paris.
Line 14 has a big inconvenient: it’s never on strike!
As someone who works with automated cargo movements at the Port of Los Angeles (via rail and crane), let me say that it all sounds good on paper…until something goes wrong. And something goes wrong everyday.
To have this type of automation you need thousands of monitored sensors to prevent accidents. And accidents (and countless delays) do routinely happen in spite of these sensors. In the end it’s far more cost effective to have an actual human at the controls…as long as they don’t fall asleep.
I love that automated line 14 in Paris. Always on time, in a city forever in Strike.
When I was in College, 15 years ago, we already knew how to build automated train/plane. However in countries where unemployment is sky rocking this is not welcome.
The history of train crashes, control systems, and safety is very long so I’ll merely recommend learning that history. There are good reasons why the only fully automatic systems are isolated systems sealed away in tubes or elevated away from the complexity of interconnected ground level systems.
The difference between US FRA standards and European standards is highlighted by the difference in crash outcomes. In July this year an 8 car train crashed in Spain in similar circumstances. Out of 218 passengers, 79 died and nearly 100 were hospitalized. In this crash, a 7 car train had 4 fatalities and 70 injuries out of 170 passengers. The US crash safety requirements are estimated to double or triple the cost of US passenger trains when compared with European standards. It shows in the crash results.
There are major differences in attitudes toward safety, risk, etc. The standards in both cases reflect the balance reached by political and social methods. The overall system impact is highly complex. Safer trains are higher cost, which shifts more transportation onto cars, which causes more ripple effects including higher death and injury rates (although commuting accident rates are well below the average for automobiles).
rjh – while your example is interesting, it’s disingenuous to imply that the US standards are inherently better based on these two examples.
For one, just to take a broader sample of cases, looking at French TGV accidents (http://en.wikipedia.org/wiki/TGV_accidents) essentially disproves your theory, many of which were at significantly higher speeds than the Metro North accident yet had much lower injury and death counts.
Second, the example you cite in Spain was at a much higher speed (120 mph, vs 80 mph) – that speed difference means that the kinetic energy is more than double in the Spain incident. The fact that there were more injuries and deaths in the Spain accident is understandable based on that alone; trying to pin it on the difference in standards is plainly illegitimate.
In this case, crash could have been prevented by a device that should cost less than $200 dollars in hardware. A GPS and a relay and a few lines of code:
if speed over 30mph by x miles from z, activate relay, hit the brakes as hard as needed to slow down below 30mph
To answer your question Phil, the trains are not self driving because of unions.
Andy…that’s simply not true. As I stated above I work with automated cargo movements (mostly involving rail) in the Los Angeles harbor. This is a highly unionized industry and automation has been introduced with the help & the blessings of the local unions.