The real scandal of the recent train crash in Hoboken is not being discussed: the super-expensive PTC technology that was supposed to prevent it is obsolete and ineffective. A bunch of college kids using off-the shelf technology could build a far superior system in months for next to nothing. The problem isn't politicians and bureaucrats reluctant to spend the billions required to implement PTC. The problem is that PTC is built on computing technology that belongs in a museum, not supposedly protecting our lives.
The Cause of the Crash
What caused the crash? It's known that the train was going too fast and failed to brake, but as of this writing, more details are not known. Much of what you read about the crash provides some details of someone's experience, but shifts into blame mode. The current governor of New Jersey, Chris Christie, is a favorite target, but various parts of government and government bureaucracy such as NJ Transit also get pot-shots. This first-person account that was published is typical:
After you're done wading through all the accusations, most writers get around to blaming the crash on the failure to implement positive train control (PTC), the government-mandated system that is supposed to prevent crashes like this.
It seems crazy: why are these people dragging their feet implementing a safety system, when passenger lives are at stake??
Failure to implement PTC
PTC has been widely implemented. But not completely. Union Pacific, for example, has already spent about $2 billion implementing it, but estimates the total cost at about $2.9 billion, so they're not done yet. That gives us our first reason why PTC isn't universally implemented, and wasn't implemented on the tracks near Hoboken: It's wildly expensive!!
Do you think PTC was designed by a bunch of modern, agile computer and software people? Or do you think it was designed by a bunch of lawyers and bureaucrats and regulators many years ago, and essentially unchanged today? Think hard, now!! Take a quick look at what goes on in PTC for example here and you'll get a feeling for it. It's wildly complicated!! Not to mention totally old-fashioned, and designed with paleolithic computing technology.
Instead of thinking, "how can I use what's out there, make a few changes, and get going with this," the people were thinking all railroad all the time. In spite of strong overlap with other systems, like trucking, PTC "had" to be totally unique to railroads. That leads to a huge pile of PTC-specific technology that is no better than what was generally available many years ago, but unique -- and therefore incredibly expensive and time-consuming to design, build and support. As a result, there isn't exactly a robust, competitive marketplace for PTC. Thus the time and expense. Thus the fact that it hasn't been deployed. Q.E.D.
PTS and Military Procurement
Clearly the people who designed and mandated PTC came from the same school of thought that dominates military procurement. Remember $640 toilet seats? That wasn't a myth. There's lots more where that came from. Here's a list from a book on the subject:
The point is whenever there is lots of money made from government procurement, companies and officials will collude to make all sorts of special requirements that standard equipment can't meet, so that the companies that play the game can make the "special" versions of whatever and have big revenues with bigger profits.
Sadly, that's exactly what's going on with PTC. It's a bad system. It's many generations obsolete. It doesn't work. And it's hundreds of times more expensive than it needs to be.
The alternatives to PTC
What could be done instead? I briefly reviewed one approach in a post about the 2015 Amtrak crash in Philadelphia. The key idea is to stop following the military procurement model, and instead use modern, off-the-shelf technology that is updated and refreshed regularly, just like you update your smartphone.
There are existing systems built for the trucking industry that could be adapted for trains. Here's a proven one that's hardware-based:
Here's another one that makes smart use of cloud technology,
which means that all the super-custom requirements for recorders that can withstand having boiling oil poured on them can be thrown out. Just as well, a system could be built completely from off-the-shelf components, making good use of powerful networking technology that can be installed on the miles of track on which it's not already available for a fraction of the cost of the super-custom, super-expensive PTC alternative.
A bureaucrat who was all wrapped up in the existing regulations could come up with all sorts of objections to this approach. They would all be bogus. Do you use a 20 year old computer? No vendor would support it. But in the isolated world of PTC, decades-old equipment is par for the course. One of the event recorders in the Hoboken train was installed in 1995. It didn't work. Not because it crashed, but it failed at some unknown time earlier. And no one knew.
By comparison, I have home security cameras from a Google company called Nest. They only cost a couple hundred dollars each. You connect them to the internet and they just work. Here's the key: when one stops working or communicating for any reason, I get an immediate notification! So I can do something about it. This is an inexpensive consumer device. A train safety device that is probably hundreds of times more expensive, using obsolete technology? Not available. This is one illustration among many. We should just get over it and vote for safety and effectiveness. We should throw out PTC and everything associated with it, and get something modern that actually works.
And, not exactly by the way, save lives.
Postscript
PTC is yet another example of the "what not how" style of government regulation that fails everywhere it's applied. It fails in train safety, in computer security and everywhere else it is applied.
Comments