This story is over 5 years old.

Why Don’t We Have Driverless Trains Yet?

CES 2015 was all about the driverless car, but autonomous rail makes more sense.
The Docklands Light Railway in London. Image: ​Robert Pittman/Flickr

The most futuristic trend to emerge from this year's annual gadget parade known as the ​Consumer Electronics Show had to be driverless cars. Mercedes, Audi, BMW, and Ford all showed off new concepts that would turn drivers into mere passengers on smooth, safe, computer-controlled rides. If cars will soon be able to go places without someone behind the wheel, why can'​t trains?

To the casual observer, it seems like trains are a far more logical mode of transportation to turn completely autonomous than cars. With few exceptions, trains are designed to travel in a single direction at a time along fixed tracks, eliminating many of the variables of the open road. The biggest problem appears to be a binary one: if something happens to the track in front of a train—someone or something enters it, or it gets broken or damaged beyond use—the train needs to stop in time. If not, keep ​on rolling.


Most trains today rely on a combination of human engineers and rail operator staff to identify and act in these types of situations. But couldn't the job be done just as easily and more efficiently with a system of cameras or sensors placed strategically along tracks, coupled with some smart computers onboard the trains themselves? "It's certainly technically possible," says David B. Clarke, director of the Center for Transportation Research at the University of Tennessee. "The technology has been around for 20 to 30 years at least."

Indeed, the first completely automated subway train went into service in New York in 1961, and a crewless freight train system was tested in Canada a year later, according to a an Office of Techno​logy report circa 1976. Today, there are a surprising number of autonomous rail systems operating around the world mainly as metros, subways, or light rail, manufactured by the likes of Siem​ens in Germany, Alstom in France, and Bombar​dier in Canada.

Credit: UITP

By the end of 2013, there were 48 fully automated public metro systems in use in 32 countries, according to the International A​ssociation of Public Transport (UITP), an advocacy group that promotes public transit. That figure doesn't even include the doz​ens of other private light rail systems also found in many of the world's metropolitan centers, such as the Air​Train at New York's JFK airport and the Disneyland​ Hong Kong Metro. While some of these systems do have staff aboard the trains, they're not actually driving them, and are instead functioning mainly in support roles, such as ticketing, rider safety, or as emergency backup if the automated system fails in some way. And more are on the way, set for Aus​tralia in 2019 and London in 2022.


However, these automated metro and light rail systems are typically confined to the area of a city at most, far smaller in their service range than the major commuter and freight rail networks that span the United States and other countries. They also don't intersect with other rail networks. Both of which underlie the key factor that makes it easier to automate these systems: they're closed. That is, "access to the track is prohibited by fences, by tunnels, bridges, platform barriers or special sensor systems in the stations," explains Burkhard Stadlmann, a professor of at the U​niversity of Applied Sciences Upper Austria. "All these systems have no obstacle detection and assume a free line."

The same is true of the growing number of high-speed rail lines in Europe, Asia, and now Cali​fornia, which have elevated tracks to get them out of the way of obstacles. Not surprisingly, many of these have semi or fully autonomous controls.

Yet in the case of existing open rail networks, "you've got a lot of miles of track" that need to be monitored for obstacles, Clarke points out. Many railways around the world have already automated some of that process and are only moving further in that direction. They're increasingly installing visible light cameras, infrared cameras, and radar-type sensors at key crossings and stations, which human dispatchers monitor for signs of trouble, radioing human train engineers well in advance.

"Ultimately what you want is for it to be interpreted without a human being in the loop, through some sort of image processing algorithm" Clarke adds. He believes this too is technically possible with current technology, so why don't we do it?

As with many things in life, the answer comes down to money. "The economic profit for sparing the driver is not that high if you have a train with 500 or 1000 passengers and one driver," Burkhard says. He and University of Austria professors Oliver Gerbauer and Wolfgang Pree have conc​eived a more affordable autonomous system that relies on smaller, lighter trains traveling along open rail in shorter intervals.

Also, many train workers are unionized, which means even when new technologies become available, the tendency is to adapt it to better support their jobs, not outsource them. "Labor agreements are not encouraging for automating the train," is how Clarke puts it. And then there's the matter of public perception of driverless trains, which Clarke notes is not often favorable. Sadly, it seems that completely automated open rail systems—ones without any human crew aboard—aren't likely for the foreseeable future.

Still, the US government favors increased automation on the basis of safety. Thanks to a 2008 law, the majority of the US rail industry is supposed to implement a semi-automated system called Positive ​Train Control by the end of this year. The sy​stem is combination of sensors along tracks and inside locomotives that can automatically change the speed of a train at any time without human intervention. Officials say it could have prevented the fatal Metro-Nort​h derailment in 2013, which occurred when a fatigued human engineer rounded a curve at three times the speed limit. "The fact is a certain amount of accidents are caused by improper action by the operator," Clarke says. "Technology will hopefully greatly reduce that, but it's not 100 percent reliable either."