Cookie consent

We use cookies to improve your experience on this website. By continuing to browse our site you agree to our use of cookies. Tell me more | Cookie preferences

Man vs. machine: Who should be at the wheel?

Do you have what it takes to drive this?

    Just Watched

    Do you have what it takes to drive this?

Do you have what it takes to drive this? 02:34

Story highlights

  • The deadly New York train wreck raises a larger question about safety innovations
  • When should safety factors like piloting, steering and braking be handed over to new tech?
  • Researchers say new tech also should maximize the benefits of good human decisions
  • An airline pilot voices misgivings over the prospect of "refuse-to-crash" automation in planes

In the early-morning stillness on the Sunday after Thanksgiving, William Rockefeller was manning the controls of a half full Metro-North train out of Poughkeepsie.

The Hudson line train was hurtling along at 82 miles an hour, not within the speed limit of 30 mph, as it approached a sharp bend in the tracks in the Bronx. The cars tumbled off the track, killing four passengers and injuring dozens of others. The lead car came to rest on its side inches from water at the intersection of the Hudson and Harlem rivers.

Rockefeller's lawyer, Jeffrey Chartier, and his union representatives say the train's hypnotic motion may have caused him to nod off -- a case of what the lawyer termed "highway hypnosis."

The engineer told investigators that moments before the derailment he was "going along and I'm in a daze. I don't know what happened," one law enforcement official said.

"I think we're really looking at a sad, really basic kind of inattentiveness," says Steven Harrod, a University of Dayton professor and expert on railway operations. "Nothing fancy: No alcohol, no drugs, maybe not even a cell phone. Just plain vanilla inattentiveness."

The deadly New York train wreck raises a larger question important to millions of U.S. travelers who rely on trains, planes and cars: When should key safety factors like piloting, steering, and braking, be handed over to emerging technology? At what point should control remain in human hands? Self-driving cars and sophisticated airline auto-pilot systems are already here. How safe is that technology? And on trains, should auto-braking technology be mandatory? As more and more Americans criss-cross the nation on land and in the sky, what does the future hold for automated safety technology?

Investigating the NYC train derailment

    Just Watched

    Investigating the NYC train derailment

Investigating the NYC train derailment 02:52
PLAY VIDEO
Who's driving your train?

    Just Watched

    Who's driving your train?

Who's driving your train? 02:12
PLAY VIDEO
What's next for NTSB investigation?

    Just Watched

    What's next for NTSB investigation?

What's next for NTSB investigation? 05:04
PLAY VIDEO
Was sleep a factor in train derailment?

    Just Watched

    Was sleep a factor in train derailment?

Was sleep a factor in train derailment? 04:04
PLAY VIDEO

The Metro-North tragedy, however, poses a bigger question that also includes planes and ground vehicles: Is the time coming soon when it will be safer for humans to take a backseat and let computers take the wheel?

The short answer, say experts, is yes.

Still, researchers are quick to add that it's important to create technology that maximizes the benefits of good decisions by pilots, drivers and engineers, while minimizing the threats that humans will eventually be distracted or get tired.

Railroad safety officials have long pushed for a system known as positive train control technology, which combines GPS, wireless radio and computers to monitor trains and stop them from colliding, derailing or speeding. But the railroad industry has opposed it because of the high cost and technological issues.

Agency issues emergency safety order after New York train derailment

"If PTC had been in place, this accident would not have happened," Harrod tells CNN. "That's pretty clear at this point. It's non-negotiable. It would not have happened."

Positive train control was designed to prevent the human errors behind roughly 40% of train accidents, rail safety experts say. The transportation safety board has urged railroads to install PTC in some form for decades. The technology already is used by many European rail systems and several American railroads.

In response to a head-on collision that killed 25 people near Los Angeles in 2005, Congress in 2008 ordered the nation's railroads to adopt PTC by December 2015.

"It's a technology that's being deployed in the military as part of the concept called network-centric warfare," says Steven Ditmeyer, a former Federal Railroad Administration official who teaches at Michigan State University. "It's the same technology that is being implemented in aviation as Next Generation Air Traffic Control. It's using digital data, communications, GPS and computers in the vehicles and control centers."

Driverless car now legal in California

    Just Watched

    Driverless car now legal in California

Driverless car now legal in California 01:02
PLAY VIDEO
Google's self driving Prius

    Just Watched

    Google's self driving Prius

Google's self driving Prius 02:14
PLAY VIDEO

On the Metro-North train in Sunday's accident, PTC would not have allowed the train to exceed the speed limit, rail experts say. It would have slowed and eventually stopped the train it if the engineer became unresponsive.

Train crash victim was 'pure goodness'

On Tuesday, National Transportation Safety Board investigator Earl Weener said PTC may have prevented the Metro-North derailment.

"For more than 20 years the NTSB has recommended the implementation of PTC technology," Weener said. "These systems provide a safety redundancy by slowing or stopping the train that isn't being operated in accordance to signals, speed limits or other operating rules. PTC is proven technology that can prevent train-to-train collisions, over-speed derailments and incursions into work zones."

Metropolitan Transportation Authority spokesman Salvatore Arena said the agency began work to install PTC on the Long Island Rail Road and Metro-North Railroad in 2009, budgeting nearly $600 million for PTC installation, including $428 million last month for a system integrator. Full implementation is estimated at $900 million.

But, in a statement, Arena said implementing PTC by the 2015 deadline will be difficult because much of the technology is still under development, untested and unproven on commuter railroads the size of Metro-North and LIRR.

"The MTA will continue its efforts to install PTC as quickly as possible and will continue to make all prudent and necessary investments to keep its network safe," Arena said.

Sunday's crash has led to calls for the government to stick to the 2015 deadline.

In a letter last week to Sen. John Rockefeller, D-West Virginia, chairman of the Senate commerce, science and transportation committee, Sen. Dianne Feinstein, D-California, wrote: "In just the last ten years, the NTSB has completed 26 investigations of train accidents in the United States that could have been prevented by PTC. These accidents claimed 65 lives and injured more than 1,000 people."

But Harrod and other experts doubt that the government will stand firm by the deadline.

"They stalled so long that they lost the window of opportunity to get it done on time," Harrod says. "I don't care had badly people scream and yell, it aint going to happen now. It does take time to implement."

Cars that 'platoon' with others

It's a crazy time in the car business, as the industry and regulators scramble to keep up with lightning-fast technology -- including automated driving systems.

"Concerns about what happens when computers take over vehicles are real and appropriate," says Sven Beiker, executive director of Stanford University's Center for Automotive Research. Biggest worry: how automated vehicles will interact on the road with cars completely controlled by humans. Self-driving cars are already here -- recently legalized by the state of California.

In addition, several, automakers, such as Ford, BMW and Volvo, already offer automated forward-collision avoidance systems, which automatically apply brakes or brake assistance when sensors and computers detect that a collision is unavoidable or likely.

By the end of the decade, our vehicles may be able to use cellular networks to "talk to each other," allowing them to coordinate movements.

Cars on similar routes will "platoon" in tightly organized bunches as they roll across streets and highways.

During this time we also may see more cars equipped with "traffic jam assist," says Beiker. "Just like a school of fish -- when you're in a traffic jam your car will be able to keep its distance from the vehicle in front of you and to the left and right of you," Beiker says.

In 10 years, says Beiker, it's plausible we'll see vehicles that will allow drivers to "sit back and monitor" their progress while cruising on long stretches of highway without much traffic. The vehicle would tell the driver when it's time to reassume control for whatever reason.

Beiker, whose facility works with Nissan, Volkswagen, Ford and other carmakers, is busy with a project that involves "wiring up" race drivers to study how they control their cars. If all goes well, researchers will learn how to translate the skills of top drivers into terms that can be understood by computers.

The push toward automation really isn't about convenience. "Safety is why were doing this," says Beiker, "basically to reduce human error." Beiker says automation will cut the number of accidents caused by distracted or tired drivers.

'Pilots like to have control over the end product'

In theory, autopilot safety systems aboard airliners might have prevented the 9/11 terrorist attacks.

The idea is called "refuse-to-crash" automation. Supporters say the technology would have made it virtually impossible for the planes to be flown into the Pentagon or the twin towers at the World Trade Center.

Here's how experts say it would work:

Airliners already have the ability to fly automatically via computer-driven auto-pilot programs that follow precise data from GPS satellites. "Refuse-to-crash" would build on that model. Locations of all buildings, mountains and other potential crash threats would be programmed into auto-pilot systems. Using that geolocation data, computers would essentially lock the plane's guidance systems from being able to collide with potential targets.

Commercial airline pilot Justin Schlechter says that kind of system wouldn't make him feel any safer.

"I just don't feel that's the kind of equipment that's essential in the cockpit when I'm flying," he said. "Pilots like to have control over the end product. I wouldn't want the system to lock onto a building and make a hard maneuver away from it that I'm not ready for."

One of the biggest challenges surrounding highly automated pilot systems "is the crew not fully understanding what the automated system is doing," said R. John Hansman, who specializes in humans and automation as director of MIT's International Center for Air Transportation.

Another hurdle, said Hansman, includes the price tag of retrofitting technology across the entire fleet of U.S. airliners. It would be enormous. Hansman estimates somewhere in the tens of billions of dollars. Then there's the federal certification process. The Federal Aviation Administration "would have to be convinced that the benefits of the technology are worth the cost and that you don't create any unexpected hazards," said Hansman.

Sometimes autopilot technology isn't enough to prevent a crash. A glaring example is Air France Flight 447 -- one of the deadliest airline disasters of the century. The Airbus A330 en route from Brazil to Paris flew into a storm and plummeted into the South Atlantic, killing 228 passengers and crew.

Flight 447 was outfitted with an automatic system designed to keep the plane from stalling. A stall occurs when there's not enough airflow moving across the wings to keep the plane flying. This could happen when the plane is flying too slowly or its nose is pointed too high.

It was a stall that eventually doomed Flight 447 after a tragic series of events, according to the final investigation report. It started when the icy storm froze the plane's air-speed sensors, which automatically turned off the aircraft's anti-stall system. Then the plane went into a stall, which triggered a warning that it was in danger. The investigation report said pilots may have "embraced the common belief that the aeroplane could not stall, and in this context a stall warning was inconsistent."

Read the report (PDF)

"The crew never understood they were in a stall situation and therefore never undertook any recovery maneuvers," chief investigator Alain Bouillard told reporters when he introduced the report.

The use of automatic systems on planes has improved safety overall, Bouillard explained, but "when it comes down to it, safety will always be based on the capacity of the pilots and the signals which they are given, which they have to understand and react to."

Pilots say maximum training on automated systems is imperative. Air France 447 reminded pilots that it's critical to be aware when automated systems are engaged and how that will effect the plane during various situations.

Another danger posed by automated systems is complacency. "You can't afford to be complacent," said Schlechter. "If you're thinking, 'Oh, if I don't fly this plane as well as I should, the stall-protection system will always save me -- and I'll be all right,' that's a very dangerous way of thinking."

So, bottom line: The technology to prevent catastrophic crashes via autopilot systems is here. But even if the most sophisticated and complicated autopilots perform as advertised, big questions remain: Do they create more problems then they're worth? Would they be too expensive?