Thursday, September 6, 2012

Self-Defense for the Self-Driving Car

In a few years? time, once we get used to the idea of Google?s self-driving cars, it?s conceivable that autonomous trucks will take over the delivery industry. But while a driverless vehicle might bring with it big advantages, such as being less prone to accidents than a big rig with a road-weary driver behind the wheel, a question remains: How will driverless cars defend themselves?

David Mascarenas, a researcher who studies cyber-physical systems at Los Alamos National Lab, says that as more robots venture out on their own, their creators are already struggling with how to protect them. During an exercise in Narragansett Bay, R.I., this summer, the U.S. Navy had to warn off at least one individual attempting to grab a miniature robot sub. In June, Cockrell School of Engineering assistant professor Todd Humphreys showed how drones could be decoyed into landing in the wrong place by deceiving their GPS. Mascarenas?s own involvement started with protecting expensive structural sensors now being placed on bridges to monitor their condition.

And in a more futuristic threat, Mascarenas says that thieves could see vehicles with no human drivers as defenseless targets. So now, before this problem arises on the road, he?s working in the lab on ways to make sure would-be robbers get the bad end of an encounter with unmanned trucks.

His work starts with teaching robots defensive driving techniques. Consider the Precision Immobilization Technique, or PIT, the standard way cops stop a fleeing vehicle (see a good example in the video below). The pursuer draws up and gives the rear end of the target a sideways nudge, causing the target vehicle to slew sideways, lose control, and stop. Mascarenas thought unmanned vehicles would be easy prey for robbers using this sort of tactic to stop trucks and steal their cargo. So he developed techniques for the robots to escape, evade, and recover from the PIT maneuver.

Mascarenas, who was involved with racing cars as a hobby, then tested the theory with experiments using scale-model vehicles?cars about 2-1/2 feet long, fitted with sensors including a camera, compass, and LIDAR, a laser radar capable of detecting the precise range and direction of other vehicles. They also had plenty of padding for the inevitable crashes.

This field test pitted a vehicle operated by a human with remote control against an unmanned vehicle. The unmanned vehicles blocked the PIT maneuver much as human driver would, by keeping directly in front of the pursuer so it could not come up alongside it. The results confirmed that the robot vehicles could be successfully programmed to execute a PIT maneuver as well as to perform actions to resist being PIT-maneuvered by another vehicle.

Defensive-driving techniques are just one part of Mascarenas?s larger effort to prevent criminals from exploiting unmanned systems. For example, he says, robots? limited navigation skills may make them vulnerable to crime. Consider a robot bellhop programmed to take your things up to your hotel room. It would be fast and efficient?but what if the robot were decoyed into taking your luggage to the wrong room, where a thief was waiting? Robots delivering drugs at a hospital could suffer the same fate if they relied upon a simple navigation system, such as following white lines or other markings on the floor, making them easy to deceive. Robots will need multiple independent navigation systems to defend against this type of scam.

One of Mascarenas?s projects makes robots less predictable and reduces their vulnerability to ambush. If you know a driverless delivery truck always goes down the same deserted street at 6:14 am, you can get there first. Mascarenas addressed this using a technique known as info-gap decision theory. This allows the robot to weigh the risk of any particular route with the possible benefits. This would apply equally to a robot submarine looking at the best way to gather oceanography data while avoiding fishing nets, or a pizza-carrying droid finding the shortest way through city streets. Crucially, the process is unpredictable: The machine will not always take the same route twice, and would-be ambushers can?t anticipate where it will be.

What about physical defences? Protection for robots might involve weaponry?in the military sphere at least?and a robust set of rules for dealing with human beings. U.S. Army work on unmanned convoys and ground robots has focused on nonlethal deterrents. These rely on laser dazzlers and pyrotechnic flash-bangs or even a skunk-like malodorant dispenser to keep people from hijacking or pilfering their robots. Such vehicles might also have lethal armament that could only be used by a remote human operator.

But, Mascarenas says, weapons are no defense against hackers who target the software and take over the robot itself. Many developers take advantage of the Robot Operating System, which is a standard piece of open-source software. But as Mascarenas points out, the Robot Operating System was originally intended for research purposes, and security was not a high priority when it was built. A few weeks ago, Mascarenas set up some robot vehicles as a honeypot to attract hackers at the DEF CON 20 Hacking Conference. His goal was to combine the honeypot with low-cost crowdsourcing techniques to find security issues associated with robots and solve them, preferably before they can be exploited by criminals.

Mascarenas compares the situation to the early days of computing, when developers were excited about new possibilities and gave little thought to the risks of hacking, , and viruses. That?s why he is trying to anticipate the problems before they occur. Mascarenas says that although there will always be a risk of criminals hijacking robots, the security problem should not be a showstopper. "It?s just something that has to be managed," he says, "And we?ll manage it."

This managing might mean better firewalls, antivirus software, smart navigation systems, and encrypted communications on unmanned vehicles. It might also mean robot drivers who are more than a match for any human who thinks he can push them off the road.

Source: http://www.popularmechanics.com/technology/military/robots/self-defense-for-the-self-driving-car-12410682?src=rss

jackpot winning numbers mega millions megamillions drawing olbermann mega millions march 30 lucky numbers odds of winning mega millions

No comments:

Post a Comment