A researcher in a robotics lab in Slovenia has reportedly been using humans as punching bags to find out the limits on how hard and fast a robot can move if it collides with people.
Borut Povse persuaded six male colleagues to let a modified production-line robot hit their arms up to 18 times with different amounts of force using either a sharp or blunt tool, New Scientist magazine reported.
The subjects could rate their pain on a scale ranging from mild to unbearable.
Povse reportedly presented a paper on his experiment at an IEEE conference being held in Istanbul this week.
He Blinded Them With Science
For the experiment, Povse and his colleagues modified a production-line robot from Epson that’s normally used for assembling coffee vending machines. They programmed the robot’s arm to hit a volunteer’s arm a total of 18 times with varying amounts of force, using either a blunt or a sharp tool.
The volunteers were instructed to rate the resulting pain in one of four categories: mild, moderate, horrible or unbearable. Povse, who used himself as a guinea pig before turning to his colleagues, reportedly said most of them rated the pain from being hit as being mild to moderate.
The team will continue their tests using an artificial human arm to see the effects of more severe collisions.
The idea is to find out the maximum speed at which a robot can move when it senses a human nearby to avoid hurting the human in the event of a collision.
Povse read a report on his work at the IEEE’s Systems, Man and Cybernetics (SMC) 2010 conference held earlier this week in Istanbul.
The IEEE did not respond to requests for comment by press time.
Domo Arigato, Mr. Roboto, May I Have Another?
Controlling a robot can be divided into three stages — perception, processing and action. The robot has to perceive what’s in its vicinity through sensors; it has to process the information it receives to figure out what to do next; and then it takes action.
The processing phase can be very complex indeed. At the basic level, it may translate raw information from the sensors directly into commands to the actuators. Actuators serve as the muscles of a robot. The most widely used are electric motors and linear actuators.
For more sophisticated tasks, the processing may involve building objects or reasoning with a quasi-cognitive model. Such models represent the robot, the world, and how the two interact. They may use pattern recognition, computer vision, mapping techniques, motion planning and other artificial intelligence techniques so the robot can figure out how to act.
What happens if the robot goes berserk? In fact, can a robot go berserk?
In a manner of speaking, yes. The most common cause of robot failure is a problem with its control software, not its hardware, according to a Rice University lesson on robot design. For example, if a robot hits something but its touch sensor isn’t triggered, the robot will get “mentally stuck” because its program will insist it keep on trying to move ahead and doesn’t recognize that it can’t move.
What will happen if the software controlling how hard and fast a robot can move when it is close to a human goes wonky? Perhaps Povse will look into testing that next.
Must Tests Hurt?
Is there a way, perhaps, to test the limits on a robot’s speed and power in the proximity of humans without having to beat people up?
Perhaps not. “It sounds obvious, but some people are more sensitive to pain than others, and I’m not sure there are any guidelines that indicate acceptable limits on the speed of a robot around humans,” Carl Howe, director, anywhere consumer research at the Yankee Group, told TechNewsWorld. “They’re probably right to conduct these tests and try to figure out the limits,” he added.
Should the force be such that test subjects can rate their pain from participation as horrible or unbearable, or is that taking the test too far?
“Why did the IEEE even let this paper be presented?” asked Jim McGregor, chief technology strategist at In-Stat. “I think this proves that not only the participants but also the experimenter should be candidates for next year’s Darwin Awards,” he told TechNewsWorld.
The Darwin Awards is a satirical honor set up to “salute the improvement of the human genome by honoring those who accidentally remove themselves from it.”