When robots fall, they fall hard, plummeting to ground in an awkward heap. It’s comedic gold until you examine the impact on the expensive and oftentimes unique robot. To prevent these high-G impacts, a team of researchers from Georgia Tech are developing an anti-falling algorithm that’ll help a robot break its fall.
The algorithm, developed by Georgia Tech professor Karen Liu and her then graduate student Sehoon Ha, allows robots to manipulate their limbs and contort their bodies in order to maximize the number of potential points that will come in contact with the ground. This strategy allows the robot to distribute the force of the fall over its body and not concentrate it on a single point. As a result, the robot will be able to withstand even a hard fall with less damage to its intricate parts.
In a series of falling tests with a small BioloidGP robot, the Georgia Tech team demonstrated the effectiveness of the algorithm. Based on measured falls, the algorithm was responsible for a considerable reduction in force one the robot, in one instance reducing the impact from 8.04G without. to 5.5G. During the heaviest push that caused a fall, the algorithm-guided robot was able to stick out a limb and do a summersault instead of crashing to the ground.
Besides preventing injury to robots, Liu also is exploring ways to prevent robots from hurting humans when they fall over. This is a much more complex problem that requires a robot to detect a nearby person while falling and then change its direction mid-fall. In their testing, the Georgia Tech team has equipped a robot with an accelerometer in its head and a motion capture camera to function as its eyes.
It’s not as easy task. Head researcher Liu points out that people manage to avoid contact in falls because of our nervous system, which reacts automatically. Liu hopes to create a similar system for robots. “That’s why we have reflexes,” Liu says. “We are thinking of building something like a nervous system for robots.”
