It’s vital for gymnasts, but learning how to fall is important even for robots where it concerns the possible damages they can take from a mere tumble. Technology is taking huge leaps in the field of robotics. More and more possibilities are being designed and developed to a point where their use becomes a certainty in the future.
However, the field is also rather costly, and the precious parts should be well protected.
Developers at the Georgia Institute of Technology have, thus, created an algorithm that will tell a robot how to minimize damage taken from a slight stumble. It could potentially alleviate the harmful impact and certainly lower the cost of repairs.
After all, every part is precious, and robots do not yet possess the ability to judge when someone will knock into them. Their sensors can only do so much.
The research was inspired by professor Karen Liu’s study on how cats adjust their position in the middle of a fall. It was observed that the most significant aspect is the angle of the actual landing. By taking a couple of simple factors into consideration, it was apparent that a robot can certainly calculate the same patterns, if only it had a tool to optimize the movements.
Liu along with Sehoon Ha, a postdoctoral associate, have created the algorithm that would perfectly teach robots how to fall gracefully.
The tool they designed can tell aid the humanoid bot in determining the exact number of contacts with the ground. They range from the numbers of shoves it receives, the order, timing and position of those contacts in order to create the best movement to minimize the damage.
They tested their findings by essentially shoving a robot at varied speeds. Ranging from a gentle nudge to a violent push, the robot was successfully able to take a step forward to regain balance, or actually perform a sequence of rolling motions. This decreased the severity of its impact to the ground.
According to Liu, they had to take into consideration the hardware constraints that do not allow a robot to be as swift as a cat, for example. Instead, they suggested a sequence that can essentially help the robot slow down in order to protect their own bodies. There won’t always be a human there to catch them.
The two researchers state that a robot can now fall as safely as possible. Be it a slight or drastic tumble, they can minimize the damage and lower the number of costly incidents. As stated by Liu, they will soon be working on another step, by providing a sort of nervous system for robots.
Image source: escapistmagazine.com