Felix Ruppert and Alexander Badri-Sprowitz at the Max Planck Institute for Intelligent Systems in Stuttgart, Germany, designed a half-metre-high robot called Morti and gave it the ability to teach itself how to walk, rather than to perform a pre-programmed step. The four-legged robot took only an hour to learn how to walk steadily, roughly the same amount of time as newborn horses need. And it's the first time that a machine learning technique has been so successfully applied to four-legged robots.
Morti is controlled by an artificial intelligence (AI) algorithm (算法) that doesn't have much information about the robot's legs, such as the exact shape of each element. "The AI, working like the central nervous system, gives walking instructions for Morti to follow. It then adjusts (调整) them based on readings from foot sensors that signal when the robot falls and loses contact with the ground. Initially (最初), Morti falls down, but after about an hour the AI finds the best way to make it walk," said Ruppert.
Because the AI learns rather than computing details of each leg's movement in advance, which can use a lot of energy, Morti walks using 42 percent less energy than when it first starts at the end of an hour-long learning process. Morti's process copies the way baby animals learn to move, as they also find the most efficient way to use their muscles by trying and initially tripping.
Dhireesha Kudithipudi at The University of Texas at San Antonio said that AI robots can often learn a specific task very well but can't readjust when the environment changes and that Morti's design, which relies on continually adjusting the robot's movements, may perform better in that regard. Ruppert said he and the team are working on adding more sensors and range of motion to Morti to make it a more animal-like robot.
