RunBot, the fastest bipedal robot in the world in relation to his height, is now also climbing climbs. Image: Poramate Manoopong et al.
Reading aloud Göttingen scientists have developed a two-legged robot that can independently learn to climb a slope. This is more difficult than it appears at the first moment, since locomotion on two legs requires a complex interaction of mechanical components and motion control. For different terrain different adjustments such as the knee joints, the hip position and the position of the body's center of gravity are needed. For their experiments, the researchers extended their previously developed robot "RunBot", which holds the record in fast walking for dynamic machines, to an infrared sensor. "This sensor measures the steepness of the ramp and allows the robot to adapt its posture and gait to the obstacle in good time, " explains Florentin Wörgötter, who led the study, to Like humans, the robot responds to the climb by leaning slightly forward and taking shorter steps. The steepest ramp the robot can climb has a slope of about fifteen degrees.

However, he first has to learn this adaptation process. If the robot meets the obstacle for the first time, it falls backwards because it has not yet learned to respond to the increase with a change in its posture and stride. Within a few tries, however, the robot manages to cope with the slope just as children learn to walk by mistake. Once the robot has learned that it needs to adjust its posture and stride to cope with a climb, it can apply that principle to ramps unknown to it.

"The special feature of the robot is that the learning process enables it to master an increase, " continues Wörgötter. The rapid change of gait is made possible by the hierarchical structure of the motion control. As long as the movement process can remain unchanged, the sensors on the legs and the body control the movement of the robot. They ensure that the joints are not overstretched or that the next step is triggered as soon as the foot touches the ground. Only when an adjustment of the movement is necessary, higher hierarchical levels intervene in the control.

A video showing the learning process of the robot climbing a slope can be seen here. display

Poramate Manoopong (University of Göttingen) et al .: PLoS Computational Biology, vol. 3, p. 7, p. E134 ddp / Tobias Becker


Recommended Editor'S Choice