2023 Author: Bryan Walter | [email protected]. Last modified: 2023-05-21 22:24
After a jump, American and Italian engineers taught the one-legged robot Salto-1P to land at the planned point with great accuracy and stop without falling. Previously, the robot could only jump without stopping. The developers presented an article about the new abilities of the robot at the ICRA 2020 conference (the preprint is available on the university website), and also talked about them in an interview for the IEEE Spectrum.
Salto is a one-legged jumping robot, the first version of which was introduced by American engineers in 2016 and then modified. The modern version of the Salto-1P controls its jumps and landing position using a three-section leg with an electric motor at the base, a flywheel on the side and two side screws. This allowed the robot to accurately calculate its jumps and jump over objects with different heights. But before, to maintain stability, he had to constantly jump, and to stop him either caught with his hands in the air, or sent to the net.
In the new version of the control algorithm, engineers led by Ronald Fearing from the University of California at Berkeley significantly improved the robot's landing accuracy and taught it to stop after that, rather than push off and jump further. Major improvements relate to angular momentum control. To do this, the robot accurately calculates and adjusts its angle of inclination by the flywheel to the moment of landing so that after touching the floor, the moment of momentum is small and sufficient for the robot to move to a vertical position. If the angle at the moment of touching the floor is less than the required one, that is, the robot is closer to the vertical position, the moment of momentum will be too large for the flywheel to be able to extinguish it, and the robot will spin and fall further along the trajectory.
In addition, the developers have improved the calculations of the level of springiness of the leg at the time of landing, as well as the calculation of the angle at the beginning of the jump. The last element is important, because after taking off, the robot can only control its rotation, and it is already unable to change the trajectory of the center of mass, so the landing point depends on the initial phase.
Improvements in the calculations of the initial phase of the jump made it possible to reduce the standard deviation from the calculated landing point from 9.2 to 1.6 centimeters. And better control over the angle of inclination when landing allowed the robot to stop and perform several accurate jumps in a row, which would be difficult to achieve without the ability to stop.
In this work, the robot performed calculations, relying not only on its readings, but also on an accurate external motion capture system. But last year, the developers created a completely independent version of the robot and demonstrated a video in which it moves several hundred jumps in a row along the street.
