Roboruku control software that automatically determines its configuration

If you have ever written software to control a roboruk or even any robot-like mechanism with more than three degrees of freedom, then you know that programming of the movement itself takes most of the time. Now imagine that you built a roboruku, connecting the links and drives as you like, and the software determined the lengths of the links, their location, the angles of rotation, and so on.

That is exactly what Columbia Research did when writing software that learns how to operate a roboruki that it controls , without having any initial understanding of the physics, geometry, or dynamic characteristics of engines. Immediately after the first launch, he does not know what form of roboruk, how its engines work, how and what each of them affects. For about a day and a half, making at first cautious movements that seem erratic, and receiving feedback on what is happening, the software, using in-depth training, configures the simulator built into it in such a way that the virtual robot box is as close to physical as possible.
')
Lipson project manager and his graduate student Robert Kwiatkowski are experimenting with an articulated roboruki that has four degrees of freedom. At the beginning of the software training, the model was inaccurate, and the order of connecting the links was not determined. After 35 hours of training, the model corresponded to reality with an accuracy of no less than four centimeters. Then the software forced the robot hands to take and move objects, while there was an additional recalibration after each move according to a model formed only by self-study.

To determine whether such software can detect damage to a roboruki, a similar but specially deformed was printed instead of one of the parts. The software determined what had changed and made appropriate changes to the model. Soon it performed the tasks of lifting and moving objects only slightly worse.

Since the internal representation of software about a roboruki device is not static, this allows it not only to perform better and better, but also to adapt to damage and changes in structures. Also, everything will continue to work with the gradual deterioration of parts and their replacement with slightly different in shape.

Of course, roboruka, managed by such software, will not very soon become as accurate as Dexter - roboruka with traditional software, which presented its developers with a prize at the 2018 Hackaday competition , but it’s interesting to see how it works: