This article is cited in 1 paper

Learning systems with discrete control

K. A. Golikov

Lomonosov Moscow State University, Faculty of Mechanics and Mathematics

Abstract: The report outlines the work to create a discrete-control system learning algorithm for acting and achieving goals. Learning is based on trials and error. The entire experience of the system is stored in the Database. The algorithm is optimized by two criteria: the accuracy of achieving the goals and the maximum reduction in training time. The reduction in training time is implemented mainly by reducing the number of trials using prediction methods and interpolation by experimental data.

Keywords: positioning, learning algorithm, robot, interpolation, approximation.