|
Determination of Q-function optimum grid applied on active magnetic bearing control task
Březina, Tomáš ; Krejsa, Jiří
AMB control task can be solved using reinforcement learning based method called Q learning. However there are certain issues remaining to solve, mainly the convergence speed. Two-phase Q learning can be used to speed up the learning process. When table is used as Q function approximation the learning speed and precision of found controllers depend highly on the Q function table grid. The paper is denoted to determination of optimum grid with respect to the properties of controllers found by given method.
|
| |
|
Stochastic policy in Q-lerning used for control of AMB
Březina, Tomáš ; Krejsa, Jiří ; Věchet, S.
A great intention is lately focused on Reinforcement Learning (RL) methods. The article is focused on improving model free RL method known as Q-learning used on active magnetic bearing model. Stochastic strategy and adaptive integration step increased the speed of learning approximately hundred times. Impossibility of using proposed improvement online is the only drawback, however it might be used for pretraining and further fined online.
|
|
Q-learning used for control of AMB: reduced state definition
Březina, Tomáš ; Krejsa, Jiří
Previous work showed that stochastic strategy improved model free RL method known as Q-learning used on active magnetic bearing (AMB) model. So far the position, velocity and acceleration were used to describe the state of the system. This paper shows simplified version of controller which uses reduced state definition - position and velocity only. Furthermore the controlled initial conditions domain and its development during learning are shown.
|
| |
| |
| |
| |
| |
| |