Publication detail

Q-learned Controller Performance of Asynchronous electromotor

MARADA, T. BŘEZINA, T.

Czech title

Q-learned controller performance of asynchronous electromotor

English title

Q-learned Controller Performance of Asynchronous electromotor

Type

journal article - other

Language

en

Original abstract

The contribution is focused on the application of Q-learning on asynchronous electromotor (AE) control. The learning process was divided into two phases – prelearning stage and tutorage stage. Tutorage stage uses conventional Q-learning procedure applied on Q function created during the prelearning stage. Tutorage is expected to be processed on real AE plant and its use for controller adaptation on possible system parameters changes is not excluded. Based on the tests the best grids of Q-function table for 2-D state and 3-D state were selected and obtained policies performance was compared with the performance of AE controlled by PID, which was tuned through Ziegler-Nichols methodology.

Czech abstract

The contribution is focused on the application of Q-learning on asynchronous electromotor (AE) control. The learning process was divided into two phases – prelearning stage and tutorage stage. Tutorage stage uses conventional Q-learning procedure applied on Q function created during the prelearning stage. Tutorage is expected to be processed on real AE plant and its use for controller adaptation on possible system parameters changes is not excluded. Based on the tests the best grids of Q-function table for 2-D state and 3-D state were selected and obtained policies performance was compared with the performance of AE controlled by PID, which was tuned through Ziegler-Nichols methodology.

English abstract

The contribution is focused on the application of Q-learning on asynchronous electromotor (AE) control. The learning process was divided into two phases – prelearning stage and tutorage stage. Tutorage stage uses conventional Q-learning procedure applied on Q function created during the prelearning stage. Tutorage is expected to be processed on real AE plant and its use for controller adaptation on possible system parameters changes is not excluded. Based on the tests the best grids of Q-function table for 2-D state and 3-D state were selected and obtained policies performance was compared with the performance of AE controlled by PID, which was tuned through Ziegler-Nichols methodology.

Keywords in English

Asynchronous electromotor, Q-learning, Control

RIV year

2004

Released

23.09.2004

Location

Warsaw, Poland

ISSN

0033-2089

Journal

Elektronika

Volume

2004

Number

8-9

Pages count

5

BIBTEX


@article{BUT45412,
  author="Tomáš {Marada} and Tomáš {Březina},
  title="Q-learned Controller Performance of Asynchronous electromotor",
  journal="Elektronika",
  year="2004",
  volume="2004",
  number="8-9",
  month="September",
  address="Warsaw, Poland",
  issn="0033-2089"
}