Creating Q-table

5 views (last 30 days)
Izzat Brgli
Izzat Brgli on 17 Mar 2021
Commented: RUBEN HERNANDEZ on 26 Apr 2022
I have made my own RL environment where the observation is a vector of 8 elements "for example [1,2,3,4,5,6,7,8]" and the action set has four elements(["up","down","right", "left"]) and I'm trying to apply Q-Learning algorithm. my question is how to create the q-table. I did read about
T = rlTable(obsinfo,actinfo)
but I got stuck without any progress. How can I create the table?
  2 Comments
hsu yi cheng
hsu yi cheng on 3 Sep 2021
Edited: hsu yi cheng on 3 Sep 2021
Hi, I'm working on a similar project as yours. May I have your file as a reference ? Thanks
RUBEN HERNANDEZ
RUBEN HERNANDEZ on 26 Apr 2022
Hi, Im' trying to simulate Q-learnig for control inverted pendulum in simulink, but i have problem to create the q-table.
For example, I have three watch signals. How can it be defined as discrete observations with their limits?

Sign in to comment.

Answers (1)

Emmanouil Tzorakoleftherakis
Edited: Emmanouil Tzorakoleftherakis on 18 Mar 2021
Did you take a look at this example? It seems to solve a similar problem.
If you want to use the provided API to create a custom grid world, I would follow this link.
If you are implemeting your own, you want to use rlFiniteSetSpec which is designed for discrete observation. rlNumericSpec is for observations continuous in nature.
  4 Comments
Izzat Brgli
Izzat Brgli on 18 Mar 2021
Edited: Izzat Brgli on 18 Mar 2021
I did what you have recommended and it worked partially. I have right now a table which has the same row's number as the observation's number and the column's number as the action's number but without header (names for columns and rows) and the question is how can I retrieve the data or store it?
Emmanouil Tzorakoleftherakis
Not sure I understand the question. If you type
yourTablename.Table
you have access to all the table elements and you can use them however you need.

Sign in to comment.

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!