Editing the Q-table before Training in Basic Grid World?

2 ビュー (過去 30 日間)
Donald Alexander
Donald Alexander 2021 年 6 月 22 日
コメント済み: Donald Alexander 2021 年 6 月 22 日
For "Train Reinforcement Learning Agent in Basic Grid World", is it posible to edit the Q-table with specific values before training begins to influence the robot's early decision making? I am attempting to create a path plan from a start to goal point and I want to provide the robot with some priori knowledge. I have run grid A* and want the Q-table to know this route before training.
I haven't been able to figure this out, so I tried to work around it by giving the robot positive rewards for going along the A* route. But the training ends up just exactly following the A*. Therefore, I want to just give the Q-table some initial values higher than 0 when starting so it probabilistically follows this route to the goal.

採用された回答

Emmanouil Tzorakoleftherakis
Emmanouil Tzorakoleftherakis 2021 年 6 月 22 日
Hello,
Please take a look at this link that mentions how you can initialize the table.
  1 件のコメント
Donald Alexander
Donald Alexander 2021 年 6 月 22 日
Thank you! I appreciate the quick response.

サインインしてコメントする。

その他の回答 (0 件)

製品


リリース

R2021a

Community Treasure Hunt

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

Start Hunting!

Translated by