Multiple goals in Grid World

4 views (last 30 days)
GCats
GCats on 15 Feb 2022
Hello everyone,
Is it possible to train the agent to reach different goals in the same episode? I'm working with the standard GridWorld environment given here: https://nl.mathworks.com/help/reinforcement-learning/ref/creategridworld.html#d123e1207 and I would like to have the agent reach for example state [2,1] and then [3,5] and finally [3,3] in this order. Is it possible to do so?
If i try to modify the terminal state option as
GW.TerminalStates = ['[2,1]'; '[3,5]'; '[3,3]'];
The agent only reaches the closest one to the initial state and not all of them. Any hints?

Answers (0)

Community Treasure Hunt

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

Start Hunting!