A college project for Machine Intelligence course. Exploring Reinforcement Learning using Direct Utility Estimation on grid world.
-
Direct Utility Estimation on three scenarios of grid world problem (4X3, 10X10 with different final reward positions).
-
Two types of Direct Utility Estimation are considered :
- Using tabular representation.
- Using simple function approximator (shown in equation 21.10 in reference).
-
Update policies within text files in
policies
folder. Text files are named after the numbers of the corresponding grid worlds. -
Run the direct utility estimation (using a specific representation on a specific grid) :
python run.py <agent_number> <grid_number> <path/to/policy/text/file>