pip install -r requirements.txt

Reinforcement Learning Implementation- Grid World

This project is meant to demonstrate a wide variety of RL algorithms in Grid World. Including Dynamic Programming : Value iterations， Policy iteration Model-free: MC，Q-learning, SARSA, Policy Gradient.

• main_.py - Just run it to view different algorithms.

• agent_.pyandgrid_env.py - Defines different agents for different algorithms and grid world for traditional algorithms(grid_env.py: actions that lead to forbidden areas or the boundaries are excluded) and DRL(grid_env_nn.py: included)

• StoredTrainingData - Trained deep neural network and V, Q Tabs.

For assignment 4

Include main_PG.py, agent_PG.py,agent_PG_e_greedy, grid_env_nn.py；

To see MC Exploring Starts,

Run main_PG.py with

#import agent.agent_PG as ag
import agent.agent_PG_e_greedy as ag

To see MC epsilon-greedy,

Run main_PG.py with

import agent.agent_PG as ag
#import agent.agent_PG_e_greedy as ag

For more information

Go through my report