rl_labs

Lab materials for the Reinforcement Learning class

Install

git clone https://github.com/inmo-jang/rl_labs.git
cd rl_labs
pip install -r requirements.txt

Simple Maze Grid

Manual Play

Deterministic Mode

python gyms/simple_maze_grid.py

• Keyboard Control
  • q or ESC: quit
  • r: reset

For Stochastic Mode

python gyms/simple_maze_grid.py --stochastic=True --epsilon=0.2

This means that an action will be randomly chosen 20% of the time.

Import

env = SimpleMazeGrid(n=5, k=3, m=2, render_option=True, random_seed=42)

# n = Length of one side of the grid
# k = Starting index from which the goal can be generated, extending to the end of the grid
# m = Number of obstacles

• Actions: Discrete(2)
• Observation Space: Discrete(n * n)
• Reward:
  • 10 when arriving the goal
  • -10 when arriving a pit
  • -1 otherwise