Q-Learning is reinforcement learning techique that is used mainly in game play and robotics.
In this project, our agent will learn by experience without teacher. The agent explores the map randomly until it reaches the goal or caught by a danger. If it reaches the goal, it will get an immediate reward or if it reaches a danger, it will get an immediate punishment. By repeating this procedure for many times, our agent learns the environment and finds optimal ways to the goal independently from starting point.
Features :
- Various simulations can be made with learning the map once.
- Some logs about simulation info can be followed.
- Various types of maps can be loaded.
- 3 degrees of learning level can be selected.
- A graphical representation of learning matrix can be viewed. |
