Basic implementation of Markov Decision Processes using Values Iteration algorithm with graphic representation in Unity
Open Unity project and find MPD gameObject in the hierarchy. There is MDP script attached to this gameObject. There are some values to change:
- Penatly - reward/penalty for each step
- Probability - probability of choosing best action
- Discount factor - gamma parameter in Bellman's quation
- Rows/Columns - size of enviroment table
- Good/Bad Rewards - value of preferable/unpreferable state
- Epsilon - stops the algorithm when absolute value of changes is less than that value
Use left mouse button to create preferable/unpreferable states by clicking on cells. Use right mouse button to create walls. Use spacebar to start the algorithm. Morover after pressing mouse wheel arows with directions (policy) will be shown.
