Nicolò Lutteri

Move To Point


The objective of this project is to create an IA with the use of Reinforcement Learning that from a starting point A reaches the end point B.



This problem can be easily solved by using, for example, A* or even easily by subtracting the coordinated, but instead, I want to try solving it with the RL strategy. The IA is a neural network with the locations X1 and X2 (the starting point) and the locations Y1 and Y2 (the final location) as input; the output of this IA is the four directions (up, down, left, and right). The reward I used for the training is -1 always; this is to create an against ambient that the IA will try to leave as soon as possible. After 200 episodes, the IA can reach point B.

Tecnology Used

Python (TensorFlow, PIL)