[q_learning]Flappy Bird

Flappy Bird

I’ve set values like below.

EPISODES = 100000       COLLISION_PENALTY = 800    POINT_REWARD = 50      ALIVE_REWARD = 1
epsilon = 0.9           EPS_DECAY = 0.99998        STEP = 500             LEARNING_RATE = 0.3
DISCOUNT = 0.95         

The observation values are UPPER and LOWER which are calculated by the coordinate of bird and the coordinate of obstacles.

(UPPER and LOWER) (result) (average of episode rewards)

I modified flappy bird game on https://github.com/Anish-Malla/Flappy-birds-game-using-pygame