fig = plt.figure(figsize=(20, 20))
bilbo = QLearningAgent(PLAYER_CHAR)
mondo = World(WORLD_DIM, bilbo=bilbo, obstacle=True)
game_ended = False
epoch = 0
anim = []
rewards = 0
env = mondo.create_env(d)
anim.append((plt.pcolormesh(env, cmap='CMRmap'), ))
while not game_ended and epoch < MAX_EPOCH:
epoch += 1
action = bilbo.get_action(0, q_table, possible_moves)
bilbo.move(inverse_possible_moves[action])()
game_ended = bilbo.game_ended()
reward = bilbo.reward()
rewards = rewards + reward
env = mondo.create_env(d)
anim.append((plt.pcolormesh(env, cmap='CMRmap'), ))
im_ani = animation.ArtistAnimation(fig,
anim,
interval=60,
repeat_delay=1000,
blit=False)
writer = animation.FFMpegWriter(fps=epoch)
for ep in range(TOT_EPISODES):
#recreate the environment
bilbo = QLearningAgent(PLAYER_CHAR)
mondo = World(WORLD_DIM, bilbo=bilbo, obstacle=True)
np.random.seed()
game_ended = False
epoch = 0
tot_reward = 0
#if ep % 10 == 0:
#a = plt.imshow(render_world(mondo.world,WORLD_DIM,q_table,ep), animated=True)
#policy.append((a,))
while not game_ended and epoch < MAX_EPOCH:
#the near it gets to the dragon the more random the movement
epoch += 1
epsilon_fear = bilbo.fear(epsilon)
action = bilbo.get_action(epsilon, q_table, possible_moves)
current_state = bilbo.get_current_state()
#treasure_gone = bilbo.treasure_gone()
old_q_val = q_table[current_state][action]
bilbo.move(inverse_possible_moves[action])()
new_state = bilbo.get_current_state()
#treasure_gone = bilbo.treasure_gone()
game_ended = bilbo.game_ended()
reward = bilbo.reward()
if reward == -DRAGON_PENALTY:
new_q_val = reward
loss += 1
elif reward == TREASURE_REWARD: