def run(self):
# initialize
dqn_agent = DQN(0.99, 0, 1, 0.001, 50000, 32, 4)
g = game.gameState()
action_0 = np.array([1, 0, 0, 0])
screen_0, reward_0, is_dead_0 = g.frameStep(action_0)
screen_0 = cv2.cvtColor(cv2.resize(screen_0, (84, 84)),
cv2.COLOR_BGR2GRAY)
_, screen_0 = cv2.threshold(screen_0, 1, 255, cv2.THRESH_BINARY)
dqn_agent.init_state(screen_0)
while True:
# get action and update the game status
action = dqn_agent.get_action()
screen_1, r, done = g.frameStep(action)
screen_1 = self.screen_handle(screen_1)
_, q_val = dqn_agent.add_replay(screen_1, action, r, done)
if done == True:
score, episode = g.ret_score()
print(score, episode)
def run(self):
# initialize
# discount factor 0.99
ag = DQN(0.99, 0, 1, 0.001, 50000, 32, 4)
g = game.gameState()
a_0 = np.array([1, 0, 0, 0])
s_0, r_0, d = g.frameStep(a_0)
s_0 = cv2.cvtColor(cv2.resize(s_0, (84, 84)), cv2.COLOR_BGR2GRAY)
_, s_0 = cv2.threshold(s_0, 1, 255, cv2.THRESH_BINARY)
ag.initState(s_0)
while True:
a = ag.getAction()
s_t1, r, done = g.frameStep(a)
s_t1 = self.screen_handle(s_t1)
ts, qv = ag.addReplay(s_t1, a, r, done)
if done == True:
sc, ep = g.retScore()
print("Epoch: {} Q-Value: {:.3f} Episode: {} Score: {}".format(ts,qv,ep,sc))
else:
print("Epoch: {} Q-Value: {:.3f}".format(ts,qv))
def run(self):
# initialize
g = game.gameState()
a_0 = np.array([1, 0, 0, 0])
s_0, r_0, done = g.frameStep(a_0)
s_0 = cv2.cvtColor(cv2.resize(s_0, (84, 84)),cv2.COLOR_BGR2GRAY)
_, s_0 = cv2.threshold(s_0, 1, 255, cv2.THRESH_BINARY)
ag = DQN()
s=ag.initState(s_0)
while True:
a = ag.get_action()
s1, r, done = g.frameStep(a)
s1 = self.screen_handle(s1)
print(s1.shape)
ts = ag.store_transition(s1, a, r, done)
qv=ag.learn()
if done == True:
sc, ep = g.retScore()
print(ts,",",qv,",",ep, ",", sc)
else:
print(ts,",",qv,",,")