def get_state_range(self, verbose=True):
env = self.env
temp_memory = Memory(500000, self.n_state, 1, 1, infinite=True)
if True: #TODO, the other get means functions
episodes = 15
for i in range(episodes):
done = False
s = env.reset()
while not done:
a = env.action_space.sample()
s_ne, reward, done, _ = env.step(a)
temp_memory.push(s, a, reward, s_ne)
s = s_ne
s, _, r, s_ne = temp_memory.get_all_sample()
s_low = np.min(s, axis=0).flatten()
s_high = np.max(s, axis=0).flatten()
low = env.observation_space.low
high = env.observation_space.high
r_high = np.max(r)
r_low = np.min(r)
# high[high == np.inf] = s_high[high == np.inf]
# low[low == -np.inf] = s_low[low == -np.inf]
high = np.array([0.407, 0.25, -0.96, 2, 8])
low = np.array([-0.407, -0.25, -1, -2, -8])
if verbose:
print("using ", temp_memory.counter,
"samples to build means for features")
print("original high and low", env.observation_space.high,
env.observation_space.low)
print("observed high and low", s_high, s_low)
print("manually high and low", high, low)
print("reward high and low", r_high, r_low)
del temp_memory, s
return low, high, r_low, r_high
