def prepare_input(parking_states):
result = build_input(env, parking_states[0]).reshape(INPUT_SIZE, 1)
for i in range(1, len(parking_states)):
result = np.column_stack(
(result, build_input(env,
parking_states[i]).reshape(INPUT_SIZE, 1)))
return result.T
def prepare_input(current_state):
return build_input(env, current_state).reshape((1, INPUT_SIZE))
return build_input(env, current_state).reshape((1, INPUT_SIZE))
if __name__ == '__main__':
network = DqnNetwork(INPUT_SIZE)
env.reset()
new_state = env.s
env.render()
total_wins = 0
steps_to_win = 0
print_values(env.s)
iteration = 0
while True:
k = 0
frame = build_input(env, new_state)
end = False
print("----NEW ROUND----")
my_reward = 0
states = list()
rewards = list()
iteration += 1
while not end:
action = network.predict(frame)
max_action = np.argmax(action)
old_state = new_state
new_state, reward, done, info = env.step(max_action)
k += 1
steps_to_win += 1
my_reward, end = calculate_reward(old_state, new_state, k)
def prepare_input(current_state, next_state):
return np.column_stack(
(build_input(env, current_state), build_input(env, next_state))).T
if __name__ == '__main__':
network = McNetwork(INPUT_SIZE)
env.reset()
env.render()
total_wins = 0
iteration = 0
performance = list()
averages = list()
avg = 0
last_plot_point = 0
episodes = 0
while True:
new_state = env.s
k = 0
frame = build_input(env, new_state)
end = False
print("----NEW ROUND----")
env.render()
my_reward = 0
iteration += 1
states = list()
rewards = list()
while not end:
action = network.predict(frame)
max_action = np.argmax(action)
old_state = new_state
new_state, reward, done, info = env.step(max_action)
k += 1
my_reward, end, reset = calculate_reward(old_state, new_state, k,