tree_actions_index = np.load("tree_actions_index.npy")
Q = np.load("treeQmatrix.npy")
env = Carom(render=True)
env.reset()
actions = env.get_actions()
nb_states = 0
index2 = 0
state = 0
episode_reward = 0
nb_branches = len(tree_actions_index[0, 0])
for i in range(tree_actions_index.shape[0]):
nb_states += nb_branches**i
for level in range(tree_actions_index.shape[0]):
chosen_branch = choose_branch(Q, state)
action_index = int(tree_actions_index[level, index2][chosen_branch])
reward, coll_r = env.step3(actions[action_index][0],
actions[action_index][1],
actions[action_index][2],
actions[action_index][3],
actions[action_index][4])
episode_reward += reward
index2 = index2 * nb_branches + chosen_branch
nextState = state * nb_branches + chosen_branch + 1
state = nextState
print("Total reward: %.3f" % (episode_reward))
theta = 5
phi = 85
V = 5
actions = env.get_fixed_actions()
nb_shots = 5
shots = []
for i in range(nb_shots):
coll_reward = 0
while coll_reward == 0:
#env.reset(pos_white, pos_yellow, pos_red)
env.reset()
action_index = int(np.random.choice(len(actions)))
action_reward, coll_reward = env.step3(actions[action_index][0],
actions[action_index][1],
actions[action_index][2],
actions[action_index][3],
actions[action_index][4])
if coll_reward == 1:
if action_index in shots:
coll_reward = 0
shots.append(action_index)
env.render = True
for i in range(nb_shots):
#env.reset(pos_white, pos_yellow, pos_red)
env.reset()
env.step3(actions[shots[i]][0], actions[shots[i]][1], actions[shots[i]][2],
actions[shots[i]][3], actions[shots[i]][4] - 1)
# env.reset(pos_white, pos_yellow, pos_red)