# pos = env.arraystate2pos(state)
# print(pos)
# optimal_action = np.zeros(2)
# action, optimal_action, a, b, theta = agent.test(env, nb_episodes=500000, visualize=False, nb_max_episode_steps=200, modif = True, pos = pos)
# env.non_random_reset(pos[0], pos[1], pos[2])
# env.render = True
# env.step(action, rand = optimal_action, a = a, b = b, theta = theta)
# state = env.reset()
# pos = env.arraystate2pos(state)
# optimal_action = np.zeros(2)
# optimal_action[0], optimal_action[1], a, b, theta = agent.test(env, nb_episodes=500000, visualize=False, nb_max_episode_steps=200, modif = True, pos = pos)
# env.non_random_reset(pos[0], pos[1], pos[2])
# env.render = True
# env.step(optimal_action, a = a, b = b, theta = theta)
nb_test = 50
for i in range(nb_test):
env.render = False
state = env.reset()
pos = env.arraystate2pos(state)
optimal_action = np.zeros(2)
action, optimal_action, a, b, theta = agent.test(env,
nb_episodes=500000,
visualize=False,
nb_max_episode_steps=200,
modif=True,
pos=pos)
env.non_random_reset(pos[0], pos[1], pos[2])
env.render = True
env.step(action, rand=optimal_action, a=a, b=b, theta=theta)
import numpy as np
from carom import Carom
import gym
from vpython import *
nb_rows = 1000
nb_good_steps = 0
env = Carom(render=True)
np.random.seed(32)
env.seed(32)
demo_table = np.load('demoTable.npy')
for i in range(10):
index = np.random.choice(1000)
env.state = demo_table[index][0]
pos_white = vector(env.state[0], env.state[1], 0)
pos_yellow = vector(env.state[2], env.state[3], 0)
pos_red = vector(env.state[4], env.state[5], 0)
env.non_random_reset(pos_white, pos_yellow, pos_red)
action = demo_table[index][1]
env.step(action)