Classes
class	Eval

Variables
int	seed = 42

string	load_path = '/home/giuseppe/src/taxons/experiments/Maze_AE_Mixed/{}'

	params = parameters.Params()

	env = gym.make(params.env_tag)

list	target_pose = [450, 500]

	obs = env.reset(desired_ball_pose=target_pose)

	target_image = env.render(mode='rgb_array')

	qpos = np.zeros(15)

	qvel = np.zeros(14)

	pose = target_pose+env.initPos[-1:]

	p = fs.Posture(*pose)

	device = torch.device("cuda" if torch.cuda.is_available() else "cpu")

	selector = ae.ConvAE(device=device, encoding_shape=params.feature_size)

	agent_type = agents.FFNeuralAgent

	pop = population.Population(agent=agent_type, pop_size=0, shapes=params.agent_shapes)

bool	done = False

int	t = 0

	agent_input = t

	action = utils.action_formatting(params.env_tag, agent['agent'](agent_input))

	reward

	info

	CoM = np.array([env.env.data.qpos[:2]])

	state = env.render(mode='rgb_array')

	surprise

	bs_point = bs_point.flatten().cpu().data.numpy()

	y

	goal = torch.Tensor(np.ascontiguousarray(target_image)).permute(2, 0, 1).unsqueeze(0).to(device)

	reconstr

	bs_space = np.stack([a[0] for a in pop['features'].values])

	diff = np.atleast_2d(bs_space - bs_point)

	dists = np.sqrt(np.sum(diff * diff, axis=1))

	closest_agent = np.argmin(dists)

	selected = pop[closest_agent]

list	images = []

list	saved_balls_pose = []

list	saved_joints_pose = []

list	saved_robot_pose = []

	f_pose = CoM

	SHOW_ARM_IN_ARRAY

	point_pose

	final_distance = np.sqrt(np.sum((target_pose - f_pose) ** 2))

Variable Documentation

test_control.action = utils.action_formatting(params.env_tag, agent['agent'](agent_input))

list test_control.agent_input = t

test_control.agent_type = agents.FFNeuralAgent

test_control.bs_point = bs_point.flatten().cpu().data.numpy()

test_control.bs_space = np.stack([a[0] for a in pop['features'].values])

test_control.closest_agent = np.argmin(dists)

test_control.CoM = np.array([env.env.data.qpos[:2]])

test_control.device = torch.device("cuda" if torch.cuda.is_available() else "cpu")

test_control.diff = np.atleast_2d(bs_space - bs_point)

test_control.dists = np.sqrt(np.sum(diff * diff, axis=1))

bool test_control.done = False

test_control.env = gym.make(params.env_tag)

test_control.f_pose = CoM

test_control.final_distance = np.sqrt(np.sum((target_pose - f_pose) ** 2))

test_control.goal = torch.Tensor(np.ascontiguousarray(target_image)).permute(2, 0, 1).unsqueeze(0).to(device)

test_control.images = []

test_control.info

string test_control.load_path = '/home/giuseppe/src/taxons/experiments/Maze_AE_Mixed/{}'

test_control.obs = env.reset(desired_ball_pose=target_pose)

test_control.p = fs.Posture(*pose)

test_control.params = parameters.Params()

test_control.point_pose

Initial value:

1 = np.array([np.sin(saved_joints_pose[:, 0]) + .9 * np.cos(np.pi / 2. - saved_joints_pose[:, 1] - saved_joints_pose[:, 0]),

2 np.cos(saved_joints_pose[:, 0]) + .9 * np.sin(np.pi / 2. - saved_joints_pose[:, 1] - saved_joints_pose[:, 0])]).transpose()

test_control.pop = population.Population(agent=agent_type, pop_size=0, shapes=params.agent_shapes)

test_control.pose = target_pose+env.initPos[-1:]

test_control.qpos = np.zeros(15)

test_control.qvel = np.zeros(14)

test_control.reconstr

test_control.reward

test_control.saved_balls_pose = []

test_control.saved_joints_pose = []

int test_control.saved_robot_pose = []

int test_control.seed = 42

test_control.selected = pop[closest_agent]

test_control.selector = ae.ConvAE(device=device, encoding_shape=params.feature_size)

test_control.SHOW_ARM_IN_ARRAY

int test_control.state = env.render(mode='rgb_array')

test_control.surprise

int test_control.t = 0

test_control.target_image = env.render(mode='rgb_array')

list test_control.target_pose = [450, 500]

test_control.y

Classes

Variables

Variable Documentation