def _thunk():
env = gym.make(env_id)
env = RandomizedEnvWrapper(env, seed + rank)
env.seed(seed + rank)
obs_shape = env.observation_space.shape # TODO: is this something we can remove
return env
#!/usr/bin/env python
import time
import numpy as np
import gym
from tqdm import tqdm
import common.envs
from common.envs.wrappers import RandomizedEnvWrapper
env = gym.make('Pusher3DOFRandomized-v0')
env = RandomizedEnvWrapper(env=env, seed=0) # env.randomize()
actions = [
np.array([0.0, 0.0, 1.0]),
np.array([0.0, 0.0, -1.0]),
np.array([0.0, 1.0, 0.0]),
np.array([0.0, -1.0, 0.0]),
np.array([1.0, 0.0, 0.0]),
np.array([-1.0, 0.0, 0.0])
]
actions.reverse()
action_change_freq = 50
for env_idx in tqdm(range(10)):
env.reset()
env.render()
for action in actions:
# /usr/bin/env python
import time
import gym
import common.envs
from common.envs.wrappers import RandomizedEnvWrapper
env = gym.make('HumanoidRandomizedEnv-v0')
env = RandomizedEnvWrapper(env=env, seed=0)
reward = 0.
env.randomize(randomized_values=[
"random", "random", "random", "random", "random", "random"
])
env.reset()
env.render()
d = False
while True:
s_, r, d, info = env.step(env.action_space.sample())
time.sleep(0.1)
if d:
env.randomize(randomized_values=[
"random", "random", "random", "random", "random", "random"
])
env.reset()
env.render()
from common.envs.wrappers import RandomizedEnvWrapper
# MODE = "MANUAL" # slow but let's you see what's happening
MODE = "SPEED" # as fast as possible
def no_op(x):
pass
if MODE == "MANUAL":
env = gym.make("ErgoReacherRandomized-Graphical-v0") # looks nice
timer = time.sleep
else:
env = gym.make("ErgoReacherRandomized-Headless-v0") # runs fast
timer = no_op
env = RandomizedEnvWrapper(env=env, seed=0)
for _ in tqdm(range(100)):
env.reset()
env.randomize(randomized_values=["random"] *
8) # 8 values to randomize over
while True:
action = env.action_space.sample()
obs, rew, done, misc = env.step(action)
timer(0.05)
if done:
break
# manual
"""
This script allows you to manually control the simulator or Duckiebot
using the keyboard arrows.
"""
import sys
from pynput import keyboard
from pynput.keyboard import Key
import numpy as np
import gym
import common.envs
from common.envs.wrappers import RandomizedEnvWrapper
env = gym.make("ErgoReacher4DOFRandomizedHardVisual-v0")
env = RandomizedEnvWrapper(env=env, seed=0)
env.reset()
env.render()
ACTION_KEYS = [
Key.up, Key.down, Key.page_up, Key.page_down, Key.right, Key.left,
Key.home, Key.end, Key.alt
]
def on_press(key):
if key == Key.tab: env.reset()
if key in ACTION_KEYS:
action = np.zeros(4)
if key != Key.alt:
import time
from timeit import default_timer as timer
import numpy as np
import tqdm
import gym
import common.envs
from common.envs.wrappers import RandomizedEnvWrapper
np.random.seed(1234)
env = gym.make('PusherRandomized-v0')
env = RandomizedEnvWrapper(env=env, seed=0)
obs = env.reset()
start = timer()
for i in tqdm.tqdm(range(int(1e6))):
env.randomize(randomized_values=["random", "random", "random"])
print(timer() - start)
# manual
"""
This script allows you to manually control the simulator or Duckiebot
using the keyboard arrows.
"""
import sys
from pynput import keyboard
from pynput.keyboard import Key
import numpy as np
import gym
import common.envs
from common.envs.wrappers import RandomizedEnvWrapper
env = gym.make('Pusher3DOFUberHard-v0')
env = RandomizedEnvWrapper(env=env, seed=0)
reward = 0.
print('hi')
env.randomize(randomized_values=["random", "random"])
env.reset()
env.render()
ACTIONS = [
np.array([0.0, 0.0, 1.0]),
np.array([0.0, 0.0, -1.0]),
np.array([0.0, 1.0, 0.0]),
np.array([0.0, -1.0, 0.0]),
np.array([1.0, 0.0, 0.0]),
np.array([-1.0, 0.0, 0.0])
import time
from timeit import default_timer as timer
import numpy as np
import tqdm
import gym
import common.envs
from common.envs.wrappers import RandomizedEnvWrapper
np.random.seed(1234)
env = gym.make('Pusher3DOFRandomized-v0')
env = RandomizedEnvWrapper(env=env, seed=0)
# obs = env.reset()
start = timer()
for i in tqdm.tqdm(range(100)):
env.randomize(randomized_values=["random", "random", "random"])
env.reset()
for _ in range(200):
obs, reward, done, _ = env.step(env.action_space.sample())
env.render()
print(obs)
env.close()
print(timer() - start)
#!/usr/bin/env python
import time
import numpy as np
import gym
from tqdm import tqdm
import common.envs
from common.envs.wrappers import RandomizedEnvWrapper
env = gym.make('Pusher3DOFRandomized-v0')
env = RandomizedEnvWrapper(env=env, seed=0) # env.randomize()
actions = [
np.array([0.0, 0.0, 1.0]),
np.array([0.0, 0.0, -1.0]),
np.array([0.0, 1.0, 0.0]),
np.array([0.0, -1.0, 0.0]),
np.array([1.0, 0.0, 0.0]),
np.array([-1.0, 0.0, 0.0])
]
actions.reverse()
discretization = 50
randomized_values = ["default", "default"]
values = np.linspace(0, 1, discretization)
for dim in range(2):
for i in tqdm(range(discretization)):
rands = randomized_values