class TestEnvironment(unittest.TestCase):
def setUp(self):
config_file = os.path.join(os.path.dirname(os.path.abspath(__file__)),
'basic_env.yaml')
self.env = DIYGym(config_file)
def test_load_environment(self):
self.assertTrue('plane' in self.env.models)
self.assertTrue('red_marble' in self.env.models)
self.assertTrue('green_marble' in self.env.models)
self.assertTrue('blue_marble' in self.env.models)
def test_spaces(self):
self.assertTrue('force' in self.env.action_space['blue_marble'].spaces)
self.assertTrue(
'camera' in self.env.observation_space['basic_env'].spaces)
self.assertTrue(
'pose' in self.env.observation_space['green_marble'].spaces)
def test_episode(self):
observation = self.env.reset()
initial_position = observation['green_marble']['pose']['position']
# try to run the blue marble into the other two
for _ in range(500):
observation, _, _, _ = self.env.step(
{'blue_marble': {
'force': [0, -100, 0]
}})
final_position = observation['green_marble']['pose']['position']
# check that the green marble has moved
self.assertNotAlmostEqual(np.linalg.norm(initial_position),
np.linalg.norm(final_position),
places=0)
observation = self.env.reset()
reset_position = observation['green_marble']['pose']['position']
# check that the green marble has been reset back to its starting position
self.assertAlmostEqual(np.linalg.norm(initial_position),
np.linalg.norm(reset_position),
places=1)
#!/usr/bin/env python
import numpy as np
import os
from diy_gym import DIYGym
if __name__ == '__main__':
config_file = os.path.join(os.path.dirname(os.path.abspath(__file__)), 'ur_high_5.yaml')
env = DIYGym(config_file)
observation = env.reset()
while True:
action = env.action_space.sample()
action['ur5_l']['controller']['linear'][:] = observation['ur_high_5']['distance_to_target']['position'] * 0.
action['ur5_r']['controller']['linear'][:] = -observation['ur_high_5']['distance_to_target']['position'] * 0.
action['ur5_l']['controller']['rotation'][:] = 0
action['ur5_r']['controller']['rotation'][:] = 0
observation, reward, terminal, info = env.step(action)
if terminal:
print('slap!')
observation = env.reset()
if __name__ == '__main__':
config_file = os.path.join(os.path.dirname(os.path.abspath(__file__)),
'r2d2_maze.yaml')
env = DIYGym(config_file)
speed = 10.
key_and_delta = {
p.B3G_LEFT_ARROW: [-1, -1, 1, 1],
p.B3G_RIGHT_ARROW: [1, 1, -1, -1],
p.B3G_UP_ARROW: [-1, -1, -1, -1],
p.B3G_DOWN_ARROW: [1, 1, 1, 1]
}
action = env.action_space.sample()
while True:
action['r2d2']['wheel_driver'][:] = 0
keys = p.getKeyboardEvents()
if 32 in keys and keys[32] & p.KEY_WAS_RELEASED:
env.reset()
for key, delta in key_and_delta.items():
if key in keys and keys[key] & p.KEY_IS_DOWN:
action['r2d2']['wheel_driver'] += np.array(delta) * speed
observation, reward, terminal, info = env.step(action)