def _init():
env_config = solo8v2vanilla.Solo8VanillaConfig()
env = gym.make('solo8vanilla-v0',
config=env_config,
normalize_actions=False)
env.obs_factory.register_observation(obs.TorsoIMU(env.robot))
env.obs_factory.register_observation(obs.MotorEncoder(env.robot))
env.termination_factory.register_termination(
terms.TimeBasedTermination(length))
stand_reward = rewards.AdditiveReward()
stand_reward.client = env.client
stand_reward.add_term(0.5, rewards.FlatTorsoReward(env.robot))
stand_reward.add_term(
0.5, rewards.TorsoHeightReward(env.robot, quad_standing_height))
home_pos_reward = rewards.MultiplicitiveReward(
1, stand_reward, rewards.SmallControlReward(env.robot),
rewards.HorizontalMoveSpeedReward(env.robot, 0))
home_pos_reward.client = env.client
env.reward_factory.register_reward(1, home_pos_reward)
return env
def test_computation(self, name, speed, target, hard, soft, low, high):
mock_client = mock.MagicMock()
mock_client.getBaseVelocity.return_value = (speed / math.sqrt(2),
speed / math.sqrt(2),
None), None
r = rewards.HorizontalMoveSpeedReward(1, target, hard, soft)
r.client = mock_client
val = r.compute()
self.assertGreaterEqual(val, low)
self.assertLessEqual(val, high)
def test_init(self):
robot_id = 5
target_speed = 0
hard_margin = 20
soft_margin = 12
r = rewards.HorizontalMoveSpeedReward(robot_id, target_speed,
hard_margin, soft_margin)
self.assertEqual(robot_id, r._robot_id)
self.assertEqual(target_speed, r._target_speed)
self.assertEqual(hard_margin, r._hard_margin)
self.assertEqual(soft_margin, r._soft_margin)
def _init():
env_config = solo8v2vanilla.Solo8VanillaConfig()
env = gym.make('solo8vanilla-v0',
config=env_config,
normalize_actions=False)
env.obs_factory.register_observation(obs.TorsoIMU(env.robot))
env.obs_factory.register_observation(obs.MotorEncoder(env.robot))
env.termination_factory.register_termination(
terms.TimeBasedTermination(length))
env.reward_factory.register_reward(
.2, rewards.SmallControlReward(env.robot))
env.reward_factory.register_reward(
.2, rewards.HorizontalMoveSpeedReward(env.robot, 0))
env.reward_factory.register_reward(.3,
rewards.FlatTorsoReward(env.robot))
env.reward_factory.register_reward(
.3, rewards.TorsoHeightReward(env.robot, quad_standing_height))
return env
def _init():
env_config = solo8v2vanilla.Solo8VanillaConfig()
env_config.max_motor_rotation = max_motor_rot
env = gym.make('solo8vanilla-v0',
config=env_config,
normalize_actions=True,
normalize_observations=True)
env.obs_factory.register_observation(obs.TorsoIMU(env.robot))
env.obs_factory.register_observation(
obs.MotorEncoder(env.robot, max_rotation=max_motor_rot))
env.termination_factory.register_termination(
terms.TimeBasedTermination(length))
stand_reward = rewards.AdditiveReward()
stand_reward.client = env.client
stand_reward.add_term(
0.5,
rewards.FlatTorsoReward(env.robot,
hard_margin=fhm,
soft_margin=fsm))
stand_reward.add_term(
0.5,
rewards.TorsoHeightReward(env.robot,
stand_height,
hard_margin=thm,
soft_margin=tsm))
home_pos_reward = rewards.MultiplicitiveReward(
1, stand_reward, rewards.SmallControlReward(env.robot, margin=scm),
rewards.HorizontalMoveSpeedReward(env.robot,
0,
hard_margin=hmhm,
soft_margin=hmsm))
env.reward_factory.register_reward(1, home_pos_reward)
return env
from gym_solo.core import termination as terms
if __name__ == '__main__':
config = solo8v2vanilla.Solo8VanillaConfig()
env: solo8v2vanilla.Solo8VanillaEnv = gym.make('solo8vanilla-v0', use_gui=True,
realtime=True, config=config)
env.obs_factory.register_observation(obs.TorsoIMU(env.robot))
env.termination_factory.register_termination(terms.PerpetualTermination())
flat = rewards.FlatTorsoReward(env.robot, hard_margin=.1, soft_margin=np.pi)
height = rewards.TorsoHeightReward(env.robot, 0.33698, 0.025, 0.15)
small_control = rewards.SmallControlReward(env.robot, margin=10)
no_move = rewards.HorizontalMoveSpeedReward(env.robot, 0, hard_margin=.5,
soft_margin=3)
stand = rewards.AdditiveReward()
stand.client = env.client
stand.add_term(0.5, flat)
stand.add_term(0.5, height)
home_pos = rewards.MultiplicitiveReward(1, stand, small_control, no_move)
env.reward_factory.register_reward(1, home_pos)
joint_params = []
num_joints = env.client.getNumJoints(env.robot)
for joint in range(num_joints):
joint_params.append(env.client.addUserDebugParameter(
'Joint {}'.format(
