dodox/fancy_gym
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

holereach025

Browse Source
This commit is contained in:
Maximilian Huettenrauch
2021-04-10 13:25:08 +02:00
parent 1fd44616bf
commit f6cef69225
4 changed files with 30 additions and 22 deletions
Download Patch File Download Diff File Expand all files Collapse all files
alr_envs/classic_control/hole_reacher.py
+4 -4
View File
@@ -99,10 +99,10 @@ class HoleReacher(gym.Env):
if self._steps == 199:
reward = - np.linalg.norm(self.end_effector - self.bottom_center_of_hole) ** 2
else:
if self.collision_penalty != 0:
reward = -self.collision_penalty
else:
reward = - np.linalg.norm(self.end_effector - self.bottom_center_of_hole) ** 2
# if self.collision_penalty != 0:
# reward = -self.collision_penalty
# else:
reward = - np.linalg.norm(self.end_effector - self.bottom_center_of_hole) ** 2 - self.collision_penalty
reward -= 5e-8 * np.sum(acc ** 2)
alr_envs/classic_control/utils.py
+9 -7
View File
@@ -52,21 +52,22 @@ def make_holereacher_env(rank, seed=0):
_env = HoleReacher(num_links=5,
allow_self_collision=False,
allow_wall_collision=False,
hole_width=0.15,
hole_width=0.25,
hole_depth=1,
hole_x=1,
hole_x=2,
collision_penalty=100)
_env = DmpEnvWrapper(_env,
num_dof=5,
num_basis=5,
duration=2,
bandwidth_factor=2,
dt=_env.dt,
learn_goal=True,
alpha_phase=3.5,
alpha_phase=2,
start_pos=_env.start_pos,
policy_type="velocity",
weights_scale=100,
weights_scale=50,
goal_scale=0.1
)
@@ -103,7 +104,7 @@ def make_holereacher_fix_goal_env(rank, seed=0):
dt=_env.dt,
learn_goal=False,
final_pos=np.array([2.02669572, -1.25966385, -1.51618198, -0.80946476, 0.02012344]),
alpha_phase=3.5,
alpha_phase=3,
start_pos=_env.start_pos,
policy_type="velocity",
weights_scale=50,
@@ -139,14 +140,15 @@ def make_holereacher_env_pmp(rank, seed=0):
_env = DetPMPEnvWrapper(_env,
num_dof=5,
num_basis=5,
width=0.025,
width=0.02,
off=0.,
policy_type="velocity",
start_pos=_env.start_pos,
duration=2,
post_traj_time=0,
dt=_env.dt,
weights_scale=0.2,
zero_start=True,
zero_start=False,
zero_goal=False
)
_env.seed(seed + rank)
alr_envs/utils/dmp_env_wrapper.py
+6 -2
View File
@@ -14,8 +14,9 @@ class DmpEnvWrapper(gym.Wrapper):
start_pos=None,
final_pos=None,
duration=1,
alpha_phase=2,
dt=0.01,
alpha_phase=2,
bandwidth_factor=3,
learn_goal=False,
post_traj_time=0.,
policy_type=None,
@@ -35,7 +36,10 @@ class DmpEnvWrapper(gym.Wrapper):
self.post_traj_steps = int(post_traj_time / dt)
phase_generator = ExpDecayPhaseGenerator(alpha_phase=alpha_phase, duration=duration)
basis_generator = DMPBasisGenerator(phase_generator, duration=duration, num_basis=self.num_basis)
basis_generator = DMPBasisGenerator(phase_generator,
duration=duration,
num_basis=self.num_basis,
basis_bandwidth_factor=bandwidth_factor)
self.dmp = dmps.DMP(num_dof=num_dof,
basis_generator=basis_generator,