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

Merge pull request #6 from ALRhub/contextual_dmp_wrapper

Browse Source 
Contextual dmp wrapper + environments
This commit is contained in:
ottofabian 2021-05-18 10:54:19 +02:00 committed by GitHub
commit e0e4d6d41c
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
23 changed files with 723 additions and 1097 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

151
alr_envs/__init__.py
View File

@ -1,7 +1,8 @@
from gym.envs.registration import register
from alr_envs.stochastic_search.functions.f_rosenbrock import Rosenbrock
# from alr_envs.utils.wrapper.dmp_wrapper import DmpWrapper
# from alr_envs.utils.mps.dmp_wrapper import DmpWrapper
# Mujoco
@ -71,6 +72,17 @@ register(
}
)
## Balancing Reacher
register(
id='Balancing-v0',
entry_point='alr_envs.mujoco:BalancingEnv',
max_episode_steps=200,
kwargs={
"n_links": 5,
}
)
register(
id='ALRBallInACupSimple-v0',
entry_point='alr_envs.mujoco:ALRBallInACupEnv',
@ -101,15 +113,7 @@ register(
# Classic control
register(
id='Balancing-v0',
entry_point='alr_envs.mujoco:BalancingEnv',
max_episode_steps=200,
kwargs={
"n_links": 5,
}
)
## Simple Reacher
register(
id='SimpleReacher-v0',
entry_point='alr_envs.classic_control:SimpleReacherEnv',
@ -119,6 +123,16 @@ register(
}
)
register(
id='SimpleReacher-v1',
entry_point='alr_envs.classic_control:SimpleReacherEnv',
max_episode_steps=200,
kwargs={
"n_links": 2,
"random_start": False
}
)
register(
id='LongSimpleReacher-v0',
entry_point='alr_envs.classic_control:SimpleReacherEnv',
@ -128,6 +142,18 @@ register(
}
)
register(
id='LongSimpleReacher-v1',
entry_point='alr_envs.classic_control:SimpleReacherEnv',
max_episode_steps=200,
kwargs={
"n_links": 5,
"random_start": False
}
)
## Viapoint Reacher
register(
id='ViaPointReacher-v0',
entry_point='alr_envs.classic_control.viapoint_reacher:ViaPointReacher',
@ -139,14 +165,47 @@ register(
}
)
## Hole Reacher
register(
id='HoleReacher-v0',
entry_point='alr_envs.classic_control.hole_reacher:HoleReacher',
entry_point='alr_envs.classic_control.hole_reacher:HoleReacherEnv',
max_episode_steps=200,
kwargs={
"n_links": 5,
"allow_self_collision": False,
"allow_wall_collision": False,
"hole_width": None,
"hole_depth": 1,
"hole_x": None,
"collision_penalty": 100,
}
)
register(
id='HoleReacher-v1',
entry_point='alr_envs.classic_control.hole_reacher:HoleReacherEnv',
max_episode_steps=200,
kwargs={
"n_links": 5,
"random_start": False,
"allow_self_collision": False,
"allow_wall_collision": False,
"hole_width": None,
"hole_depth": 1,
"hole_x": None,
"collision_penalty": 100,
}
)
register(
id='HoleReacher-v2',
entry_point='alr_envs.classic_control.hole_reacher:HoleReacherEnv',
max_episode_steps=200,
kwargs={
"n_links": 5,
"random_start": False,
"allow_self_collision": False,
"allow_wall_collision": False,
"hole_width": 0.25,
"hole_depth": 1,
"hole_x": 2,
@ -155,6 +214,25 @@ register(
)
# MP environments
## Simple Reacher
versions = ["SimpleReacher-v0", "SimpleReacher-v1", "LongSimpleReacher-v0", "LongSimpleReacher-v1"]
for v in versions:
name = v.split("-")
register(
id=f'{name[0]}DMP-{name[1]}',
entry_point='alr_envs.utils.make_env_helpers:make_dmp_env',
# max_episode_steps=1,
kwargs={
"name": f"alr_envs:{v}",
"num_dof": 2 if "long" not in v.lower() else 5,
"num_basis": 5,
"duration": 2,
"alpha_phase": 2,
"learn_goal": True,
"policy_type": "velocity",
"weights_scale": 50,
}
)
register(
id='ViaPointReacherDMP-v0',
@ -172,23 +250,33 @@ register(
}
)
register(
id='HoleReacherDMP-v0',
entry_point='alr_envs.utils.make_env_helpers:make_dmp_env',
# max_episode_steps=1,
kwargs={
"name": "alr_envs:HoleReacher-v0",
"num_dof": 5,
"num_basis": 5,
"duration": 2,
"learn_goal": True,
"alpha_phase": 2,
"bandwidth_factor": 2,
"policy_type": "velocity",
"weights_scale": 50,
"goal_scale": 0.1
}
)
## Hole Reacher
versions = ["v0", "v1", "v2"]
for v in versions:
register(
id=f'HoleReacherDMP-{v}',
entry_point='alr_envs.utils.make_env_helpers:make_dmp_env',
# max_episode_steps=1,
kwargs={
"name": f"alr_envs:HoleReacher-{v}",
"num_dof": 5,
"num_basis": 5,
"duration": 2,
"learn_goal": True,
"alpha_phase": 2,
"bandwidth_factor": 2,
"policy_type": "velocity",
"weights_scale": 50,
"goal_scale": 0.1
}
)
# register(
# id='HoleReacherDetPMP-v0',
# entry_point='alr_envs.classic_control.hole_reacher:holereacher_detpmp',
# # max_episode_steps=1,
# # TODO: add mp kwargs
# )
# TODO: properly add final_pos
register(
@ -208,12 +296,7 @@ register(
}
)
# register(
# id='HoleReacherDetPMP-v0',
# entry_point='alr_envs.classic_control.hole_reacher:holereacher_detpmp',
# # max_episode_steps=1,
# # TODO: add mp kwargs
# )
## Ball in Cup
register(
id='ALRBallInACupSimpleDMP-v0',

2
alr_envs/classic_control/__init__.py
View File

@ -1,3 +1,3 @@
from alr_envs.classic_control.simple_reacher import SimpleReacherEnv
from alr_envs.classic_control.viapoint_reacher import ViaPointReacher
from alr_envs.classic_control.hole_reacher import HoleReacher
from alr_envs.classic_control.hole_reacher import HoleReacherEnv

314
alr_envs/classic_control/hole_reacher.py
View File

@ -1,27 +1,36 @@
from typing import Union
import gym
import numpy as np
import matplotlib.pyplot as plt
import numpy as np
from gym.utils import seeding
from matplotlib import patches
from alr_envs.classic_control.utils import check_self_collision
from alr_envs.utils.mps.mp_environments import MPEnv
class HoleReacher(gym.Env):
class HoleReacherEnv(MPEnv):
def __init__(self, n_links, hole_x, hole_width, hole_depth, allow_self_collision=False,
allow_wall_collision=False, collision_penalty=1000):
def __init__(self, n_links: int, hole_x: Union[None, float] = None, hole_depth: Union[None, float] = None,
hole_width: float = 1., random_start: bool = False, allow_self_collision: bool = False,
allow_wall_collision: bool = False, collision_penalty: bool = 1000):
self.n_links = n_links
self.link_lengths = np.ones((n_links, 1))
# task
self.hole_x = hole_x # x-position of center of hole
self.hole_width = hole_width # width of hole
self.hole_depth = hole_depth # depth of hole
self.random_start = random_start
self.bottom_center_of_hole = np.hstack([hole_x, -hole_depth])
self.top_center_of_hole = np.hstack([hole_x, 0])
self.left_wall_edge = np.hstack([hole_x - self.hole_width / 2, 0])
self.right_wall_edge = np.hstack([hole_x + self.hole_width / 2, 0])
# provided initial parameters
self._hole_x = hole_x # x-position of center of hole
self._hole_width = hole_width # width of hole
self._hole_depth = hole_depth # depth of hole
# temp container for current env state
self._tmp_hole_x = None
self._tmp_hole_width = None
self._tmp_hole_depth = None
self._goal = None # x-y coordinates for reaching the center at the bottom of the hole
# collision
self.allow_self_collision = allow_self_collision
@ -32,91 +41,77 @@ class HoleReacher(gym.Env):
self._joints = None
self._joint_angles = None
self._angle_velocity = None
self.start_pos = np.hstack([[np.pi / 2], np.zeros(self.n_links - 1)])
self.start_vel = np.zeros(self.n_links)
self._start_pos = np.hstack([[np.pi / 2], np.zeros(self.n_links - 1)])
self._start_vel = np.zeros(self.n_links)
self.dt = 0.01
# self.time_limit = 2
action_bound = np.pi * np.ones((self.n_links,))
state_bound = np.hstack([
[np.pi] * self.n_links, # cos
[np.pi] * self.n_links, # sin
[np.inf] * self.n_links, # velocity
[np.inf], # hole width
[np.inf], # hole depth
[np.inf] * 2, # x-y coordinates of target distance
[np.inf] # env steps, because reward start after n steps TODO: Maybe
])
self.action_space = gym.spaces.Box(low=-action_bound, high=action_bound, shape=action_bound.shape)
self.observation_space = gym.spaces.Box(low=-state_bound, high=state_bound, shape=state_bound.shape)
# containers for plotting
self.metadata = {'render.modes': ["human", "partial"]}
self.fig = None
rect_1 = patches.Rectangle((-self.n_links, -1),
self.n_links + self.hole_x - self.hole_width / 2, 1,
fill=True, edgecolor='k', facecolor='k')
rect_2 = patches.Rectangle((self.hole_x + self.hole_width / 2, -1),
self.n_links - self.hole_x + self.hole_width / 2, 1,
fill=True, edgecolor='k', facecolor='k')
rect_3 = patches.Rectangle((self.hole_x - self.hole_width / 2, -1), self.hole_width,
1 - self.hole_depth,
fill=True, edgecolor='k', facecolor='k')
self.patches = [rect_1, rect_2, rect_3]
self._steps = 0
self.seed()
@property
def end_effector(self):
return self._joints[self.n_links].T
def step(self, action: np.ndarray):
"""
A single step with an action in joint velocity space
"""
def configure(self, context):
pass
self._angle_velocity = action
self._joint_angles = self._joint_angles + self.dt * self._angle_velocity
self._update_joints()
acc = (action - self._angle_velocity) / self.dt
reward, info = self._get_reward(acc)
info.update({"is_collided": self._is_collided})
self._steps += 1
done = self._is_collided
return self._get_obs().copy(), reward, done, info
def reset(self):
self._joint_angles = self.start_pos
self._angle_velocity = self.start_vel
if self.random_start:
# Maybe change more than dirst seed
first_joint = self.np_random.uniform(np.pi / 4, 3 * np.pi / 4)
self._joint_angles = np.hstack([[first_joint], np.zeros(self.n_links - 1)])
self._start_pos = self._joint_angles.copy()
else:
self._joint_angles = self._start_pos
self._generate_hole()
self._set_patches()
self._angle_velocity = self._start_vel
self._joints = np.zeros((self.n_links + 1, 2))
self._update_joints()
self._steps = 0
return self._get_obs().copy()
def step(self, action: np.ndarray):
"""
a single step with an action in joint velocity space
"""
vel = action # + 0.01 * np.random.randn(self.num_links)
acc = (vel - self._angle_velocity) / self.dt
self._angle_velocity = vel
self._joint_angles = self._joint_angles + self.dt * self._angle_velocity
self._update_joints()
# rew = self._reward()
# compute reward directly in step function
success = False
reward = 0
if not self._is_collided:
if self._steps == 199:
dist = np.linalg.norm(self.end_effector - self.bottom_center_of_hole)
reward = - dist ** 2
success = dist < 0.005
else:
dist = np.linalg.norm(self.end_effector - self.bottom_center_of_hole)
# if self.collision_penalty != 0:
# reward = -self.collision_penalty
# else:
reward = - dist ** 2 - self.collision_penalty
reward -= 5e-8 * np.sum(acc ** 2)
info = {"is_collided": self._is_collided, "is_success": success}
self._steps += 1
# done = self._steps * self.dt > self.time_limit or self._is_collided
done = self._is_collided
return self._get_obs().copy(), reward, done, info
def _generate_hole(self):
self._tmp_hole_x = self.np_random.uniform(0.5, 3.5, 1) if self._hole_x is None else np.copy(self._hole_x)
self._tmp_hole_width = self.np_random.uniform(0.5, 0.1, 1) if self._hole_width is None else np.copy(
self._hole_width)
# TODO we do not want this right now.
self._tmp_hole_depth = self.np_random.uniform(1, 1, 1) if self._hole_depth is None else np.copy(
self._hole_depth)
self._goal = np.hstack([self._tmp_hole_x, -self._tmp_hole_depth])
def _update_joints(self):
"""
@ -124,7 +119,7 @@ class HoleReacher(gym.Env):
Returns:
"""
line_points_in_taskspace = self.get_forward_kinematics(num_points_per_link=20)
line_points_in_taskspace = self._get_forward_kinematics(num_points_per_link=20)
self._joints[1:, 0] = self._joints[0, 0] + line_points_in_taskspace[:, -1, 0]
self._joints[1:, 1] = self._joints[0, 1] + line_points_in_taskspace[:, -1, 1]
@ -138,48 +133,65 @@ class HoleReacher(gym.Env):
self_collision = True
if not self.allow_wall_collision:
wall_collision = self.check_wall_collision(line_points_in_taskspace)
wall_collision = self._check_wall_collision(line_points_in_taskspace)
self._is_collided = self_collision or wall_collision
def _get_reward(self, acc: np.ndarray):
success = False
reward = -np.inf
if not self._is_collided:
dist = 0
# return reward only in last time step
if self._steps == 199:
dist = np.linalg.norm(self.end_effector - self._goal)
success = dist < 0.005
else:
# Episode terminates when colliding, hence return reward
dist = np.linalg.norm(self.end_effector - self._goal)
reward = -self.collision_penalty
reward -= dist ** 2
reward -= 5e-8 * np.sum(acc ** 2)
info = {"is_success": success}
return reward, info
def _get_obs(self):
theta = self._joint_angles
return np.hstack([
np.cos(theta),
np.sin(theta),
self._angle_velocity,
self.end_effector - self.bottom_center_of_hole,
self._tmp_hole_width,
self._tmp_hole_depth,
self.end_effector - self._goal,
self._steps
])
def get_forward_kinematics(self, num_points_per_link=1):
def _get_forward_kinematics(self, num_points_per_link=1):
theta = self._joint_angles[:, None]
if num_points_per_link > 1:
intermediate_points = np.linspace(0, 1, num_points_per_link)
else:
intermediate_points = 1
intermediate_points = np.linspace(0, 1, num_points_per_link) if num_points_per_link > 1 else 1
accumulated_theta = np.cumsum(theta, axis=0)
endeffector = np.zeros(shape=(self.n_links, num_points_per_link, 2))
end_effector = np.zeros(shape=(self.n_links, num_points_per_link, 2))
x = np.cos(accumulated_theta) * self.link_lengths * intermediate_points
y = np.sin(accumulated_theta) * self.link_lengths * intermediate_points
endeffector[0, :, 0] = x[0, :]
endeffector[0, :, 1] = y[0, :]
end_effector[0, :, 0] = x[0, :]
end_effector[0, :, 1] = y[0, :]
for i in range(1, self.n_links):
endeffector[i, :, 0] = x[i, :] + endeffector[i - 1, -1, 0]
endeffector[i, :, 1] = y[i, :] + endeffector[i - 1, -1, 1]
end_effector[i, :, 0] = x[i, :] + end_effector[i - 1, -1, 0]
end_effector[i, :, 1] = y[i, :] + end_effector[i - 1, -1, 1]
return np.squeeze(endeffector + self._joints[0, :])
return np.squeeze(end_effector + self._joints[0, :])
def check_wall_collision(self, line_points):
def _check_wall_collision(self, line_points):
# all points that are before the hole in x
r, c = np.where(line_points[:, :, 0] < (self.hole_x - self.hole_width / 2))
r, c = np.where(line_points[:, :, 0] < (self._tmp_hole_x - self._tmp_hole_width / 2))
# check if any of those points are below surface
nr_line_points_below_surface_before_hole = np.sum(line_points[r, c, 1] < 0)
@ -188,7 +200,7 @@ class HoleReacher(gym.Env):
return True
# all points that are after the hole in x
r, c = np.where(line_points[:, :, 0] > (self.hole_x + self.hole_width / 2))
r, c = np.where(line_points[:, :, 0] > (self._tmp_hole_x + self._tmp_hole_width / 2))
# check if any of those points are below surface
nr_line_points_below_surface_after_hole = np.sum(line_points[r, c, 1] < 0)
@ -197,11 +209,11 @@ class HoleReacher(gym.Env):
return True
# all points that are above the hole
r, c = np.where((line_points[:, :, 0] > (self.hole_x - self.hole_width / 2)) & (
line_points[:, :, 0] < (self.hole_x + self.hole_width / 2)))
r, c = np.where((line_points[:, :, 0] > (self._tmp_hole_x - self._tmp_hole_width / 2)) & (
line_points[:, :, 0] < (self._tmp_hole_x + self._tmp_hole_width / 2)))
# check if any of those points are below surface
nr_line_points_below_surface_in_hole = np.sum(line_points[r, c, 1] < -self.hole_depth)
nr_line_points_below_surface_in_hole = np.sum(line_points[r, c, 1] < -self._tmp_hole_depth)
if nr_line_points_below_surface_in_hole > 0:
return True
@ -210,61 +222,85 @@ class HoleReacher(gym.Env):
def render(self, mode='human'):
if self.fig is None:
# Create base figure once on the beginning. Afterwards only update
plt.ion()
self.fig = plt.figure()
# plt.ion()
# plt.pause(0.01)
else:
plt.figure(self.fig.number)
ax = self.fig.add_subplot(1, 1, 1)
# limits
lim = np.sum(self.link_lengths) + 0.5
ax.set_xlim([-lim, lim])
ax.set_ylim([-1.1, lim])
self.line, = ax.plot(self._joints[:, 0], self._joints[:, 1], 'ro-', markerfacecolor='k')
self._set_patches()
self.fig.show()
self.fig.gca().set_title(
f"Iteration: {self._steps}, distance: {self.end_effector - self._goal}")
if mode == "human":
plt.cla()
plt.title(f"Iteration: {self._steps}, distance: {self.end_effector - self.bottom_center_of_hole}")
# Arm
plt.plot(self._joints[:, 0], self._joints[:, 1], 'ro-', markerfacecolor='k')
# arm
self.line.set_data(self._joints[:, 0], self._joints[:, 1])
# Add the patch to the Axes
[plt.gca().add_patch(rect) for rect in self.patches]
lim = np.sum(self.link_lengths) + 0.5
plt.xlim([-lim, lim])
plt.ylim([-1.1, lim])
# plt.draw()
plt.pause(1e-4) # pushes window to foreground, which is annoying.
# self.fig.canvas.flush_events()
self.fig.canvas.draw()
self.fig.canvas.flush_events()
elif mode == "partial":
if self._steps == 1:
# fig, ax = plt.subplots()
# Add the patch to the Axes
[plt.gca().add_patch(rect) for rect in self.patches]
# plt.pause(0.01)
if self._steps % 20 == 0 or self._steps in [1, 199] or self._is_collided:
# Arm
plt.plot(self._joints[:, 0], self._joints[:, 1], 'ro-', markerfacecolor='k', alpha=self._steps / 200)
# ax.plot(line_points_in_taskspace[:, 0, 0],
# line_points_in_taskspace[:, 0, 1],
# line_points_in_taskspace[:, -1, 0],
# line_points_in_taskspace[:, -1, 1], marker='o', color='k', alpha=t / 200)
plt.plot(self._joints[:, 0], self._joints[:, 1], 'ro-', markerfacecolor='k',
alpha=self._steps / 200)
lim = np.sum(self.link_lengths) + 0.5
plt.xlim([-lim, lim])
plt.ylim([-1.1, lim])
plt.pause(0.01)
def _set_patches(self):
if self.fig is not None:
self.fig.gca().patches = []
left_block = patches.Rectangle((-self.n_links, -self._tmp_hole_depth),
self.n_links + self._tmp_hole_x - self._tmp_hole_width / 2,
self._tmp_hole_depth,
fill=True, edgecolor='k', facecolor='k')
right_block = patches.Rectangle((self._tmp_hole_x + self._tmp_hole_width / 2, -self._tmp_hole_depth),
self.n_links - self._tmp_hole_x + self._tmp_hole_width / 2,
self._tmp_hole_depth,
fill=True, edgecolor='k', facecolor='k')
hole_floor = patches.Rectangle((self._tmp_hole_x - self._tmp_hole_width / 2, -self._tmp_hole_depth),
self._tmp_hole_width,
1 - self._tmp_hole_depth,
fill=True, edgecolor='k', facecolor='k')
elif mode == "final":
if self._steps == 199 or self._is_collided:
# fig, ax = plt.subplots()
# Add the patch to the Axes
self.fig.gca().add_patch(left_block)
self.fig.gca().add_patch(right_block)
self.fig.gca().add_patch(hole_floor)
# Add the patch to the Axes
[plt.gca().add_patch(rect) for rect in self.patches]
@property
def active_obs(self):
return np.hstack([
[self.random_start] * self.n_links, # cos
[self.random_start] * self.n_links, # sin
[self.random_start] * self.n_links, # velocity
[self._hole_width is None], # hole width
[self._hole_depth is None], # hole width
[True] * 2, # x-y coordinates of target distance
[False] # env steps
])
plt.xlim(-self.n_links, self.n_links), plt.ylim(-1, self.n_links)
# Arm
plt.plot(self._joints[:, 0], self._joints[:, 1], 'ro-', markerfacecolor='k')
@property
def start_pos(self) -> Union[float, int, np.ndarray]:
return self._start_pos
plt.pause(0.01)
@property
def goal_pos(self) -> Union[float, int, np.ndarray]:
raise ValueError("Goal position is not available and has to be learnt based on the environment.")
def seed(self, seed=None):
self.np_random, seed = seeding.np_random(seed)
return [seed]
@property
def end_effector(self):
return self._joints[self.n_links].T
def close(self):
if self.fig is not None:
@ -274,22 +310,20 @@ class HoleReacher(gym.Env):
if __name__ == '__main__':
nl = 5
render_mode = "human" # "human" or "partial" or "final"
env = HoleReacher(n_links=nl, allow_self_collision=False, allow_wall_collision=False, hole_width=0.15,
hole_depth=1, hole_x=1)
env.reset()
# env.render(mode=render_mode)
env = HoleReacherEnv(n_links=nl, allow_self_collision=False, allow_wall_collision=False, hole_width=None,
hole_depth=1, hole_x=None)
obs = env.reset()
for i in range(200):
# objective.load_result("/tmp/cma")
# test with random actions
ac = 2 * env.action_space.sample()
# ac[0] += np.pi/2
obs, rew, d, info = env.step(ac)
env.render(mode=render_mode)
print(rew)
if d:
break
env.reset()
env.close()

216
alr_envs/classic_control/simple_reacher.py
View File

@ -1,39 +1,41 @@
import gym
from typing import Iterable, Union
import matplotlib.pyplot as plt
import numpy as np
from gym import spaces
from gym.utils import seeding
from alr_envs.utils.utils import angle_normalize
from alr_envs.utils.mps.mp_environments import MPEnv
# if os.environ.get("DISPLAY", None):
# mpl.use('Qt5Agg')
class SimpleReacherEnv(gym.Env):
class SimpleReacherEnv(MPEnv):
"""
Simple Reaching Task without any physics simulation.
Returns no reward until 150 time steps. This allows the agent to explore the space, but requires precise actions
towards the end of the trajectory.
"""
def __init__(self, n_links):
def __init__(self, n_links: int, target: Union[None, Iterable] = None, random_start: bool = True):
super().__init__()
self.link_lengths = np.ones(n_links)
self.n_links = n_links
self.dt = 0.1
self._goal_pos = None
self.random_start = random_start
self._joints = None
self._joint_angle = None
self._joint_angles = None
self._angle_velocity = None
self._start_pos = np.zeros(self.n_links)
self._start_vel = np.zeros(self.n_links)
self.max_torque = 1 # 10
self._target = target # provided target value
self._goal = None # updated goal value, does not change when target != None
self.max_torque = 1
self.steps_before_reward = 199
action_bound = np.ones((self.n_links,))
action_bound = np.ones((self.n_links,)) * self.max_torque
state_bound = np.hstack([
[np.pi] * self.n_links, # cos
[np.pi] * self.n_links, # sin
@ -44,49 +46,50 @@ class SimpleReacherEnv(gym.Env):
self.action_space = spaces.Box(low=-action_bound, high=action_bound, shape=action_bound.shape)
self.observation_space = spaces.Box(low=-state_bound, high=state_bound, shape=state_bound.shape)
self.fig = None
# containers for plotting
self.metadata = {'render.modes': ["human"]}
self.fig = None
self._steps = 0
self.seed()
def step(self, action: np.ndarray):
"""
A single step with action in torque space
"""
# action = self._add_action_noise(action)
action = np.clip(action, -self.max_torque, self.max_torque)
ac = np.clip(action, -self.max_torque, self.max_torque)
self._angle_velocity = self._angle_velocity + self.dt * action
self._joint_angle = angle_normalize(self._joint_angle + self.dt * self._angle_velocity)
self._angle_velocity = self._angle_velocity + self.dt * ac
self._joint_angles = self._joint_angles + self.dt * self._angle_velocity
self._update_joints()
self._steps += 1
reward, info = self._get_reward(action)
# done = np.abs(self.end_effector - self._goal_pos) < 0.1
self._steps += 1
done = False
return self._get_obs().copy(), reward, done, info
def _add_action_noise(self, action: np.ndarray):
"""
add unobserved Gaussian Noise N(0,0.01) to the actions
Args:
action:
def reset(self):
Returns: actions with noise
# TODO: maybe do initialisation more random?
# Sample only orientation of first link, i.e. the arm is always straight.
if self.random_start:
self._joint_angles = np.hstack([[self.np_random.uniform(-np.pi, np.pi)], np.zeros(self.n_links - 1)])
self._start_pos = self._joint_angles.copy()
else:
self._joint_angles = self._start_pos
"""
return self.np_random.normal(0, 0.1, *action.shape) + action
self._generate_goal()
def _get_obs(self):
theta = self._joint_angle
return np.hstack([
np.cos(theta),
np.sin(theta),
self._angle_velocity,
self.end_effector - self._goal_pos,
self._steps
])
self._angle_velocity = self._start_vel
self._joints = np.zeros((self.n_links + 1, 2))
self._update_joints()
self._steps = 0
return self._get_obs().copy()
def _update_joints(self):
"""
@ -94,15 +97,14 @@ class SimpleReacherEnv(gym.Env):
Returns:
"""
angles = np.cumsum(self._joint_angle)
angles = np.cumsum(self._joint_angles)
x = self.link_lengths * np.vstack([np.cos(angles), np.sin(angles)])
self._joints[1:] = self._joints[0] + np.cumsum(x.T, axis=0)
def _get_reward(self, action: np.ndarray):
diff = self.end_effector - self._goal_pos
diff = self.end_effector - self._goal
reward_dist = 0
# TODO: Is this the best option
if self._steps >= self.steps_before_reward:
reward_dist -= np.linalg.norm(diff)
# reward_dist = np.exp(-0.1 * diff ** 2).mean()
@ -112,62 +114,112 @@ class SimpleReacherEnv(gym.Env):
reward = reward_dist - reward_ctrl
return reward, dict(reward_dist=reward_dist, reward_ctrl=reward_ctrl)
def reset(self):
def _get_obs(self):
theta = self._joint_angles
return np.hstack([
np.cos(theta),
np.sin(theta),
self._angle_velocity,
self.end_effector - self._goal,
self._steps
])
# TODO: maybe do initialisation more random?
# Sample only orientation of first link, i.e. the arm is always straight.
self._joint_angle = np.hstack([[self.np_random.uniform(-np.pi, np.pi)], np.zeros(self.n_links - 1)])
self._angle_velocity = np.zeros(self.n_links)
self._joints = np.zeros((self.n_links + 1, 2))
self._update_joints()
self._steps = 0
def _generate_goal(self):
self._goal_pos = self._get_random_goal()
return self._get_obs().copy()
if self._target is None:
def _get_random_goal(self):
center = self._joints[0]
total_length = np.sum(self.link_lengths)
goal = np.array([total_length, total_length])
while np.linalg.norm(goal) >= total_length:
goal = self.np_random.uniform(low=-total_length, high=total_length, size=2)
else:
goal = np.copy(self._target)
# Sample uniformly in circle with radius R around center of reacher.
R = np.sum(self.link_lengths)
r = R * np.sqrt(self.np_random.uniform())
theta = self.np_random.uniform() * 2 * np.pi
return center + r * np.stack([np.cos(theta), np.sin(theta)])
self._goal = goal
def render(self, mode='human'): # pragma: no cover
if self.fig is None:
# Create base figure once on the beginning. Afterwards only update
plt.ion()
self.fig = plt.figure()
ax = self.fig.add_subplot(1, 1, 1)
# limits
lim = np.sum(self.link_lengths) + 0.5
ax.set_xlim([-lim, lim])
ax.set_ylim([-lim, lim])
self.line, = ax.plot(self._joints[:, 0], self._joints[:, 1], 'ro-', markerfacecolor='k')
goal_pos = self._goal.T
self.goal_point, = ax.plot(goal_pos[0], goal_pos[1], 'gx')
self.goal_dist, = ax.plot([self.end_effector[0], goal_pos[0]], [self.end_effector[1], goal_pos[1]], 'g--')
self.fig.show()
self.fig.gca().set_title(f"Iteration: {self._steps}, distance: {self.end_effector - self._goal}")
# goal
goal_pos = self._goal.T
if self._steps == 1:
self.goal_point.set_data(goal_pos[0], goal_pos[1])
# arm
self.line.set_data(self._joints[:, 0], self._joints[:, 1])
# distance between end effector and goal
self.goal_dist.set_data([self.end_effector[0], goal_pos[0]], [self.end_effector[1], goal_pos[1]])
self.fig.canvas.draw()
self.fig.canvas.flush_events()
@property
def active_obs(self):
return np.hstack([
[self.random_start] * self.n_links, # cos
[self.random_start] * self.n_links, # sin
[self.random_start] * self.n_links, # velocity
[True] * 2, # x-y coordinates of target distance
[False] # env steps
])
@property
def start_pos(self):
return self._start_pos
@property
def goal_pos(self):
raise ValueError("Goal position is not available and has to be learnt based on the environment.")
def seed(self, seed=None):
self.np_random, seed = seeding.np_random(seed)
return [seed]
def render(self, mode='human'): # pragma: no cover
if self.fig is None:
self.fig = plt.figure()
plt.ion()
plt.show()
else:
plt.figure(self.fig.number)
plt.cla()
plt.title(f"Iteration: {self._steps}, distance: {self.end_effector - self._goal_pos}")
# Arm
plt.plot(self._joints[:, 0], self._joints[:, 1], 'ro-', markerfacecolor='k')
# goal
goal_pos = self._goal_pos.T
plt.plot(goal_pos[0], goal_pos[1], 'gx')
# distance between end effector and goal
plt.plot([self.end_effector[0], goal_pos[0]], [self.end_effector[1], goal_pos[1]], 'g--')
lim = np.sum(self.link_lengths) + 0.5
plt.xlim([-lim, lim])
plt.ylim([-lim, lim])
# plt.draw()
# plt.pause(1e-4) pushes window to foreground, which is annoying.
self.fig.canvas.flush_events()
def close(self):
del self.fig
@property
def end_effector(self):
return self._joints[self.n_links].T
if __name__ == '__main__':
nl = 5
render_mode = "human" # "human" or "partial" or "final"
env = SimpleReacherEnv(n_links=nl)
obs = env.reset()
print("First", obs)
for i in range(2000):
# objective.load_result("/tmp/cma")
# test with random actions
ac = 2 * env.action_space.sample()
# ac = np.ones(env.action_space.shape)
obs, rew, d, info = env.step(ac)
env.render(mode=render_mode)
print(obs[env.active_obs].shape)
if d or i % 200 == 0:
env.reset()
env.close()

225
alr_envs/classic_control/viapoint_reacher.py
View File

@ -1,19 +1,31 @@
from typing import Iterable, Union
import gym
import matplotlib.pyplot as plt
import numpy as np
from gym.utils import seeding
from alr_envs.classic_control.utils import check_self_collision
from alr_envs.utils.mps.mp_environments import MPEnv
class ViaPointReacher(gym.Env):
class ViaPointReacher(MPEnv):
def __init__(self, n_links, allow_self_collision=False, collision_penalty=1000):
self.num_links = n_links
def __init__(self, n_links, random_start: bool = True, via_target: Union[None, Iterable] = None,
target: Union[None, Iterable] = None, allow_self_collision=False, collision_penalty=1000):
self.n_links = n_links
self.link_lengths = np.ones((n_links, 1))
# task
self.via_point = np.ones(2)
self.goal_point = np.array((n_links, 0))
self.random_start = random_start
# provided initial parameters
self._target = target # provided target value
self._via_target = via_target # provided via point target value
# temp container for current env state
self._via_point = np.ones(2)
self._goal = np.array((n_links, 0))
# collision
self.allow_self_collision = allow_self_collision
@ -23,78 +35,93 @@ class ViaPointReacher(gym.Env):
self._joints = None
self._joint_angles = None
self._angle_velocity = None
self.start_pos = np.hstack([[np.pi / 2], np.zeros(self.num_links - 1)])
self.start_vel = np.zeros(self.num_links)
self._start_pos = np.hstack([[np.pi / 2], np.zeros(self.n_links - 1)])
self._start_vel = np.zeros(self.n_links)
self.weight_matrix_scale = 1
self._steps = 0
self.dt = 0.01
# self.time_limit = 2
action_bound = np.pi * np.ones((self.num_links,))
action_bound = np.pi * np.ones((self.n_links,))
state_bound = np.hstack([
[np.pi] * self.num_links, # cos
[np.pi] * self.num_links, # sin
[np.inf] * self.num_links, # velocity
[np.pi] * self.n_links, # cos
[np.pi] * self.n_links, # sin
[np.inf] * self.n_links, # velocity
[np.inf] * 2, # x-y coordinates of via point distance
[np.inf] * 2, # x-y coordinates of target distance
[np.inf] # env steps, because reward start after n steps TODO: Maybe
[np.inf] # env steps, because reward start after n steps
])
self.action_space = gym.spaces.Box(low=-action_bound, high=action_bound, shape=action_bound.shape)
self.observation_space = gym.spaces.Box(low=-state_bound, high=state_bound, shape=state_bound.shape)
# containers for plotting
self.metadata = {'render.modes': ["human", "partial"]}
self.fig = None
@property
def end_effector(self):
return self._joints[self.num_links].T
def configure(self, context):
pass
def reset(self):
self._joint_angles = self.start_pos
self._angle_velocity = self.start_vel
self._joints = np.zeros((self.num_links + 1, 2))
self._update_joints()
self._steps = 0
return self._get_obs().copy()
self.seed()
def step(self, action: np.ndarray):
"""
a single step with an action in joint velocity space
"""
vel = action
acc = (vel - self._angle_velocity) / self.dt
self._angle_velocity = vel
self._joint_angles = self._joint_angles + self.dt * self._angle_velocity
self._update_joints()
dist_reward = 0
if not self._is_collided:
if self._steps == 100:
dist_reward = np.linalg.norm(self.end_effector - self.via_point)
elif self._steps == 199:
dist_reward = np.linalg.norm(self.end_effector - self.goal_point)
acc = (vel - self._angle_velocity) / self.dt
reward, info = self._get_reward(acc)
# TODO: Do we need that?
reward = - dist_reward ** 2
reward -= 5e-8 * np.sum(acc**2)
if self._is_collided:
reward -= self.collision_penalty
info = {"is_collided": self._is_collided}
info.update({"is_collided": self._is_collided})
self._steps += 1
# done = self._steps * self.dt > self.time_limit or self._is_collided
done = self._is_collided
return self._get_obs().copy(), reward, done, info
def reset(self):
if self.random_start:
# Maybe change more than dirst seed
first_joint = self.np_random.uniform(np.pi / 4, 3 * np.pi / 4)
self._joint_angles = np.hstack([[first_joint], np.zeros(self.n_links - 1)])
self._start_pos = self._joint_angles.copy()
else:
self._joint_angles = self._start_pos
self._generate_goal()
self._angle_velocity = self._start_vel
self._joints = np.zeros((self.n_links + 1, 2))
self._update_joints()
self._steps = 0
return self._get_obs().copy()
def _generate_goal(self):
# TODO: Maybe improve this later, this can yield quite a lot of invalid settings
total_length = np.sum(self.link_lengths)
# rejection sampled point in inner circle with 0.5*Radius
if self._via_target is None:
via_target = np.array([total_length, total_length])
while np.linalg.norm(via_target) >= 0.5 * total_length:
via_target = self.np_random.uniform(low=-0.5 * total_length, high=0.5 * total_length, size=2)
else:
via_target = np.copy(self._via_target)
# rejection sampled point in outer circle
if self._target is None:
goal = np.array([total_length, total_length])
while np.linalg.norm(goal) >= total_length or np.linalg.norm(goal) <= 0.5 * total_length:
goal = self.np_random.uniform(low=-total_length, high=total_length, size=2)
else:
goal = np.copy(self._target)
self._via_target = via_target
self._goal = goal
def _update_joints(self):
"""
update _joints to get new end effector position. The other links are only required for rendering.
@ -115,14 +142,38 @@ class ViaPointReacher(gym.Env):
self._is_collided = self_collision
def _get_reward(self, acc):
success = False
reward = -np.inf
if not self._is_collided:
dist = np.inf
# return intermediate reward for via point
if self._steps == 100:
dist = np.linalg.norm(self.end_effector - self._via_point)
# return reward in last time step for goal
elif self._steps == 199:
dist = np.linalg.norm(self.end_effector - self._goal)
success = dist < 0.005
else:
# Episode terminates when colliding, hence return reward
dist = np.linalg.norm(self.end_effector - self._goal)
reward = -self.collision_penalty
reward -= dist ** 2
reward -= 5e-8 * np.sum(acc ** 2)
info = {"is_success": success}
return reward, info
def _get_obs(self):
theta = self._joint_angles
return np.hstack([
np.cos(theta),
np.sin(theta),
self._angle_velocity,
self.end_effector - self.via_point,
self.end_effector - self.goal_point,
self.end_effector - self._via_point,
self.end_effector - self._goal,
self._steps
])
@ -133,7 +184,7 @@ class ViaPointReacher(gym.Env):
accumulated_theta = np.cumsum(theta, axis=0)
endeffector = np.zeros(shape=(self.num_links, num_points_per_link, 2))
endeffector = np.zeros(shape=(self.n_links, num_points_per_link, 2))
x = np.cos(accumulated_theta) * self.link_lengths * intermediate_points
y = np.sin(accumulated_theta) * self.link_lengths * intermediate_points
@ -141,33 +192,46 @@ class ViaPointReacher(gym.Env):
endeffector[0, :, 0] = x[0, :]
endeffector[0, :, 1] = y[0, :]
for i in range(1, self.num_links):
for i in range(1, self.n_links):
endeffector[i, :, 0] = x[i, :] + endeffector[i - 1, -1, 0]
endeffector[i, :, 1] = y[i, :] + endeffector[i - 1, -1, 1]
return np.squeeze(endeffector + self._joints[0, :])
def render(self, mode='human'):
goal_pos = self._goal.T
via_pos = self._via_point.T
if self.fig is None:
# Create base figure once on the beginning. Afterwards only update
plt.ion()
self.fig = plt.figure()
# plt.ion()
# plt.pause(0.01)
else:
plt.figure(self.fig.number)
ax = self.fig.add_subplot(1, 1, 1)
# limits
lim = np.sum(self.link_lengths) + 0.5
ax.set_xlim([-lim, lim])
ax.set_ylim([-lim, lim])
self.line, = ax.plot(self._joints[:, 0], self._joints[:, 1], 'ro-', markerfacecolor='k')
self.goal_point_plot, = ax.plot(goal_pos[0], goal_pos[1], 'go')
self.via_point_plot, = ax.plot(via_pos[0], via_pos[1], 'gx')
self.fig.show()
self.fig.gca().set_title(f"Iteration: {self._steps}, distance: {self.end_effector - self._goal}")
if mode == "human":
plt.cla()
plt.title(f"Iteration: {self._steps}")
# goal
if self._steps == 1:
self.goal_point_plot.set_data(goal_pos[0], goal_pos[1])
self.via_point_plot.set_data(via_pos[0], goal_pos[1])
# Arm
plt.plot(self._joints[:, 0], self._joints[:, 1], 'ro-', markerfacecolor='k')
# arm
self.line.set_data(self._joints[:, 0], self._joints[:, 1])
lim = np.sum(self.link_lengths) + 0.5
plt.xlim([-lim, lim])
plt.ylim([-lim, lim])
# plt.draw()
plt.pause(1e-4) # pushes window to foreground, which is annoying.
# self.fig.canvas.flush_events()
self.fig.canvas.draw()
self.fig.canvas.flush_events()
elif mode == "partial":
if self._steps == 1:
@ -196,12 +260,39 @@ class ViaPointReacher(gym.Env):
# Add the patch to the Axes
[plt.gca().add_patch(rect) for rect in self.patches]
plt.xlim(-self.num_links, self.num_links), plt.ylim(-1, self.num_links)
plt.xlim(-self.n_links, self.n_links), plt.ylim(-1, self.n_links)
# Arm
plt.plot(self._joints[:, 0], self._joints[:, 1], 'ro-', markerfacecolor='k')
plt.pause(0.01)
@property
def active_obs(self):
return np.hstack([
[self.random_start] * self.n_links, # cos
[self.random_start] * self.n_links, # sin
[self.random_start] * self.n_links, # velocity
[self._via_target is None] * 2, # x-y coordinates of via point distance
[True] * 2, # x-y coordinates of target distance
[False] # env steps
])
@property
def start_pos(self) -> Union[float, int, np.ndarray]:
return self._start_pos
@property
def goal_pos(self) -> Union[float, int, np.ndarray]:
raise ValueError("Goal position is not available and has to be learnt based on the environment.")
def seed(self, seed=None):
self.np_random, seed = seeding.np_random(seed)
return [seed]
@property
def end_effector(self):
return self._joints[self.n_links].T
def close(self):
if self.fig is not None:
plt.close(self.fig)

50
alr_envs/mujoco/ball_in_a_cup/utils.py
View File

@ -1,5 +1,5 @@
from alr_envs.utils.wrapper.detpmp_wrapper import DetPMPWrapper
from alr_envs.utils.wrapper.dmp_wrapper import DmpWrapper
from alr_envs.utils.mps.detpmp_wrapper import DetPMPWrapper
from alr_envs.utils.mps.dmp_wrapper import DmpWrapper
from alr_envs.mujoco.ball_in_a_cup.ball_in_a_cup import ALRBallInACupEnv
@ -17,19 +17,8 @@ def make_contextual_env(rank, seed=0):
def _init():
env = ALRBallInACupEnv(reward_type="contextual_goal")
env = DetPMPWrapper(env,
num_dof=7,
num_basis=5,
width=0.005,
policy_type="motor",
start_pos=env.start_pos,
duration=3.5,
post_traj_time=4.5,
dt=env.dt,
weights_scale=0.5,
zero_start=True,
zero_goal=True
)
env = DetPMPWrapper(env, num_dof=7, num_basis=5, width=0.005, duration=3.5, dt=env.dt, post_traj_time=4.5,
policy_type="motor", weights_scale=0.5, zero_start=True, zero_goal=True)
env.seed(seed + rank)
return env
@ -51,19 +40,8 @@ def make_env(rank, seed=0):
def _init():
env = ALRBallInACupEnv(reward_type="simple")
env = DetPMPWrapper(env,
num_dof=7,
num_basis=5,
width=0.005,
policy_type="motor",
start_pos=env.start_pos,
duration=3.5,
post_traj_time=4.5,
dt=env.dt,
weights_scale=0.2,
zero_start=True,
zero_goal=True
)
env = DetPMPWrapper(env, num_dof=7, num_basis=5, width=0.005, duration=3.5, dt=env.dt, post_traj_time=4.5,
policy_type="motor", weights_scale=0.2, zero_start=True, zero_goal=True)
env.seed(seed + rank)
return env
@ -85,20 +63,8 @@ def make_simple_env(rank, seed=0):
def _init():
env = ALRBallInACupEnv(reward_type="simple")
env = DetPMPWrapper(env,
num_dof=3,
num_basis=5,
width=0.005,
off=-0.1,
policy_type="motor",
start_pos=env.start_pos[1::2],
duration=3.5,
post_traj_time=4.5,
dt=env.dt,
weights_scale=0.25,
zero_start=True,
zero_goal=True
)
env = DetPMPWrapper(env, num_dof=3, num_basis=5, width=0.005, duration=3.5, dt=env.dt, post_traj_time=4.5,
policy_type="motor", weights_scale=0.25, zero_start=True, zero_goal=True, off=-0.1)
env.seed(seed + rank)
return env

47
alr_envs/mujoco/beerpong/utils.py
View File

@ -1,4 +1,4 @@
from alr_envs.utils.wrapper.detpmp_wrapper import DetPMPWrapper
from alr_envs.utils.mps.detpmp_wrapper import DetPMPWrapper
from alr_envs.mujoco.beerpong.beerpong import ALRBeerpongEnv
from alr_envs.mujoco.beerpong.beerpong_simple import ALRBeerpongEnv as ALRBeerpongEnvSimple
@ -17,19 +17,8 @@ def make_contextual_env(rank, seed=0):
def _init():
env = ALRBeerpongEnv()
env = DetPMPWrapper(env,
num_dof=7,
num_basis=5,
width=0.005,
policy_type="motor",
start_pos=env.start_pos,
duration=3.5,
post_traj_time=4.5,
dt=env.dt,
weights_scale=0.5,
zero_start=True,
zero_goal=True
)
env = DetPMPWrapper(env, num_dof=7, num_basis=5, width=0.005, duration=3.5, dt=env.dt, post_traj_time=4.5,
policy_type="motor", weights_scale=0.5, zero_start=True, zero_goal=True)
env.seed(seed + rank)
return env
@ -51,19 +40,8 @@ def make_env(rank, seed=0):
def _init():
env = ALRBeerpongEnvSimple()
env = DetPMPWrapper(env,
num_dof=7,
num_basis=5,
width=0.005,
policy_type="motor",
start_pos=env.start_pos,
duration=3.5,
post_traj_time=4.5,
dt=env.dt,
weights_scale=0.25,
zero_start=True,
zero_goal=True
)
env = DetPMPWrapper(env, num_dof=7, num_basis=5, width=0.005, duration=3.5, dt=env.dt, post_traj_time=4.5,
policy_type="motor", weights_scale=0.25, zero_start=True, zero_goal=True)
env.seed(seed + rank)
return env
@ -85,19 +63,8 @@ def make_simple_env(rank, seed=0):
def _init():
env = ALRBeerpongEnvSimple()
env = DetPMPWrapper(env,
num_dof=3,
num_basis=5,
width=0.005,
policy_type="motor",
start_pos=env.start_pos[1::2],
duration=3.5,
post_traj_time=4.5,
dt=env.dt,
weights_scale=0.5,
zero_start=True,
zero_goal=True
)
env = DetPMPWrapper(env, num_dof=3, num_basis=5, width=0.005, duration=3.5, dt=env.dt, post_traj_time=4.5,
policy_type="motor", weights_scale=0.5, zero_start=True, zero_goal=True)
env.seed(seed + rank)
return env

6
alr_envs/mujoco/reacher/alr_reacher.py
View File

@ -2,12 +2,12 @@ import os
import numpy as np
from gym import utils
from gym.envs.mujoco import mujoco_env
from gym.envs.mujoco import MujocoEnv
import alr_envs.utils.utils as alr_utils
class ALRReacherEnv(mujoco_env.MujocoEnv, utils.EzPickle):
class ALRReacherEnv(MujocoEnv, utils.EzPickle):
def __init__(self, steps_before_reward=200, n_links=5, balance=False):
utils.EzPickle.__init__(**locals())
@ -31,7 +31,7 @@ class ALRReacherEnv(mujoco_env.MujocoEnv, utils.EzPickle):
else:
raise ValueError(f"Invalid number of links {n_links}, only 5 or 7 allowed.")
mujoco_env.MujocoEnv.__init__(self, os.path.join(os.path.dirname(__file__), "assets", file_name), 2)
MujocoEnv.__init__(self, os.path.join(os.path.dirname(__file__), "assets", file_name), 2)
def step(self, a):
self._steps += 1

88
alr_envs/utils/legacy/detpmp_env_wrapper.py
View File

@ -1,88 +0,0 @@
from alr_envs.utils.policies import get_policy_class
from mp_lib import det_promp
import numpy as np
import gym
class DetPMPEnvWrapper(gym.Wrapper):
def __init__(self,
env,
num_dof,
num_basis,
width,
off=0.01,
start_pos=None,
duration=1,
dt=0.01,
post_traj_time=0.,
policy_type=None,
weights_scale=1,
zero_start=False,
zero_goal=False,
):
super(DetPMPEnvWrapper, self).__init__(env)
self.num_dof = num_dof
self.num_basis = num_basis
self.dim = num_dof * num_basis
self.pmp = det_promp.DeterministicProMP(n_basis=num_basis, n_dof=num_dof, width=width, off=off,
zero_start=zero_start, zero_goal=zero_goal)
weights = np.zeros(shape=(num_basis, num_dof))
self.pmp.set_weights(duration, weights)
self.weights_scale = weights_scale
self.duration = duration
self.dt = dt
self.post_traj_steps = int(post_traj_time / dt)
self.start_pos = start_pos
self.zero_start = zero_start
policy_class = get_policy_class(policy_type)
self.policy = policy_class(env)
def __call__(self, params, contexts=None):
params = np.atleast_2d(params)
rewards = []
infos = []
for p, c in zip(params, contexts):
reward, info = self.rollout(p, c)
rewards.append(reward)
infos.append(info)
return np.array(rewards), infos
def rollout(self, params, context=None, render=False):
""" This function generates a trajectory based on a DMP and then does the usual loop over reset and step"""
params = np.reshape(params, newshape=(self.num_basis, self.num_dof)) * self.weights_scale
self.pmp.set_weights(self.duration, params)
t, des_pos, des_vel, des_acc = self.pmp.compute_trajectory(1 / self.dt, 1.)
if self.zero_start:
des_pos += self.start_pos[None, :]
if self.post_traj_steps > 0:
des_pos = np.vstack([des_pos, np.tile(des_pos[-1, :], [self.post_traj_steps, 1])])
des_vel = np.vstack([des_vel, np.zeros(shape=(self.post_traj_steps, self.num_dof))])
self._trajectory = des_pos
self._velocity = des_vel
rews = []
infos = []
self.env.configure(context)
self.env.reset()
for t, pos_vel in enumerate(zip(des_pos, des_vel)):
ac = self.policy.get_action(pos_vel[0], pos_vel[1])
obs, rew, done, info = self.env.step(ac)
rews.append(rew)
infos.append(info)
if render:
self.env.render(mode="human")
if done:
break
reward = np.sum(rews)
return reward, info

173
alr_envs/utils/legacy/dmp_async_vec_env.py
View File

@ -1,173 +0,0 @@
import gym
from gym.error import (AlreadyPendingCallError, NoAsyncCallError)
from gym.vector.utils import concatenate, create_empty_array
from gym.vector.async_vector_env import AsyncState
import numpy as np
import multiprocessing as mp
import sys
def _worker(index, env_fn, pipe, parent_pipe, shared_memory, error_queue):
assert shared_memory is None
env = env_fn()
parent_pipe.close()
try:
while True:
command, data = pipe.recv()
if command == 'reset':
observation = env.reset()
pipe.send((observation, True))
elif command == 'step':
observation, reward, done, info = env.step(data)
if done:
observation = env.reset()
pipe.send(((observation, reward, done, info), True))
elif command == 'rollout':
rewards = []
infos = []
for p, c in zip(*data):
reward, info = env.rollout(p, c)
rewards.append(reward)
infos.append(info)
pipe.send(((rewards, infos), (True, ) * len(rewards)))
elif command == 'seed':
env.seed(data)
pipe.send((None, True))
elif command == 'close':
env.close()
pipe.send((None, True))
break
elif command == 'idle':
pipe.send((None, True))
elif command == '_check_observation_space':
pipe.send((data == env.observation_space, True))
else:
raise RuntimeError('Received unknown command `{0}`. Must '
'be one of {`reset`, `step`, `seed`, `close`, '
'`_check_observation_space`}.'.format(command))
except (KeyboardInterrupt, Exception):
error_queue.put((index,) + sys.exc_info()[:2])
pipe.send((None, False))
finally:
env.close()
class DmpAsyncVectorEnv(gym.vector.AsyncVectorEnv):
def __init__(self, env_fns, n_samples, observation_space=None, action_space=None,
shared_memory=False, copy=True, context="spawn", daemon=True, worker=_worker):
super(DmpAsyncVectorEnv, self).__init__(env_fns,
observation_space=observation_space,
action_space=action_space,
shared_memory=shared_memory,
copy=copy,
context=context,
daemon=daemon,
worker=worker)
# we need to overwrite the number of samples as we may sample more than num_envs
self.observations = create_empty_array(self.single_observation_space,
n=n_samples,
fn=np.zeros)
def __call__(self, params, contexts=None):
return self.rollout(params, contexts)
def rollout_async(self, params, contexts):
"""
Parameters
----------
params : iterable of samples from `action_space`
List of actions.
"""
self._assert_is_running()
if self._state != AsyncState.DEFAULT:
raise AlreadyPendingCallError('Calling `rollout_async` while waiting '
'for a pending call to `{0}` to complete.'.format(
self._state.value), self._state.value)
params = np.atleast_2d(params)
split_params = np.array_split(params, np.minimum(len(params), self.num_envs))
if contexts is None:
split_contexts = np.array_split([None, ] * len(params), np.minimum(len(params), self.num_envs))
else:
split_contexts = np.array_split(contexts, np.minimum(len(contexts), self.num_envs))
assert np.all([len(p) == len(c) for p, c in zip(split_params, split_contexts)])
for pipe, param, context in zip(self.parent_pipes, split_params, split_contexts):
pipe.send(('rollout', (param, context)))
for pipe in self.parent_pipes[len(split_params):]:
pipe.send(('idle', None))
self._state = AsyncState.WAITING_ROLLOUT
def rollout_wait(self, timeout=None):
"""
Parameters
----------
timeout : int or float, optional
Number of seconds before the call to `step_wait` times out. If
`None`, the call to `step_wait` never times out.
Returns
-------
observations : sample from `observation_space`
A batch of observations from the vectorized environment.
rewards : `np.ndarray` instance (dtype `np.float_`)
A vector of rewards from the vectorized environment.
dones : `np.ndarray` instance (dtype `np.bool_`)
A vector whose entries indicate whether the episode has ended.
infos : list of dict
A list of auxiliary diagnostic information.
"""
self._assert_is_running()
if self._state != AsyncState.WAITING_ROLLOUT:
raise NoAsyncCallError('Calling `rollout_wait` without any prior call '
'to `rollout_async`.', AsyncState.WAITING_ROLLOUT.value)
if not self._poll(timeout):
self._state = AsyncState.DEFAULT
raise mp.TimeoutError('The call to `rollout_wait` has timed out after '
'{0} second{1}.'.format(timeout, 's' if timeout > 1 else ''))
results, successes = zip(*[pipe.recv() for pipe in self.parent_pipes])
results = [r for r in results if r is not None]
self._raise_if_errors(successes)
self._state = AsyncState.DEFAULT
rewards, infos = [_flatten_list(r) for r in zip(*results)]
# for now, we ignore the observations and only return the rewards
# if not self.shared_memory:
# self.observations = concatenate(observations_list, self.observations,
# self.single_observation_space)
# return (deepcopy(self.observations) if self.copy else self.observations,
# np.array(rewards), np.array(dones, dtype=np.bool_), infos)
return np.array(rewards), infos
def rollout(self, actions, contexts):
self.rollout_async(actions, contexts)
return self.rollout_wait()
def _flatten_obs(obs):
assert isinstance(obs, (list, tuple))
assert len(obs) > 0
if isinstance(obs[0], dict):
keys = obs[0].keys()
return {k: np.stack([o[k] for o in obs]) for k in keys}
else:
return np.stack(obs)
def _flatten_list(l):
assert isinstance(l, (list, tuple))
assert len(l) > 0
assert all([len(l_) > 0 for l_ in l])
return [l__ for l_ in l for l__ in l_]

125
alr_envs/utils/legacy/dmp_env_wrapper.py
View File

@ -1,125 +0,0 @@
from alr_envs.utils.policies import get_policy_class
from mp_lib.phase import ExpDecayPhaseGenerator
from mp_lib.basis import DMPBasisGenerator
from mp_lib import dmps
import numpy as np
import gym
class DmpEnvWrapper(gym.Wrapper):
def __init__(self,
env,
num_dof,
num_basis,
start_pos=None,
final_pos=None,
duration=1,
dt=0.01,
alpha_phase=2,
bandwidth_factor=3,
learn_goal=False,
post_traj_time=0.,
policy_type=None,
weights_scale=1.,
goal_scale=1.,
):
super(DmpEnvWrapper, self).__init__(env)
self.num_dof = num_dof
self.num_basis = num_basis
self.dim = num_dof * num_basis
if learn_goal:
self.dim += num_dof
self.learn_goal = learn_goal
self.duration = duration # seconds
time_steps = int(duration / dt)
self.t = np.linspace(0, duration, time_steps)
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_bandwidth_factor=bandwidth_factor)
self.dmp = dmps.DMP(num_dof=num_dof,
basis_generator=basis_generator,
phase_generator=phase_generator,
num_time_steps=time_steps,
dt=dt
)
self.dmp.dmp_start_pos = start_pos.reshape((1, num_dof))
dmp_weights = np.zeros((num_basis, num_dof))
if learn_goal:
dmp_goal_pos = np.zeros(num_dof)
else:
dmp_goal_pos = final_pos
self.dmp.set_weights(dmp_weights, dmp_goal_pos)
self.weights_scale = weights_scale
self.goal_scale = goal_scale
policy_class = get_policy_class(policy_type)
self.policy = policy_class(env)
def __call__(self, params, contexts=None):
params = np.atleast_2d(params)
rewards = []
infos = []
for p, c in zip(params, contexts):
reward, info = self.rollout(p, c)
rewards.append(reward)
infos.append(info)
return np.array(rewards), infos
def goal_and_weights(self, params):
if len(params.shape) > 1:
assert params.shape[1] == self.dim
else:
assert len(params) == self.dim
params = np.reshape(params, [1, self.dim])
if self.learn_goal:
goal_pos = params[0, -self.num_dof:]
weight_matrix = np.reshape(params[:, :-self.num_dof], [self.num_basis, self.num_dof])
else:
goal_pos = self.dmp.dmp_goal_pos.flatten()
assert goal_pos is not None
weight_matrix = np.reshape(params, [self.num_basis, self.num_dof])
return goal_pos * self.goal_scale, weight_matrix * self.weights_scale
def rollout(self, params, context=None, render=False):
""" This function generates a trajectory based on a DMP and then does the usual loop over reset and step"""
goal_pos, weight_matrix = self.goal_and_weights(params)
self.dmp.set_weights(weight_matrix, goal_pos)
trajectory, velocity = self.dmp.reference_trajectory(self.t)
if self.post_traj_steps > 0:
trajectory = np.vstack([trajectory, np.tile(trajectory[-1, :], [self.post_traj_steps, 1])])
velocity = np.vstack([velocity, np.zeros(shape=(self.post_traj_steps, self.num_dof))])
self._trajectory = trajectory
self._velocity = velocity
rews = []
infos = []
self.env.configure(context)
self.env.reset()
for t, pos_vel in enumerate(zip(trajectory, velocity)):
ac = self.policy.get_action(pos_vel[0], pos_vel[1])
obs, rew, done, info = self.env.step(ac)
rews.append(rew)
infos.append(info)
if render:
self.env.render(mode="human")
if done:
break
reward = np.sum(rews)
return reward, info

28
alr_envs/utils/legacy/dmp_env_wrapper_example.py
View File

@ -1,28 +0,0 @@
from alr_envs.utils.legacy.utils import make_holereacher_env
import numpy as np
if __name__ == "__main__":
n_samples = 1
n_cpus = 4
dim = 30
# env = DmpAsyncVectorEnv([make_viapointreacher_env(i) for i in range(n_cpus)],
# n_samples=n_samples)
test_env = make_holereacher_env(0)()
# params = np.random.randn(n_samples, dim)
params = np.array([[1.386102, -3.29980525, 4.70402733, 1.3966668, 0.73774902,
3.14676681, -4.98644416, 6.20303193, 1.30502127, -0.09330522,
7.62656797, -5.76893033, 3.4706711, -0.6944142, -3.33442788,
12.31421548, -0.72760271, -6.9090723, 7.02903814, -8.7236836,
1.4805914, 0.53185824, -5.46626893, 0.69692163, 13.58472666,
0.77199316, 2.02906724, -3.0203244, -1.00533159, -0.57417351]])
# params = np.hstack([50 * np.random.randn(n_samples, 25), np.tile(np.array([np.pi/2, -np.pi/4, -np.pi/4, -np.pi/4, -np.pi/4]), [n_samples, 1])])
rew, info = test_env.rollout(params, render=True)
print(rew)
# out = env(params)
# print(out)

28
alr_envs/utils/legacy/dmp_pd_control_example.py
View File

@ -1,28 +0,0 @@
from alr_envs.mujoco.ball_in_a_cup.utils import make_simple_dmp_env
import numpy as np
if __name__ == "__main__":
dim = 15
n_cpus = 4
# n_samples = 10
#
# vec_env = DmpAsyncVectorEnv([make_simple_env(i) for i in range(n_cpus)],
# n_samples=n_samples)
#
# params = np.tile(1 * np.random.randn(n_samples, dim), (10, 1))
#
# rewards, infos = vec_env(params)
# print(rewards)
#
non_vec_env = make_simple_dmp_env(0, 0)()
# params = 0.5 * np.random.randn(dim)
params = np.array([-2.63357598, -1.04950296, -0.44330737, 0.52950017, 4.29247739,
4.52473661, -0.05685977, -0.76796851, 3.71540499, 1.22631059,
2.20412438, 3.91588129, -0.12652723, -3.0788211 , 0.56204464])
out2 = non_vec_env.rollout(params, render=True )
print(out2)

156
alr_envs/utils/legacy/utils.py
View File

@ -1,156 +0,0 @@
import alr_envs.classic_control.hole_reacher as hr
import alr_envs.classic_control.viapoint_reacher as vpr
from alr_envs.utils.wrapper.dmp_wrapper import DmpWrapper
from alr_envs.utils.wrapper.detpmp_wrapper import DetPMPWrapper
import numpy as np
def make_viapointreacher_env(rank, seed=0):
"""
Utility function for multiprocessed env.
:param env_id: (str) the environment ID
:param num_env: (int) the number of environments you wish to have in subprocesses
:param seed: (int) the initial seed for RNG
:param rank: (int) index of the subprocess
:returns a function that generates an environment
"""
def _init():
_env = vpr.ViaPointReacher(n_links=5,
allow_self_collision=False,
collision_penalty=1000)
_env = DmpWrapper(_env,
num_dof=5,
num_basis=5,
duration=2,
alpha_phase=2.5,
dt=_env.dt,
start_pos=_env.start_pos,
learn_goal=False,
policy_type="velocity",
weights_scale=50)
_env.seed(seed + rank)
return _env
return _init
def make_holereacher_env(rank, seed=0):
"""
Utility function for multiprocessed env.
:param env_id: (str) the environment ID
:param num_env: (int) the number of environments you wish to have in subprocesses
:param seed: (int) the initial seed for RNG
:param rank: (int) index of the subprocess
:returns a function that generates an environment
"""
def _init():
_env = hr.HoleReacher(n_links=5,
allow_self_collision=False,
allow_wall_collision=False,
hole_width=0.25,
hole_depth=1,
hole_x=2,
collision_penalty=100)
_env = DmpWrapper(_env,
num_dof=5,
num_basis=5,
duration=2,
bandwidth_factor=2,
dt=_env.dt,
learn_goal=True,
alpha_phase=2,
start_pos=_env.start_pos,
policy_type="velocity",
weights_scale=50,
goal_scale=0.1
)
_env.seed(seed + rank)
return _env
return _init
def make_holereacher_fix_goal_env(rank, seed=0):
"""
Utility function for multiprocessed env.
:param env_id: (str) the environment ID
:param num_env: (int) the number of environments you wish to have in subprocesses
:param seed: (int) the initial seed for RNG
:param rank: (int) index of the subprocess
:returns a function that generates an environment
"""
def _init():
_env = hr.HoleReacher(n_links=5,
allow_self_collision=False,
allow_wall_collision=False,
hole_width=0.15,
hole_depth=1,
hole_x=1,
collision_penalty=100)
_env = DmpWrapper(_env,
num_dof=5,
num_basis=5,
duration=2,
dt=_env.dt,
learn_goal=False,
final_pos=np.array([2.02669572, -1.25966385, -1.51618198, -0.80946476, 0.02012344]),
alpha_phase=2,
start_pos=_env.start_pos,
policy_type="velocity",
weights_scale=50,
goal_scale=1
)
_env.seed(seed + rank)
return _env
return _init
def make_holereacher_env_pmp(rank, seed=0):
"""
Utility function for multiprocessed env.
:param env_id: (str) the environment ID
:param num_env: (int) the number of environments you wish to have in subprocesses
:param seed: (int) the initial seed for RNG
:param rank: (int) index of the subprocess
:returns a function that generates an environment
"""
def _init():
_env = hr.HoleReacher(n_links=5,
allow_self_collision=False,
allow_wall_collision=False,
hole_width=0.15,
hole_depth=1,
hole_x=1,
collision_penalty=1000)
_env = DetPMPWrapper(_env,
num_dof=5,
num_basis=5,
width=0.02,
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_goal=False
)
_env.seed(seed + rank)
return _env
return _init

4
alr_envs/utils/make_env_helpers.py
View File

@ -1,5 +1,5 @@
from alr_envs.utils.wrapper.dmp_wrapper import DmpWrapper
from alr_envs.utils.wrapper.detpmp_wrapper import DetPMPWrapper
from alr_envs.utils.mps.dmp_wrapper import DmpWrapper
from alr_envs.utils.mps.detpmp_wrapper import DetPMPWrapper
import gym
from gym.vector.utils import write_to_shared_memory
import sys

40
alr_envs/utils/mp_env_async_sampler.py
View File

@ -4,8 +4,8 @@ import numpy as np
from _collections import defaultdict
def make_env(env_id, rank, seed=0):
env = gym.make(env_id)
def make_env(env_id, rank, seed=0, **env_kwargs):
env = gym.make(env_id, **env_kwargs)
env.seed(seed + rank)
return lambda: env
@ -45,9 +45,9 @@ class AlrMpEnvSampler:
An asynchronous sampler for non contextual MPWrapper environments. A sampler object can be called with a set of
parameters and returns the corresponding final obs, rewards, dones and info dicts.
"""
def __init__(self, env_id, num_envs, seed=0):
def __init__(self, env_id, num_envs, seed=0, **env_kwargs):
self.num_envs = num_envs
self.env = AsyncVectorEnv([make_env(env_id, seed, i) for i in range(num_envs)])
self.env = AsyncVectorEnv([make_env(env_id, seed, i, **env_kwargs) for i in range(num_envs)])
def __call__(self, params):
params = np.atleast_2d(params)
@ -56,6 +56,7 @@ class AlrMpEnvSampler:
vals = defaultdict(list)
for p in split_params:
self.env.reset()
obs, reward, done, info = self.env.step(p)
vals['obs'].append(obs)
vals['reward'].append(reward)
@ -67,6 +68,37 @@ class AlrMpEnvSampler:
_flatten_list(vals['done'])[:n_samples], _flatten_list(vals['info'])[:n_samples]
class AlrContextualMpEnvSampler:
"""
An asynchronous sampler for contextual MPWrapper environments. A sampler object can be called with a set of
parameters and returns the corresponding final obs, rewards, dones and info dicts.
"""
def __init__(self, env_id, num_envs, seed=0, **env_kwargs):
self.num_envs = num_envs
self.env = AsyncVectorEnv([make_env(env_id, seed, i, **env_kwargs) for i in range(num_envs)])
def __call__(self, dist, n_samples):
repeat = int(np.ceil(n_samples / self.env.num_envs))
vals = defaultdict(list)
for i in range(repeat):
new_contexts = self.env.reset()
vals['new_contexts'].append(new_contexts)
new_samples, new_contexts = dist.sample(new_contexts)
vals['new_samples'].append(new_samples)
obs, reward, done, info = self.env.step(new_samples)
vals['obs'].append(obs)
vals['reward'].append(reward)
vals['done'].append(done)
vals['info'].append(info)
# do not return values above threshold
return np.vstack(vals['new_samples'])[:n_samples], np.vstack(vals['new_contexts'])[:n_samples], \
np.vstack(vals['obs'])[:n_samples], np.hstack(vals['reward'])[:n_samples], \
_flatten_list(vals['done'])[:n_samples], _flatten_list(vals['info'])[:n_samples]
if __name__ == "__main__":
env_name = "alr_envs:ALRBallInACupSimpleDMP-v0"
n_cpu = 8

0
alr_envs/utils/legacy/__init__.py → alr_envs/utils/mps/__init__.py
View File

21
alr_envs/utils/wrapper/detpmp_wrapper.py → alr_envs/utils/mps/detpmp_wrapper.py
View File

@ -2,26 +2,27 @@ import gym
import numpy as np
from mp_lib import det_promp
from alr_envs.utils.wrapper.mp_wrapper import MPWrapper
from alr_envs.utils.mps.mp_environments import MPEnv
from alr_envs.utils.mps.mp_wrapper import MPWrapper
class DetPMPWrapper(MPWrapper):
def __init__(self, env, num_dof, num_basis, width, start_pos=None, duration=1, dt=0.01, post_traj_time=0.,
policy_type=None, weights_scale=1, zero_start=False, zero_goal=False, **mp_kwargs):
# self.duration = duration # seconds
def __init__(self, env: MPEnv, num_dof: int, num_basis: int, width: int, duration: int = 1, dt: float = 0.01,
post_traj_time: float = 0., policy_type: str = None, weights_scale: float = 1.,
zero_start: bool = False, zero_goal: bool = False, **mp_kwargs):
self.duration = duration # seconds
super().__init__(env, num_dof, duration, dt, post_traj_time, policy_type, weights_scale,
num_basis=num_basis, width=width, start_pos=start_pos, zero_start=zero_start,
zero_goal=zero_goal)
super().__init__(env, num_dof, dt, duration, post_traj_time, policy_type, weights_scale, num_basis=num_basis,
width=width, zero_start=zero_start, zero_goal=zero_goal, **mp_kwargs)
self.dt = dt
action_bounds = np.inf * np.ones((self.mp.n_basis * self.mp.n_dof))
self.action_space = gym.spaces.Box(low=-action_bounds, high=action_bounds, dtype=np.float32)
self.start_pos = start_pos
self.dt = dt
def initialize_mp(self, num_dof: int, duration: int, dt: float, num_basis: int = 5, width: float = None,
start_pos: np.ndarray = None, zero_start: bool = False, zero_goal: bool = False):
zero_start: bool = False, zero_goal: bool = False):
pmp = det_promp.DeterministicProMP(n_basis=num_basis, n_dof=num_dof, width=width, off=0.01,
zero_start=zero_start, zero_goal=zero_goal)

48
alr_envs/utils/wrapper/dmp_wrapper.py → alr_envs/utils/mps/dmp_wrapper.py
View File

@ -1,17 +1,18 @@
from mp_lib.phase import ExpDecayPhaseGenerator
from mp_lib.basis import DMPBasisGenerator
from mp_lib import dmps
import numpy as np
import gym
import numpy as np
from mp_lib import dmps
from mp_lib.basis import DMPBasisGenerator
from mp_lib.phase import ExpDecayPhaseGenerator
from alr_envs.utils.wrapper.mp_wrapper import MPWrapper
from alr_envs.utils.mps.mp_environments import MPEnv
from alr_envs.utils.mps.mp_wrapper import MPWrapper
class DmpWrapper(MPWrapper):
def __init__(self, env: gym.Env, num_dof: int, num_basis: int, start_pos: np.ndarray = None,
final_pos: np.ndarray = None, duration: int = 1, alpha_phase: float = 2., dt: float = None,
learn_goal: bool = False, return_to_start: bool = False, post_traj_time: float = 0.,
def __init__(self, env: MPEnv, num_dof: int, num_basis: int,
duration: int = 1, alpha_phase: float = 2., dt: float = None,
learn_goal: bool = False, post_traj_time: float = 0.,
weights_scale: float = 1., goal_scale: float = 1., bandwidth_factor: float = 3.,
policy_type: str = None, render_mode: str = None):
@ -21,8 +22,6 @@ class DmpWrapper(MPWrapper):
env:
num_dof:
num_basis:
start_pos:
final_pos:
duration:
alpha_phase:
dt:
@ -35,25 +34,17 @@ class DmpWrapper(MPWrapper):
self.learn_goal = learn_goal
dt = env.dt if hasattr(env, "dt") else dt
assert dt is not None
start_pos = start_pos if start_pos is not None else env.start_pos if hasattr(env, "start_pos") else None
assert start_pos is not None
if learn_goal:
final_pos = np.zeros_like(start_pos) # arbitrary, will be learned
else:
final_pos = final_pos if final_pos is not None else start_pos if return_to_start else None
assert final_pos is not None
self.t = np.linspace(0, duration, int(duration / dt))
self.goal_scale = goal_scale
super().__init__(env, num_dof, duration, dt, post_traj_time, policy_type, weights_scale, render_mode,
num_basis=num_basis, start_pos=start_pos, final_pos=final_pos, alpha_phase=alpha_phase,
bandwidth_factor=bandwidth_factor)
super().__init__(env, num_dof, dt, duration, post_traj_time, policy_type, weights_scale, render_mode,
num_basis=num_basis, alpha_phase=alpha_phase, bandwidth_factor=bandwidth_factor)
action_bounds = np.inf * np.ones((np.prod(self.mp.dmp_weights.shape) + (num_dof if learn_goal else 0)))
self.action_space = gym.spaces.Box(low=-action_bounds, high=action_bounds, dtype=np.float32)
def initialize_mp(self, num_dof: int, duration: int, dt: float, num_basis: int = 5, start_pos: np.ndarray = None,
final_pos: np.ndarray = None, alpha_phase: float = 2., bandwidth_factor: float = 3.):
def initialize_mp(self, num_dof: int, duration: int, dt: float, num_basis: int = 5, alpha_phase: float = 2.,
bandwidth_factor: int = 3):
phase_generator = ExpDecayPhaseGenerator(alpha_phase=alpha_phase, duration=duration)
basis_generator = DMPBasisGenerator(phase_generator, duration=duration, num_basis=num_basis,
@ -62,12 +53,6 @@ class DmpWrapper(MPWrapper):
dmp = dmps.DMP(num_dof=num_dof, basis_generator=basis_generator, phase_generator=phase_generator,
num_time_steps=int(duration / dt), dt=dt)
dmp.dmp_start_pos = start_pos.reshape((1, num_dof))
weights = np.zeros((num_basis, num_dof))
goal_pos = np.zeros(num_dof) if self.learn_goal else final_pos
dmp.set_weights(weights, goal_pos)
return dmp
def goal_and_weights(self, params):
@ -77,16 +62,15 @@ class DmpWrapper(MPWrapper):
if self.learn_goal:
goal_pos = params[0, -self.mp.num_dimensions:] # [num_dof]
params = params[:, :-self.mp.num_dimensions] # [1,num_dof]
# weight_matrix = np.reshape(params[:, :-self.num_dof], [self.num_basis, self.num_dof])
else:
goal_pos = self.mp.dmp_goal_pos.flatten()
goal_pos = self.env.goal_pos
assert goal_pos is not None
# weight_matrix = np.reshape(params, [self.num_basis, self.num_dof])
weight_matrix = np.reshape(params, self.mp.dmp_weights.shape)
weight_matrix = np.reshape(params, self.mp.dmp_weights.shape) # [num_basis, num_dof]
return goal_pos * self.goal_scale, weight_matrix * self.weights_scale
def mp_rollout(self, action):
self.mp.dmp_start_pos = self.env.start_pos
goal_pos, weight_matrix = self.goal_and_weights(action)
self.mp.set_weights(weight_matrix, goal_pos)
return self.mp.reference_trajectory(self.t)

33
alr_envs/utils/mps/mp_environments.py Normal file
View File

@ -0,0 +1,33 @@
from abc import abstractmethod
from typing import Union
import gym
import numpy as np
class MPEnv(gym.Env):
@property
@abstractmethod
def active_obs(self):
"""Returns boolean mask for each observation entry
whether the observation is returned for the contextual case or not.
This effectively allows to filter unwanted or unnecessary observations from the full step-based case.
"""
return np.ones(self.observation_space.shape, dtype=bool)
@property
@abstractmethod
def start_pos(self) -> Union[float, int, np.ndarray]:
"""
Returns the starting position of the joints
"""
raise NotImplementedError()
@property
def goal_pos(self) -> Union[float, int, np.ndarray]:
"""
Returns the current final position of the joints for the MP.
By default this returns the starting position.
"""
return self.start_pos

55
alr_envs/utils/wrapper/mp_wrapper.py → alr_envs/utils/mps/mp_wrapper.py
View File

@ -1,32 +1,24 @@
from abc import ABC, abstractmethod
from collections import defaultdict
import gym
import numpy as np
from alr_envs.utils.mps.mp_environments import MPEnv
from alr_envs.utils.policies import get_policy_class
class MPWrapper(gym.Wrapper, ABC):
def __init__(self,
env: gym.Env,
num_dof: int,
duration: int = 1,
dt: float = None,
post_traj_time: float = 0.,
policy_type: str = None,
weights_scale: float = 1.,
render_mode: str = None,
**mp_kwargs
):
def __init__(self, env: MPEnv, num_dof: int, dt: float, duration: int = 1, post_traj_time: float = 0.,
policy_type: str = None, weights_scale: float = 1., render_mode: str = None, **mp_kwargs):
super().__init__(env)
# self.num_dof = num_dof
# self.num_basis = num_basis
# self.duration = duration # seconds
# adjust observation space to reduce version
obs_sp = self.env.observation_space
self.observation_space = gym.spaces.Box(low=obs_sp.low[self.env.active_obs],
high=obs_sp.high[self.env.active_obs],
dtype=obs_sp.dtype)
# dt = env.dt if hasattr(env, "dt") else dt
assert dt is not None # this should never happen as MPWrapper is a base class
self.post_traj_steps = int(post_traj_time / dt)
@ -40,8 +32,11 @@ class MPWrapper(gym.Wrapper, ABC):
self.render_mode = render_mode
self.render_kwargs = {}
# TODO: not yet final
# TODO: @Max I think this should not be in this class, this functionality should be part of your sampler.
def __call__(self, params, contexts=None):
"""
Can be used to provide a batch of parameter sets
"""
params = np.atleast_2d(params)
obs = []
rewards = []
@ -61,6 +56,9 @@ class MPWrapper(gym.Wrapper, ABC):
def configure(self, context):
self.env.configure(context)
def reset(self):
return self.env.reset()[self.env.active_obs]
def step(self, action: np.ndarray):
""" This function generates a trajectory based on a DMP and then does the usual loop over reset and step"""
trajectory, velocity = self.mp_rollout(action)
@ -73,14 +71,9 @@ class MPWrapper(gym.Wrapper, ABC):
# self._velocity = velocity
rewards = 0
# infos = defaultdict(list)
# TODO: @Max Why do we need this configure, states should be part of the model
# TODO: Ask Onur if the context distribution needs to be outside the environment
# TODO: For now create a new env with each context
# self.env.configure(context)
obs = self.env.reset()
info = {}
# create random obs as the reset function is called externally
obs = self.env.observation_space.sample()
for t, pos_vel in enumerate(zip(trajectory, velocity)):
ac = self.policy.get_action(pos_vel[0], pos_vel[1])
@ -94,7 +87,7 @@ class MPWrapper(gym.Wrapper, ABC):
break
done = True
return obs, rewards, done, info
return obs[self.env.active_obs], rewards, done, info
def render(self, mode='human', **kwargs):
"""Only set render options here, such that they can be used during the rollout.
@ -102,18 +95,6 @@ class MPWrapper(gym.Wrapper, ABC):
self.render_mode = mode
self.render_kwargs = kwargs
# def __call__(self, actions):
# return self.step(actions)
# params = np.atleast_2d(params)
# rewards = []
# infos = []
# for p, c in zip(params, contexts):
# reward, info = self.rollout(p, c)
# rewards.append(reward)
# infos.append(info)
#
# return np.array(rewards), infos
@abstractmethod
def mp_rollout(self, action):
"""

0
alr_envs/utils/wrapper/__init__.py
View File

10
example.py
View File

@ -46,7 +46,7 @@ def example_dmp():
obs = env.reset()
def example_async(n_cpu=4, seed=int('533D', 16)):
def example_async(env_id="alr_envs:HoleReacherDMP-v0", n_cpu=4, seed=int('533D', 16)):
def make_env(env_id, seed, rank):
env = gym.make(env_id)
env.seed(seed + rank)
@ -73,7 +73,7 @@ def example_async(n_cpu=4, seed=int('533D', 16)):
# do not return values above threshold
return (*map(lambda v: np.stack(v)[:n_samples], vals.values()),)
envs = gym.vector.AsyncVectorEnv([make_env("alr_envs:HoleReacherDMP-v0", seed, i) for i in range(n_cpu)])
envs = gym.vector.AsyncVectorEnv([make_env(env_id, seed, i) for i in range(n_cpu)])
obs = envs.reset()
print(sample(envs, 16))
@ -82,6 +82,6 @@ def example_async(n_cpu=4, seed=int('533D', 16)):
if __name__ == '__main__':
# example_mujoco()
# example_dmp()
# example_async()
env = gym.make("alr_envs:HoleReacherDMP-v0", context=0.1)
print()
example_async("alr_envs:LongSimpleReacherDMP-v0", 4)
# env = gym.make("alr_envs:HoleReacherDMP-v0", context=0.1)
# env = gym.make("alr_envs:HoleReacherDMP-v1")