-
Notifications
You must be signed in to change notification settings - Fork 0
/
robodesk.py
116 lines (98 loc) · 4 KB
/
robodesk.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
import os
import gym
import embodied
import numpy as np
class RoboDesk(embodied.Env):
def __init__(self, task, mode, repeat=1, length=500, resets=True):
assert mode in ('train', 'eval')
# TODO: This env variable is meant for headless GPU machines but may fail
# on CPU-only machines.
if 'MUJOCO_GL' not in os.environ:
os.environ['MUJOCO_GL'] = 'egl'
try:
from robodesk import robodesk
except ImportError:
import robodesk
task, reward = task.rsplit('_', 1)
if mode == 'eval':
reward = 'success'
assert reward in ('dense', 'sparse', 'success'), reward
self._gymenv = robodesk.RoboDesk(task, reward, repeat, length)
from . import from_gym
self._env = from_gym.FromGym(self._gymenv)
@property
def obs_space(self):
return self._env.obs_space
@property
def act_space(self):
return self._env.act_space
def step(self, action):
obs = self._env.step(action)
obs['is_terminal'] = False
return obs
class RoboDeskMulti(embodied.Env):
def __init__(self, _, task_sequence = ["flat_block_in_bin", "upright_block_off_table", "push_green"], repeat=8, length=500, resets=True, image_size=64):
# note: we ignore task variable that's passed in
# if 'MUJOCO_GL' not in os.environ:
# os.environ['MUJOCO_GL'] = 'egl'
# import robodesk
from absl import logging
logging.set_verbosity(logging.ERROR)
# self._gymenv = robodesk.RoboDesk(task = task_sequence[0], reward = 'dense', action_repeat = repeat, episode_length = length * repeat, image_size = image_size)
from . import robodesk_hd
self._gymenv = robodesk_hd.RoboDeskHD(task = task_sequence[0], reward = 'dense', action_repeat = repeat, episode_length = length * repeat, image_size = image_size)
self._gymenv = RobodeskWrapper(self._gymenv, task_sequence = task_sequence)
from . import from_gym
self._env = from_gym.FromGym(self._gymenv)
@property
def obs_space(self):
return self._env.obs_space
@property
def act_space(self):
return self._env.act_space
def step(self, action):
obs = self._env.step(action)
obs['is_terminal'] = False
return obs
class RobodeskWrapper(gym.Wrapper):
def __init__(self, env, task_sequence = ["flat_block_in_bin", "upright_block_off_table", "push_green"]):
# if output size is none, don't resize the image from wrapped env
super().__init__(env)
self._task_sequence = task_sequence
self._task_num = 0
self.env.task = self._task_sequence[self._task_num]
self._need_new_goal = False
self._all_goals_done = False
@property
def observation_space(self): # TODO get observation spaces to handle downsampling of images
obs_space = self.env.observation_space
obs_space.spaces["i_tasks_completed"] = gym.spaces.Box(low=0, high=len(self._task_sequence), dtype=np.float32, shape=(1,))
return obs_space
def reset(self, seed=None):
self._task_num = 0
self.env.task = self._task_sequence[self._task_num]
self._need_new_goal = False
self._all_goals_done = False
ob = self.env.reset()
return self._add_info(ob), {}
def step(self, ac): # TODO downsample images if necessary
if self._need_new_goal:
self._next_task()
ob, rew, done, info = self.env.step(ac)
# check if next task is needed
self._need_new_goal = self.env._get_task_reward(self.env.task, 'success') > 0.5
# give bonus reward
if self._need_new_goal and not self._all_goals_done:
rew += 300
self._all_goals_done = self._all_goals_done or (self._task_num == len(self._task_sequence)-1)
return self._add_info(ob), rew, done, done, info
def _next_task(self):
# move to next task
self._task_num = min(len(self._task_sequence) - 1, self._task_num + 1)
self.env.task = self._task_sequence[self._task_num]
self._need_new_goal = False
def _add_info(self, ob):
# adds logging info to observations
completed_tasks = (self._task_num + 1) if self._all_goals_done else self._task_num
ob.update({"i_tasks_completed": completed_tasks})
return ob