-
Notifications
You must be signed in to change notification settings - Fork 438
/
factory_task_nut_bolt_pick.py
451 lines (350 loc) · 22.5 KB
/
factory_task_nut_bolt_pick.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
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
# Copyright (c) 2021-2023, NVIDIA Corporation
# All rights reserved.
#
# Redistribution and use in source and binary forms, with or without
# modification, are permitted provided that the following conditions are met:
#
# 1. Redistributions of source code must retain the above copyright notice, this
# list of conditions and the following disclaimer.
#
# 2. Redistributions in binary form must reproduce the above copyright notice,
# this list of conditions and the following disclaimer in the documentation
# and/or other materials provided with the distribution.
#
# 3. Neither the name of the copyright holder nor the names of its
# contributors may be used to endorse or promote products derived from
# this software without specific prior written permission.
#
# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
# AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
# DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
# FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
# DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
# SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
# CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
# OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
"""Factory: Class for nut-bolt pick task.
Inherits nut-bolt environment class and abstract task class (not enforced). Can be executed with
python train.py task=FactoryTaskNutBoltPick
"""
import hydra
import omegaconf
import os
import torch
from isaacgym import gymapi, gymtorch
from isaacgymenvs.utils import torch_jit_utils as torch_utils
import isaacgymenvs.tasks.factory.factory_control as fc
from isaacgymenvs.tasks.factory.factory_env_nut_bolt import FactoryEnvNutBolt
from isaacgymenvs.tasks.factory.factory_schema_class_task import FactoryABCTask
from isaacgymenvs.tasks.factory.factory_schema_config_task import FactorySchemaConfigTask
from isaacgymenvs.utils import torch_jit_utils
class FactoryTaskNutBoltPick(FactoryEnvNutBolt, FactoryABCTask):
def __init__(self, cfg, rl_device, sim_device, graphics_device_id, headless, virtual_screen_capture, force_render):
"""Initialize instance variables. Initialize environment superclass."""
super().__init__(cfg, rl_device, sim_device, graphics_device_id, headless, virtual_screen_capture, force_render)
self.cfg = cfg
self._get_task_yaml_params()
self._acquire_task_tensors()
self.parse_controller_spec()
if self.cfg_task.sim.disable_gravity:
self.disable_gravity()
if self.viewer is not None:
self._set_viewer_params()
def _get_task_yaml_params(self):
"""Initialize instance variables from YAML files."""
cs = hydra.core.config_store.ConfigStore.instance()
cs.store(name='factory_schema_config_task', node=FactorySchemaConfigTask)
self.cfg_task = omegaconf.OmegaConf.create(self.cfg)
self.max_episode_length = self.cfg_task.rl.max_episode_length # required instance var for VecTask
asset_info_path = '../../assets/factory/yaml/factory_asset_info_nut_bolt.yaml' # relative to Gym's Hydra search path (cfg dir)
self.asset_info_nut_bolt = hydra.compose(config_name=asset_info_path)
self.asset_info_nut_bolt = self.asset_info_nut_bolt['']['']['']['']['']['']['assets']['factory']['yaml'] # strip superfluous nesting
ppo_path = 'train/FactoryTaskNutBoltPickPPO.yaml' # relative to Gym's Hydra search path (cfg dir)
self.cfg_ppo = hydra.compose(config_name=ppo_path)
self.cfg_ppo = self.cfg_ppo['train'] # strip superfluous nesting
def _acquire_task_tensors(self):
"""Acquire tensors."""
# Grasp pose tensors
nut_grasp_heights = self.bolt_head_heights + self.nut_heights * 0.5 # nut COM
self.nut_grasp_pos_local = nut_grasp_heights * torch.tensor([0.0, 0.0, 1.0], device=self.device).repeat(
(self.num_envs, 1))
self.nut_grasp_quat_local = torch.tensor([0.0, 1.0, 0.0, 0.0], device=self.device).unsqueeze(0).repeat(
self.num_envs, 1)
# Keypoint tensors
self.keypoint_offsets = self._get_keypoint_offsets(
self.cfg_task.rl.num_keypoints) * self.cfg_task.rl.keypoint_scale
self.keypoints_gripper = torch.zeros((self.num_envs, self.cfg_task.rl.num_keypoints, 3),
dtype=torch.float32,
device=self.device)
self.keypoints_nut = torch.zeros_like(self.keypoints_gripper, device=self.device)
self.identity_quat = torch.tensor([0.0, 0.0, 0.0, 1.0], device=self.device).unsqueeze(0).repeat(self.num_envs,
1)
def _refresh_task_tensors(self):
"""Refresh tensors."""
# Compute pose of nut grasping frame
self.nut_grasp_quat, self.nut_grasp_pos = torch_jit_utils.tf_combine(self.nut_quat,
self.nut_pos,
self.nut_grasp_quat_local,
self.nut_grasp_pos_local)
# Compute pos of keypoints on gripper and nut in world frame
for idx, keypoint_offset in enumerate(self.keypoint_offsets):
self.keypoints_gripper[:, idx] = torch_jit_utils.tf_combine(self.fingertip_midpoint_quat,
self.fingertip_midpoint_pos,
self.identity_quat,
keypoint_offset.repeat(self.num_envs, 1))[1]
self.keypoints_nut[:, idx] = torch_jit_utils.tf_combine(self.nut_grasp_quat,
self.nut_grasp_pos,
self.identity_quat,
keypoint_offset.repeat(self.num_envs, 1))[1]
def pre_physics_step(self, actions):
"""Reset environments. Apply actions from policy. Simulation step called after this method."""
env_ids = self.reset_buf.nonzero(as_tuple=False).squeeze(-1)
if len(env_ids) > 0:
self.reset_idx(env_ids)
self.actions = actions.clone().to(self.device) # shape = (num_envs, num_actions); values = [-1, 1]
self._apply_actions_as_ctrl_targets(actions=self.actions,
ctrl_target_gripper_dof_pos=self.asset_info_franka_table.franka_gripper_width_max,
do_scale=True)
def post_physics_step(self):
"""Step buffers. Refresh tensors. Compute observations and reward. Reset environments."""
self.progress_buf[:] += 1
# In this policy, episode length is constant
is_last_step = (self.progress_buf[0] == self.max_episode_length - 1)
if self.cfg_task.env.close_and_lift:
# At this point, robot has executed RL policy. Now close gripper and lift (open-loop)
if is_last_step:
self._close_gripper(sim_steps=self.cfg_task.env.num_gripper_close_sim_steps)
self._lift_gripper(sim_steps=self.cfg_task.env.num_gripper_lift_sim_steps)
self.refresh_base_tensors()
self.refresh_env_tensors()
self._refresh_task_tensors()
self.compute_observations()
self.compute_reward()
def compute_observations(self):
"""Compute observations."""
# Shallow copies of tensors
obs_tensors = [self.fingertip_midpoint_pos,
self.fingertip_midpoint_quat,
self.fingertip_midpoint_linvel,
self.fingertip_midpoint_angvel,
self.nut_grasp_pos,
self.nut_grasp_quat]
self.obs_buf = torch.cat(obs_tensors, dim=-1) # shape = (num_envs, num_observations)
return self.obs_buf
def compute_reward(self):
"""Update reward and reset buffers."""
self._update_reset_buf()
self._update_rew_buf()
def _update_reset_buf(self):
"""Assign environments for reset if successful or failed."""
# If max episode length has been reached
self.reset_buf[:] = torch.where(self.progress_buf[:] >= self.max_episode_length - 1,
torch.ones_like(self.reset_buf),
self.reset_buf)
def _update_rew_buf(self):
"""Compute reward at current timestep."""
keypoint_reward = -self._get_keypoint_dist()
action_penalty = torch.norm(self.actions, p=2, dim=-1) * self.cfg_task.rl.action_penalty_scale
self.rew_buf[:] = keypoint_reward * self.cfg_task.rl.keypoint_reward_scale \
- action_penalty * self.cfg_task.rl.action_penalty_scale
# In this policy, episode length is constant across all envs
is_last_step = (self.progress_buf[0] == self.max_episode_length - 1)
if is_last_step:
# Check if nut is picked up and above table
lift_success = self._check_lift_success(height_multiple=3.0)
self.rew_buf[:] += lift_success * self.cfg_task.rl.success_bonus
self.extras['successes'] = torch.mean(lift_success.float())
def reset_idx(self, env_ids):
"""Reset specified environments."""
self._reset_franka(env_ids)
self._reset_object(env_ids)
self._randomize_gripper_pose(env_ids, sim_steps=self.cfg_task.env.num_gripper_move_sim_steps)
self._reset_buffers(env_ids)
def _reset_franka(self, env_ids):
"""Reset DOF states and DOF targets of Franka."""
self.dof_pos[env_ids] = torch.cat(
(torch.tensor(self.cfg_task.randomize.franka_arm_initial_dof_pos, device=self.device),
torch.tensor([self.asset_info_franka_table.franka_gripper_width_max], device=self.device),
torch.tensor([self.asset_info_franka_table.franka_gripper_width_max], device=self.device)),
dim=-1).unsqueeze(0).repeat((self.num_envs, 1)) # shape = (num_envs, num_dofs)
self.dof_vel[env_ids] = 0.0 # shape = (num_envs, num_dofs)
self.ctrl_target_dof_pos[env_ids] = self.dof_pos[env_ids]
multi_env_ids_int32 = self.franka_actor_ids_sim[env_ids].flatten()
self.gym.set_dof_state_tensor_indexed(self.sim,
gymtorch.unwrap_tensor(self.dof_state),
gymtorch.unwrap_tensor(multi_env_ids_int32),
len(multi_env_ids_int32))
def _reset_object(self, env_ids):
"""Reset root states of nut and bolt."""
# shape of root_pos = (num_envs, num_actors, 3)
# shape of root_quat = (num_envs, num_actors, 4)
# shape of root_linvel = (num_envs, num_actors, 3)
# shape of root_angvel = (num_envs, num_actors, 3)
# Randomize root state of nut
nut_noise_xy = 2 * (torch.rand((self.num_envs, 2), dtype=torch.float32, device=self.device) - 0.5) # [-1, 1]
nut_noise_xy = nut_noise_xy @ torch.diag(
torch.tensor(self.cfg_task.randomize.nut_pos_xy_initial_noise, device=self.device))
self.root_pos[env_ids, self.nut_actor_id_env, 0] = self.cfg_task.randomize.nut_pos_xy_initial[0] + nut_noise_xy[
env_ids, 0]
self.root_pos[env_ids, self.nut_actor_id_env, 1] = self.cfg_task.randomize.nut_pos_xy_initial[1] + nut_noise_xy[
env_ids, 1]
self.root_pos[
env_ids, self.nut_actor_id_env, 2] = self.cfg_base.env.table_height - self.bolt_head_heights.squeeze(-1)
self.root_quat[env_ids, self.nut_actor_id_env] = torch.tensor([0.0, 0.0, 0.0, 1.0], dtype=torch.float32,
device=self.device).repeat(len(env_ids), 1)
self.root_linvel[env_ids, self.nut_actor_id_env] = 0.0
self.root_angvel[env_ids, self.nut_actor_id_env] = 0.0
# Randomize root state of bolt
bolt_noise_xy = 2 * (torch.rand((self.num_envs, 2), dtype=torch.float32, device=self.device) - 0.5) # [-1, 1]
bolt_noise_xy = bolt_noise_xy @ torch.diag(
torch.tensor(self.cfg_task.randomize.bolt_pos_xy_noise, device=self.device))
self.root_pos[env_ids, self.bolt_actor_id_env, 0] = self.cfg_task.randomize.bolt_pos_xy_initial[0] + \
bolt_noise_xy[env_ids, 0]
self.root_pos[env_ids, self.bolt_actor_id_env, 1] = self.cfg_task.randomize.bolt_pos_xy_initial[1] + \
bolt_noise_xy[env_ids, 1]
self.root_pos[env_ids, self.bolt_actor_id_env, 2] = self.cfg_base.env.table_height
self.root_quat[env_ids, self.bolt_actor_id_env] = torch.tensor([0.0, 0.0, 0.0, 1.0], dtype=torch.float32,
device=self.device).repeat(len(env_ids), 1)
self.root_linvel[env_ids, self.bolt_actor_id_env] = 0.0
self.root_angvel[env_ids, self.bolt_actor_id_env] = 0.0
nut_bolt_actor_ids_sim = torch.cat((self.nut_actor_ids_sim[env_ids],
self.bolt_actor_ids_sim[env_ids]),
dim=0)
self.gym.set_actor_root_state_tensor_indexed(self.sim,
gymtorch.unwrap_tensor(self.root_state),
gymtorch.unwrap_tensor(nut_bolt_actor_ids_sim),
len(nut_bolt_actor_ids_sim))
def _reset_buffers(self, env_ids):
"""Reset buffers."""
self.reset_buf[env_ids] = 0
self.progress_buf[env_ids] = 0
def _set_viewer_params(self):
"""Set viewer parameters."""
cam_pos = gymapi.Vec3(-1.0, -1.0, 1.0)
cam_target = gymapi.Vec3(0.0, 0.0, 0.5)
self.gym.viewer_camera_look_at(self.viewer, None, cam_pos, cam_target)
def _apply_actions_as_ctrl_targets(self, actions, ctrl_target_gripper_dof_pos, do_scale):
"""Apply actions from policy as position/rotation targets."""
# Interpret actions as target pos displacements and set pos target
pos_actions = actions[:, 0:3]
if do_scale:
pos_actions = pos_actions @ torch.diag(torch.tensor(self.cfg_task.rl.pos_action_scale, device=self.device))
self.ctrl_target_fingertip_midpoint_pos = self.fingertip_midpoint_pos + pos_actions
# Interpret actions as target rot (axis-angle) displacements
rot_actions = actions[:, 3:6]
if do_scale:
rot_actions = rot_actions @ torch.diag(torch.tensor(self.cfg_task.rl.rot_action_scale, device=self.device))
# Convert to quat and set rot target
angle = torch.norm(rot_actions, p=2, dim=-1)
axis = rot_actions / angle.unsqueeze(-1)
rot_actions_quat = torch_utils.quat_from_angle_axis(angle, axis)
if self.cfg_task.rl.clamp_rot:
rot_actions_quat = torch.where(angle.unsqueeze(-1).repeat(1, 4) > self.cfg_task.rl.clamp_rot_thresh,
rot_actions_quat,
torch.tensor([0.0, 0.0, 0.0, 1.0], device=self.device).repeat(self.num_envs,
1))
self.ctrl_target_fingertip_midpoint_quat = torch_utils.quat_mul(rot_actions_quat, self.fingertip_midpoint_quat)
if self.cfg_ctrl['do_force_ctrl']:
# Interpret actions as target forces and target torques
force_actions = actions[:, 6:9]
if do_scale:
force_actions = force_actions @ torch.diag(
torch.tensor(self.cfg_task.rl.force_action_scale, device=self.device))
torque_actions = actions[:, 9:12]
if do_scale:
torque_actions = torque_actions @ torch.diag(
torch.tensor(self.cfg_task.rl.torque_action_scale, device=self.device))
self.ctrl_target_fingertip_contact_wrench = torch.cat((force_actions, torque_actions), dim=-1)
self.ctrl_target_gripper_dof_pos = ctrl_target_gripper_dof_pos
self.generate_ctrl_signals()
def _get_keypoint_offsets(self, num_keypoints):
"""Get uniformly-spaced keypoints along a line of unit length, centered at 0."""
keypoint_offsets = torch.zeros((num_keypoints, 3), device=self.device)
keypoint_offsets[:, -1] = torch.linspace(0.0, 1.0, num_keypoints, device=self.device) - 0.5
return keypoint_offsets
def _get_keypoint_dist(self):
"""Get keypoint distance."""
keypoint_dist = torch.sum(torch.norm(self.keypoints_nut - self.keypoints_gripper, p=2, dim=-1), dim=-1)
return keypoint_dist
def _close_gripper(self, sim_steps=20):
"""Fully close gripper using controller. Called outside RL loop (i.e., after last step of episode)."""
self._move_gripper_to_dof_pos(gripper_dof_pos=0.0, sim_steps=sim_steps)
def _move_gripper_to_dof_pos(self, gripper_dof_pos, sim_steps=20):
"""Move gripper fingers to specified DOF position using controller."""
delta_hand_pose = torch.zeros((self.num_envs, self.cfg_task.env.numActions),
device=self.device) # No hand motion
self._apply_actions_as_ctrl_targets(delta_hand_pose, gripper_dof_pos, do_scale=False)
# Step sim
for _ in range(sim_steps):
self.render()
self.gym.simulate(self.sim)
def _lift_gripper(self, franka_gripper_width=0.0, lift_distance=0.3, sim_steps=20):
"""Lift gripper by specified distance. Called outside RL loop (i.e., after last step of episode)."""
delta_hand_pose = torch.zeros([self.num_envs, 6], device=self.device)
delta_hand_pose[:, 2] = lift_distance
# Step sim
for _ in range(sim_steps):
self._apply_actions_as_ctrl_targets(delta_hand_pose, franka_gripper_width, do_scale=False)
self.render()
self.gym.simulate(self.sim)
def _check_lift_success(self, height_multiple):
"""Check if nut is above table by more than specified multiple times height of nut."""
lift_success = torch.where(
self.nut_pos[:, 2] > self.cfg_base.env.table_height + self.nut_heights.squeeze(-1) * height_multiple,
torch.ones((self.num_envs,), device=self.device),
torch.zeros((self.num_envs,), device=self.device))
return lift_success
def _randomize_gripper_pose(self, env_ids, sim_steps):
"""Move gripper to random pose."""
# Set target pos above table
self.ctrl_target_fingertip_midpoint_pos = \
torch.tensor([0.0, 0.0, self.cfg_base.env.table_height], device=self.device) \
+ torch.tensor(self.cfg_task.randomize.fingertip_midpoint_pos_initial, device=self.device)
self.ctrl_target_fingertip_midpoint_pos = self.ctrl_target_fingertip_midpoint_pos.unsqueeze(0).repeat(self.num_envs, 1)
fingertip_midpoint_pos_noise = \
2 * (torch.rand((self.num_envs, 3), dtype=torch.float32, device=self.device) - 0.5) # [-1, 1]
fingertip_midpoint_pos_noise = \
fingertip_midpoint_pos_noise @ torch.diag(torch.tensor(self.cfg_task.randomize.fingertip_midpoint_pos_noise,
device=self.device))
self.ctrl_target_fingertip_midpoint_pos += fingertip_midpoint_pos_noise
# Set target rot
ctrl_target_fingertip_midpoint_euler = torch.tensor(self.cfg_task.randomize.fingertip_midpoint_rot_initial,
device=self.device).unsqueeze(0).repeat(self.num_envs, 1)
fingertip_midpoint_rot_noise = \
2 * (torch.rand((self.num_envs, 3), dtype=torch.float32, device=self.device) - 0.5) # [-1, 1]
fingertip_midpoint_rot_noise = fingertip_midpoint_rot_noise @ torch.diag(
torch.tensor(self.cfg_task.randomize.fingertip_midpoint_rot_noise, device=self.device))
ctrl_target_fingertip_midpoint_euler += fingertip_midpoint_rot_noise
self.ctrl_target_fingertip_midpoint_quat = torch_utils.quat_from_euler_xyz(
ctrl_target_fingertip_midpoint_euler[:, 0],
ctrl_target_fingertip_midpoint_euler[:, 1],
ctrl_target_fingertip_midpoint_euler[:, 2])
# Step sim and render
for _ in range(sim_steps):
self.refresh_base_tensors()
self.refresh_env_tensors()
self._refresh_task_tensors()
pos_error, axis_angle_error = fc.get_pose_error(
fingertip_midpoint_pos=self.fingertip_midpoint_pos,
fingertip_midpoint_quat=self.fingertip_midpoint_quat,
ctrl_target_fingertip_midpoint_pos=self.ctrl_target_fingertip_midpoint_pos,
ctrl_target_fingertip_midpoint_quat=self.ctrl_target_fingertip_midpoint_quat,
jacobian_type=self.cfg_ctrl['jacobian_type'],
rot_error_type='axis_angle')
delta_hand_pose = torch.cat((pos_error, axis_angle_error), dim=-1)
actions = torch.zeros((self.num_envs, self.cfg_task.env.numActions), device=self.device)
actions[:, :6] = delta_hand_pose
self._apply_actions_as_ctrl_targets(actions=actions,
ctrl_target_gripper_dof_pos=self.asset_info_franka_table.franka_gripper_width_max,
do_scale=False)
self.gym.simulate(self.sim)
self.render()
self.dof_vel[env_ids, :] = torch.zeros_like(self.dof_vel[env_ids])
# Set DOF state
multi_env_ids_int32 = self.franka_actor_ids_sim[env_ids].flatten()
self.gym.set_dof_state_tensor_indexed(self.sim,
gymtorch.unwrap_tensor(self.dof_state),
gymtorch.unwrap_tensor(multi_env_ids_int32),
len(multi_env_ids_int32))