-
Notifications
You must be signed in to change notification settings - Fork 5
/
ppo.py
150 lines (130 loc) · 4.78 KB
/
ppo.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
import numpy as np
from stable_baselines3 import PPO
from sofa_env.scenes.rope_threading.rope_threading_env import RenderMode, ObservationType, RopeThreadingEnv
from sofa_zoo.common.sb3_setup import configure_learning_pipeline
from sofa_zoo.common.lapgym_experiment_parameters import CONFIG, PPO_KWARGS
if __name__ == "__main__":
add_render_callback = True
continuous_actions = True
normalize_reward = True
reward_clip = np.inf
# observations, bimanual, randomized, eyes
parameters = ["STATE", "True", "True", "2"]
observation_type = ObservationType[parameters[0]]
image_based = observation_type in [ObservationType.RGB, ObservationType.RGBD]
eye_configs = {
"1": [
(60, 10, 0, 90),
],
"2": [
(60, 10, 0, 90),
(10, 10, 0, 90),
],
}
bimanual_grasp = parameters[1] == "True"
randomized_eye = parameters[2] == "True"
image_based = observation_type in [ObservationType.RGB, ObservationType.RGBD]
env_kwargs = {
"image_shape": (64, 64),
"window_size": (600, 600),
"observation_type": observation_type,
"time_step": 0.01,
"frame_skip": 10,
"settle_steps": 20,
"render_mode": RenderMode.HEADLESS if image_based or add_render_callback else RenderMode.NONE,
"reward_amount_dict": {
"passed_eye": 10.0,
"lost_eye": -20.0, # more than passed_eye
"goal_reached": 100.0,
"distance_to_active_eye": -0.0,
"lost_grasp": -0.1,
"collision": -0.1,
"floor_collision": -0.1,
"bimanual_grasp": 0.0,
"moved_towards_eye": 200.0,
"moved_away_from_eye": -200.0,
"workspace_violation": -0.01,
"state_limit_violation": -0.01,
"distance_to_lost_rope": -0.0,
"delta_distance_to_lost_rope": -0.0,
"fraction_rope_passed": 0.0,
"delta_fraction_rope_passed": 200.0,
},
"create_scene_kwargs": {
"eye_config": eye_configs[parameters[3]],
"randomize_gripper": True,
"start_grasped": True,
"randomize_grasp_index": True,
},
"on_reset_callbacks": None,
"color_eyes": True,
"individual_agents": False,
"only_right_gripper": not bimanual_grasp,
"fraction_of_rope_to_pass": 0.05,
"num_rope_tracking_points": 10,
}
if bimanual_grasp:
env_kwargs["reward_amount_dict"]["bimanual_grasp"] = 100.0
env_kwargs["reward_amount_dict"]["distance_to_bimanual_grasp"] = -0.0
env_kwargs["reward_amount_dict"]["delta_distance_to_bimanual_grasp"] = -200.0
if randomized_eye:
env_kwargs["create_scene_kwargs"]["eye_reset_noise"] = {
"low": np.array([-20.0, -20.0, 0.0, -15]),
"high": np.array([20.0, 20.0, 0.0, 15]),
}
config = {"max_episode_steps": 200 + 150 * (len(eye_configs[parameters[3]]) - 1), **CONFIG}
if image_based:
ppo_kwargs = PPO_KWARGS["image_based"]
else:
ppo_kwargs = PPO_KWARGS["state_based"]
info_keywords = [
"distance_to_active_eye",
"lost_grasps",
"recovered_lost_grasps",
"passed_eyes",
"lost_eyes",
"collisions",
"floor_collisions",
"successful_task",
"rew_delta_distance",
"rew_absolute_distance",
"rew_losing_eyes",
"rew_losing_grasp",
"rew_collisions",
"rew_floor_collisions",
"rew_workspace_violation",
"rew_state_limit_violation",
"rew_dist_to_lost_rope",
"rew_delt_dist_to_lost_rope",
"rew_passed_eyes",
"rew_bimanual_grasp",
"rew_dist_to_bimanual_grasp",
"rew_delt_dist_to_bimanual_grasp",
"rew_fraction_passed",
"rew_delta_fraction_passed",
]
config["ppo_config"] = ppo_kwargs
config["env_kwargs"] = env_kwargs
config["info_keywords"] = info_keywords
model, callback = configure_learning_pipeline(
env_class=RopeThreadingEnv,
env_kwargs=env_kwargs,
pipeline_config=config,
monitoring_keywords=info_keywords,
normalize_observations=False if image_based else True,
algo_class=PPO,
algo_kwargs=ppo_kwargs,
render=add_render_callback,
normalize_reward=normalize_reward,
reward_clip=reward_clip,
use_watchdog_vec_env=True,
watchdog_vec_env_timeout=20.0,
reset_process_on_env_reset=False,
)
model.learn(
total_timesteps=config["total_timesteps"],
callback=callback,
tb_log_name= f"{parameters[0]}_{parameters[1]}Biman_{parameters[2]}Random_{parameters[3]}",
)
log_path = str(model.logger.dir)
model.save(log_path + "saved_model.pth")