-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathcollect_data.py
210 lines (189 loc) · 7.61 KB
/
collect_data.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
# import hydra
# import other_utils
from taskEnv import ComeHereEnv
# from agent.action_decoder import ActionDecoder
# from agent.features import FeatureFusion
# from agent.ppo import PPOAgent
# import torch
import ligent
import time
import numpy as np
import matplotlib
import matplotlib.pyplot as plt
import os
import math
import h5py
matplotlib.use('Agg')
def save_pic(np_arr, pic_path):
fig, ax = plt.subplots()
ax.imshow(np_arr)
ax.axis('off')
fig.set_size_inches(np_arr.shape[1]/100, np_arr.shape[0]/100)
plt.savefig(pic_path,bbox_inches='tight', pad_inches=0, transparent=True)
plt.close()
# def degree(a, b):
# inner_product = lambda a,b: max(min(a[0]*b[0]+a[1]*b[1], 1), -1)
# outer_product = lambda a,b: a[0]*b[1]-a[1]*b[0]
# degree = math.acos(inner_product(a, b)) / math.pi * 180
# return degree
def distance(a,b):
return ((a[0]-b[0])**2 + (a[1]-b[1])**2)**0.5
def inference_action(info):
position_player = info['game_states']['player']['position']
position_mate = info['game_states']['playmate']['position']
face_direction_player = info['game_states']['player']['forward']
face_direction_mate = info['game_states']['playmate']['forward']
mate2player = np.array([position_player['x']-position_mate['x'], position_player['z']-position_mate['z']])
mate2player /= (mate2player[0]**2 + mate2player[1]**2)**0.5
mate_direction = np.array([face_direction_mate['x'], face_direction_mate['z']])
mate_direction /= (mate_direction[0]**2 + mate_direction[1]**2)**0.5
def deg(v1,v2):
# 计算向量的点积
dot_product = v1[0] * v2[0] + v1[1] * v2[1] # 计算向量的叉积
cross_product = v1[0] * v2[1] - v1[1] * v2[0] # 计算带符号夹角（弧度）
angle = math.atan2(cross_product, dot_product)
ret_degree = angle/math.pi * 180
# print(degree)
return ret_degree
degree = deg(mate2player, mate_direction)
# inner_product = lambda a,b: max(min(a[0]*b[0]+a[1]*b[1], 1), -1)
# outer_product = lambda a,b: a[0]*b[1]-a[1]*b[0]
# degree = math.acos(inner_product(mate_direction, mate2player)) / math.pi * 180
def action_encoder(action_inst):
inst = np.zeros(1)
# inst[1] = action_inst['move_forward']
# inst[2] = 1 if action_inst['look_yaw']==30 else 0
# inst = action_inst['move_forward'] == 1
if action_inst['move_forward']==1:
inst = 1
elif action_inst['look_yaw'] == 30:
inst = 2
else:
inst = 0
return inst
if abs(degree) <= 15:
action_env = {
"move_right": 0,
"move_forward": 1,
"look_yaw": 0.0,
"look_pitch": 0.0,
"jump": False,
"grab": False,
"speak": "",
}
action_str = 'forward'
elif degree>=-30 and degree<-15:
action_env = {
"move_right": 0,
"move_forward": 0,
"look_yaw": -30,
"look_pitch": 0.0,
"jump": False,
"grab": False,
"speak": "",
}
action_str = "left30d"
else:
action_env = {
"move_right": 0,
"move_forward": 0,
"look_yaw": 30,
"look_pitch": 0.0,
"jump": False,
"grab": False,
"speak": "",
}
action_str = 'right30d'
return action_env, action_encoder(action_env), action_str
def collect(env,episodes:int):
env_decoder = ComeHereEnv(distance_reward=10, success_reward=200, distance_min=1.2, step_penalty=1, episode_len=100, is_debug=True)
action_noop = {
"move_right": 0,
"move_forward": 0,
"look_yaw": 0.0,
"look_pitch": 0.0,
"jump": False,
"grab": False,
"speak": "",
}
returns = []
if not os.path.exists('./dataset'):
os.makedirs('./dataset')
with h5py.File('./dataset/Episode1000.h5', 'w') as f:
data_nums = 0
obs_set = f.create_dataset('obs_V', (0,56,56,3), maxshape=(None, 56, 56, 3), dtype='i')
action_set = f.create_dataset('action', (0,2), maxshape=(None, 2), dtype='i')
for episode in range(episodes):
path = f"./obs_visions/episode_{episode}"
if not os.path.exists(path):
os.makedirs(path)
state_img, _ = env.reset()
env_decoder.reset()
done, blocked = False, False
(state_img, state_text), _, _, info = env.step(**action_noop)
while not (done or blocked):
last_state_img = state_img
action_env, action_code, action_str = inference_action(info)
(state_img, _), _, _, info = env.step(**action_env)
reward, done, blocked, cumulate_reward, elspsed_step, distance_info = env_decoder.step(info)
save_pic(last_state_img, pic_path=path+f'/{elspsed_step:03d}_{action_str}_{round(reward,1)}.png')
data_nums += 1
obs_set.resize((data_nums, 56,56,3))
action_set.resize((data_nums, 2))
obs_set[data_nums-1] = last_state_img
action_set[data_nums-1] = action_code
os.rename(path, path + f"_{elspsed_step}_{distance_info}_{cumulate_reward}")
print(returns)
env.close()
def debug(env):
action_noop = {
"move_right": 0,
"move_forward": 0,
"look_yaw": 0.0,
"look_pitch": 0.0,
"jump": False,
"grab": False,
"speak": "",
}
action_right15 = {
"move_right": 0,
"move_forward": 1,
"look_yaw": 0,
"look_pitch": 0.0,
"jump": False,
"grab": False,
"speak": "",
}
ligent.set_scenes_dir("C:/Users/19355/Desktop/drlProject/LIGENT/custom_scenes")
# env = ligent.Environment(path="C:/Users/19355/Desktop/drlProject/05272014_fix_multi_rotate/305272014_fix_multi_rotate/LIGENT.exe")
env_decoder = ComeHereEnv(distance_reward=10, success_reward=200, distance_min=1.2, step_penalty=1, episode_len=100, is_debug=True)
(state_img, _), _, _, info = env.step(**action_noop)
print(info['game_states']['playmate']['forward'], flush=True)
last_forward = info['game_states']['playmate']['forward']
last_position = info['game_states']['playmate']['position']
last_time = time.time()
for step in range(360//15):
(state_img, _), _, _, info = env.step(**action_right15)
print(f"Cost {time.time()-last_time} s!")
last_time = time.time()
# reward, done, blocked, cumulate_reward, elspsed_step, distance_info = env_decoder.step(info)
print(info['game_states']['playmate']['forward'], flush=True)
current_forward = info['game_states']['playmate']['forward']
current_position = info['game_states']['playmate']['position']
print(degree([last_forward['x'],last_forward['z']],[current_forward['x'],current_forward['z']]))
print(distance([current_position['x'],current_position['z']],[last_position['x'],last_position['z']]))
# time.sleep(3)
last_forward = current_forward
last_position = current_position
env.close()
ligent.set_scenes_dir("")
if __name__ == "__main__":
# with Display(visible=False) as disp:
try:
ligent.set_scenes_dir("")
env = ligent.Environment(path="C:/Users/19355/Desktop/drlProject/05272014_fix_multi_rotate/305272014_fix_multi_rotate/LIGENT.exe")
collect(env,1000)
# debug(env)
except:
env.close()
# debug()