-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathfw_pg.py
263 lines (233 loc) · 12.1 KB
/
fw_pg.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
from tqdm import tqdm
import numpy as np
import torch
import collections
import random
from agent.fix_rule_no_att.agent import Agent
MAX_STEP = 2000
class ReplayBuffer:
def __init__(self, capacity):
self.buffer = collections.deque(maxlen=capacity)
def add(self, state, action, reward, next_state, done):
self.buffer.append((state, action, reward, next_state, done))
def sample(self, batch_size):
transitions = random.sample(self.buffer, batch_size)
state, action, reward, next_state, done = zip(*transitions)
return np.array(state), action, reward, np.array(next_state), done
def size(self):
return len(self.buffer)
def moving_average(a, window_size):
cumulative_sum = np.cumsum(np.insert(a, 0, 0))
middle = (cumulative_sum[window_size:] - cumulative_sum[:-window_size]) / window_size
r = np.arange(1, window_size - 1, 2)
begin = np.cumsum(a[:window_size - 1])[::2] / r
end = (np.cumsum(a[:-window_size:-1])[::2] / r)[::-1]
return np.concatenate((begin, middle, end))
def train_on_policy_agent(env, agent, num_episodes):
return_list = []
for i in range(10):
with tqdm(total=int(num_episodes / 10), desc='Iteration %d' % i) as pbar:
for i_episode in range(int(num_episodes / 10)):
episode_return = 0
transition_dict = {'states': [], 'actions': [], 'next_states': [], 'rewards': [], 'dones': []}
state = env.reset()
done = False
while not done:
action = agent.take_action(state)
next_state, reward, done, _ = env.step(action)
transition_dict['states'].append(state)
transition_dict['actions'].append(action)
transition_dict['next_states'].append(next_state)
transition_dict['rewards'].append(reward)
transition_dict['dones'].append(done)
state = next_state
episode_return += reward
return_list.append(episode_return)
agent.update(transition_dict)
if (i_episode + 1) % 10 == 0:
pbar.set_postfix({'episode': '%d' % (num_episodes / 10 * i + i_episode + 1),
'return': '%.3f' % np.mean(return_list[-10:])})
pbar.update(1)
return return_list
def train_off_policy_agent(env, agent, num_episodes, replay_buffer, minimal_size, batch_size):
return_list = []
red_detector_action = []
blue_agent = Agent()
blue_agent_obs_ind = blue_agent.get_obs_ind()
size_x, size_y = env.get_map_size()
red_detector_num, red_fighter_num, blue_detector_num, blue_fighter_num = env.get_unit_num()
# set map info to blue agent
blue_agent.set_map_info(size_x, size_y, blue_detector_num, blue_fighter_num)
for i in range(10):
with tqdm(total=int(num_episodes / 10), desc='Iteration %d' % i) as pbar:
for i_episode in range(int(num_episodes / 10)):
episode_return = 0
# state: [-0.22665706 0.97397465 -0.44414896]
# state = env.reset() # 状态
env.reset()
state_org = env.get_obs_raw()
state_buffer = []
x1 = state_org[0]['fighter_obs_list'][0]['pos_x']
x2 = state_org[1]['fighter_obs_list'][0]['pos_x']
y1 = state_org[0]['fighter_obs_list'][0]['pos_y']
y2 = state_org[1]['fighter_obs_list'][0]['pos_y']
angle2 = np.float(state_org[0]['fighter_obs_list'][0]['course'])
angle2 = angle2 * np.pi / 180
k = (y2 - y1) / (x2 - x1 + 1e-6)
# 计算angle1
if x2 > x1 and y2 > y1:
angle = np.arctan(k)
elif x2 > x1 and y2 < y1:
angle = np.arctan(k) + np.pi * 2
elif x2 > x1 and y2 == y1:
angle = 0
elif x1 > x2:
angle = np.arctan(k) + np.pi
elif x1 == x2 and y1 > y2:
angle = np.pi * 3 / 2
elif x1 == x2 and y1 < y2:
angle = np.pi / 2
elif x1 == x2 and y1 == y2:
angle = angle2
else:
print('warning!')
# T = np.abs(angle - angle2)
T = angle-angle2
# state_buffer.append(np.cos(abs(T - np.pi)))
state_buffer.append(T)
state_buffer.append(angle2)
# state_buffer.append(np.sin(angle - angle2))
# state_buffer.append(abs(T - np.pi))
state = state_buffer
# print('state:',state)
done = False
for j in range(MAX_STEP):
action = np.array(agent.take_action(state)) # 策略网络输入状态，输出动作[-1,1]
# env.render = True # print('action:',action) ##动作预处理
obs_list = []
next_obs_list = []
r_action_list = []
b_action_list = []
if j == 0:
red_obs_dict, blue_obs_dict = env.get_obs()
blue_detector_action, blue_fighter_action = blue_agent.get_action(blue_obs_dict, j)
b_action_list.append([blue_fighter_action[0]['course'], 1, 0, 0])
# get red action
x1 = red_obs_dict['fighter_obs_list'][0]['pos_x']
y1 = red_obs_dict['fighter_obs_list'][0]['pos_y']
x2 = blue_obs_dict['fighter_obs_list'][0]['pos_x']
y2 = blue_obs_dict['fighter_obs_list'][0]['pos_y']
angle2 = red_obs_dict['fighter_obs_list'][0]['course']
# print('坐标：',x1,x2,y1,y2)
k = (y2 - y1) / (x2 - x1 + 1e-6)
# print('y:',y1,y2)
# print('敌机坐标:',x2,y2) angle2 = red_obs_dict['fighter_obs_list'][0]['course']
angle2 = angle2 * np.pi / 180
# print('angle2:',angle2)
# 计算angle
if x2 > x1 and y2 > y1:
angle = np.arctan(k)
elif x2 > x1 and y2 < y1:
angle = np.arctan(k) + np.pi * 2
elif x2 > x1 and y2 == y1:
angle = 0
elif x1 > x2:
angle = np.arctan(k) + np.pi
elif x1 == x2 and y1 > y2:
angle = np.pi * 3 / 2
elif x1 == x2 and y1 < y2:
angle = np.pi / 2
elif x1 == x2 and y1 == y2:
angle = angle2
else:
print('warning!')
# 计算angle2
temp_angle = angle
# T = np.abs(angle - angle2)
T = angle - angle2
# print('angle:',angle)
# obs_list.append(np.cos(abs(T - np.pi)))
obs_list.append(T)
obs_list.append(angle2)
d = float(np.sqrt((x1 - x2) ** 2 + (y1 - y2) ** 2))
if d <= 120:
attack = blue_obs_dict['fighter_obs_list'][0]['id']
print('attack:',attack)
else:
attack = 0
# state_buffer.append(np.sin(angle - angle2))
# obs_list.append(abs(T - np.pi))
# obs_list.append(angle - angle2)
env.step(np.array([red_detector_action]),np.array([[float((action+1)*180),1,0,0]]),
np.array([blue_detector_action]), np.array(b_action_list))
###################################################
red_obs_dict, blue_obs_dict = env.get_obs()
blue_detector_action, blue_fighter_action = blue_agent.get_action(blue_obs_dict, j+1)
x1 = red_obs_dict['fighter_obs_list'][0]['pos_x']
y1 = red_obs_dict['fighter_obs_list'][0]['pos_y']
x2 = blue_obs_dict['fighter_obs_list'][0]['pos_x']
y2 = blue_obs_dict['fighter_obs_list'][0]['pos_y']
# print('采取动作后航向：',red_obs_dict['fighter_obs_list'][0]['course'])
# print('动作值：',action*180)
k = (y2 - y1) / (x2 - x1 + 1e-6)
# 计算angle
angle2 = (action+1)*180
angle2 = angle2 * np.pi / 180
# 计算angle
if x2 > x1 and y2 > y1:
angle = np.arctan(k)
elif x2 > x1 and y2 < y1:
angle = np.arctan(k) + np.pi * 2
elif x2 > x1 and y2 == y1:
angle = 0
elif x1 > x2:
angle = np.arctan(k) + np.pi
elif x1 == x2 and y1 > y2:
angle = np.pi * 3 / 2
elif x1 == x2 and y1 < y2:
angle = np.pi / 2
elif x1 == x2 and y1 == y2:
angle = angle2
else:
print('warning!')
# d_std = np.sqrt((y2 - y1) ** 2 + (x2 - x1) ** 2) / 10
# T = np.abs(angle - angle2)
T = angle - angle2
# temp_delta = temp_angle - angle2
# print('temp_delta:',temp_delta)
next_obs_list.append(float(T))
next_obs_list.append(float(angle2))
# state_buffer.append(np.sin(angle - angle2))
# next_obs_list.append(abs(T - np.pi))
# next_obs_list.append(temp_delta)
# next_obs_list.append(d_std)
done = env.get_done()
r = -(angle2-temp_angle) ** 2
################################################
# next_state, reward, done, _ = env.step(action) # 环境采取动作返还下一步状态，奖励，是否结束
replay_buffer.add(obs_list, action, r[0], next_obs_list, done)
# print('info:',state, action, reward, next_state, done)
# sample # # [-0.87174714 0.48995608 -0.22461547] [0.02978992648422718] -6.919596472865825 [-0.8753325 0.48352158 0.14732009] False
state = next_obs_list
episode_return += r
if replay_buffer.size() > minimal_size:
b_s, b_a, b_r, b_ns, b_d = replay_buffer.sample(batch_size)
transition_dict = {'states': b_s, 'actions': b_a, 'next_states': b_ns, 'rewards': b_r, 'dones': b_d}
agent.update(transition_dict) # 更新
return_list.append(episode_return)
# ep_reward += reward
# print('Episode:', episode, 'Reward:%i' % int(episode_return))
if (i_episode+1) % 10 == 0:
pbar.set_postfix({'episode': '%d' % (num_episodes / 10 * i + i_episode + 1),
'return': '%.3f' % np.mean(return_list[-10:])})
pbar.update(1)
return return_list
def compute_advantage(gamma, lmbda, td_delta):
td_delta = td_delta.detach().numpy()
advantage_list = []
advantage = 0.0
for delta in td_delta[::-1]:
advantage = gamma * lmbda * advantage + delta
advantage_list.append(advantage)
advantage_list.reverse()
return torch.tensor(advantage_list, dtype=torch.float)