-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathreferee.py
More file actions
161 lines (127 loc) · 5.82 KB
/
referee.py
File metadata and controls
161 lines (127 loc) · 5.82 KB
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
import rospy
import random
from gazebo_msgs.msg import ModelState
from sensor_msgs.msg import LaserScan
import math
import torch
import agent
import policy
import human_controller
class Referee(object):
def __init__(self):
self.epsilon = 0.05
self.green_safe = [30 for i in range(360)]
self.red_safe = [30 for i in range(360)]
rospy.init_node("referee", anonymous='True')
self.kidnapper_pub = rospy.Publisher('/gazebo/set_model_state', ModelState, queue_size=1)
rospy.Subscriber('/hokuyo_safe_green', LaserScan, self.callback)
rospy.Subscriber('/hokuyo_safe_red', LaserScan, self.callback2)
self.rate = rospy.Rate(20)
self.time = 0
self.agent_green = agent.Agent(name='green')
self.agent_red = agent.Agent(name='red')
self.expert = human_controller.Controller()
self.target_green = "goal_box"
self.target_red = "goal_box_0"
self.green_target = [-5.8, 5.37]
self.red_target = [5.53, -5.5]
self.agent_green.update_target(self.green_target)
self.agent_red.update_target(self.red_target)
self.generate_target('green')
self.generate_target('red')
self.dagger_policy = policy.Policy()
def reset_game(self):
self.agent_green.reset()
self.agent_red.reset()
agent_red_pose_x, agent_red_pose_y = self.generate_random_pose()
self.kidnapper(self.agent_red.name, [agent_red_pose_x, agent_red_pose_y])
agent_green_pose_x, agent_green_pose_y = self.generate_random_pose()
self.kidnapper(self.agent_green.name, [agent_green_pose_x, agent_green_pose_y])
target_red_pose_x ,target_red_pose_y = self.generate_random_pose()
target_green_pose_x, target_green_pose_y = self.generate_random_pose()
self.kidnapper(self.target_green, [target_green_pose_x, target_green_pose_y])
self.kidnapper(self.target_red, [target_red_pose_x, target_red_pose_y])
self.green_safe = [30 for i in range(360)]
self.red_safe = [30 for i in range(360)]
self.agent_green.update_target(self.green_target)
self.agent_red.update_target(self.red_target)
self.generate_target('green')
self.generate_target('red')
def generate_random_pose(self):
pose_x = (13.6 * random.random()) - 6.8
pose_y = (13.6 * random.random()) - 6.8
while abs(pose_x) + abs(pose_y) < 6:
pose_x = (13.6 * random.random()) - 6.8
pose_y = (13.6 * random.random()) - 6.8
return pose_x, pose_y
def generate_target(self, color):
if color == 'green':
pose = self.generate_random_pose()
self.green_target = pose
print('kidnapped:', self.target_green)
self.kidnapper(self.target_green, pose)
if color == 'red':
pose = self.generate_random_pose()
self.red_target = pose
self.kidnapper(self.target_red, pose)
def tick(self):
self.agent_green.update_target(self.green_target)
self.agent_red.update_target(self.red_target)
safety = self.check_collision()
if safety[0]:
self.agent_green.reset()
pose_x, pose_y = self.generate_random_pose()
self.kidnapper(self.agent_green.name, [pose_x, pose_y])
if safety[1]:
self.agent_red.reset()
pose_x, pose_y = self.generate_random_pose()
self.kidnapper(self.agent_red.name, [pose_x, pose_y])
self.check_goal()
expert_action = self.expert.command
green_state= self.agent_green.ask_state()
self.dagger_policy.store_transition(green_state, expert_action)
random_epsilon = random.random()
if random_epsilon > self.epsilon:
self.agent_green.tick(self.expert.command)
else:
self.agent_green.tick(torch.max(self.dagger_policy.choose_action(green_state)[0], 0)[1].item())
red_state = self.agent_red.ask_state()
self.agent_red.tick(torch.max(self.dagger_policy.choose_action(red_state)[0], 0)[1].item())
self.dagger_policy.learn()
if self.epsilon < 0.95:
self.epsilon += 0.000015
if self.time % 100 == 0:
print(self.epsilon)
self.expert.command = 0
if self.dagger_policy.train_num % 3000 == 1:
self.dagger_policy.save()
self.rate.sleep()
self.time += 1
def check_collision(self):
return [(min(self.green_safe) < 0.52) or (self.dist() < 1.05), (min(self.red_safe) < 0.52) or (self.dist() < 1.05)]
def check_goal(self):
if self.time % 100 == 0:
print("goal_dist", math.sqrt((self.agent_green.x - self.green_target[0]) ** 2 + (self.agent_green.y - self.green_target[1]) ** 2))
if math.sqrt((self.agent_green.x - self.green_target[0]) ** 2 + (self.agent_green.y - self.green_target[1]) ** 2) < 1.5:
self.agent_green.update_target(self.green_target)
self.generate_target('green')
if math.sqrt((self.agent_red.x - self.red_target[0]) ** 2 + (self.agent_red.y - self.red_target[1]) ** 2) < 1.5:
self.agent_red.update_target(self.red_target)
self.generate_target('red')
def kidnapper(self, target, pose):
modelstate = ModelState()
modelstate.model_name = target
modelstate.pose.position.x = pose[0]
modelstate.pose.position.y = pose[1]
modelstate.pose.position.z = 0.125
self.kidnapper_pub.publish(modelstate)
self.rate.sleep()
def dist(self):
return math.sqrt((self.agent_green.x - self.agent_red.x) ** 2 + (self.agent_green.y - self.agent_red.y) ** 2)
def callback(self, data):
self.green_safe = data.ranges
def callback2(self, data):
self.red_safe = data.ranges
game = Referee()
while not rospy.is_shutdown():
game.tick()