-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathCart_Pole.py
188 lines (156 loc) · 5.94 KB
/
Cart_Pole.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
#Imports
import math
import gym
from gym import spaces, logger
from gym.utils import seeding
import numpy as np
#CartPole Gym Enviornment
class CartPoleEnv(gym.Env):
#Metadata
metadata = {
'render.modes': ['human', 'rgb_array'],
'video.frames_per_second' : 50
}
#Enviornment Initialization Function
def __init__(self):
#Gravity
self.gravity = 9.8
#Mass
self.masscart = 1.0
self.masspole = 0.1
self.total_mass = (self.masspole + self.masscart)
#Length
self.length = 0.5
self.polemass_length = (self.masspole * self.length)
#Force
self.force_mag = 10.0
self.tau = 0.02
self.kinematics_integrator = 'euler'
#Angle at which episode fails
self.theta_threshold_radians = 12 * 2 * math.pi / 360
self.x_threshold = 2.4
#Angle limits using numpy array
high = np.array([
self.x_threshold * 2,
np.finfo(np.float32).max,
self.theta_threshold_radians * 2,
np.finfo(np.float32).max])
#Action Space
self.action_space = spaces.Discrete(2)
#Observation Space
self.observation_space = spaces.Box(-high, high, dtype=np.float32)
#Seed
self.seed()
self.viewer = None
self.state = None
self.steps_beyond_done = None
#Randomized seed function
def seed(self, seed=None):
self.np_random, seed = seeding.np_random(seed)
return [seed]
#Step function
def step(self, action):
assert self.action_space.contains(action), "%r (%s) invalid"%(action, type(action))
state = self.state
x, x_dot, theta, theta_dot = state
force = self.force_mag if action==1 else -self.force_mag
#Mathmatics and Constants
costheta = math.cos(theta)
sintheta = math.sin(theta)
temp = (force + self.polemass_length * theta_dot * theta_dot * sintheta) / self.total_mass
thetaacc = (self.gravity * sintheta - costheta* temp) / (self.length * (4.0/3.0 - self.masspole * costheta * costheta / self.total_mass))
xacc = temp - self.polemass_length * thetaacc * costheta / self.total_mass
#Kinematics
if self.kinematics_integrator == 'euler':
x = x + self.tau * x_dot
x_dot = x_dot + self.tau * xacc
theta = theta + self.tau * theta_dot
theta_dot = theta_dot + self.tau * thetaacc
else:
x_dot = x_dot + self.tau * xacc
x = x + self.tau * x_dot
theta_dot = theta_dot + self.tau * thetaacc
theta = theta + self.tau * theta_dot
self.state = (x,x_dot,theta,theta_dot)
done = x < -self.x_threshold \
or x > self.x_threshold \
or theta < -self.theta_threshold_radians \
or theta > self.theta_threshold_radians
done = bool(done)
#Reward Structure
if not done:
reward = 1.0
elif self.steps_beyond_done is None:
self.steps_beyond_done = 0
reward = 1.0
else:
if self.steps_beyond_done == 0:
logger.warn("Warning call reset when done")
self.steps_beyond_done += 1
reward = 0.0
#Return state and reward
return np.array(self.state), reward, done, {}
#Reset enviorment
def reset(self):
self.state = self.np_random.uniform(low=-0.05, high=0.05, size=(4,))
self.steps_beyond_done = None
return np.array(self.state)
#Render Enviorment
def render(self, mode='human'):
#Screen
screen_width = 600
screen_height = 400
world_width = self.x_threshold*2
scale = screen_width/world_width
#Cart metrics
carty = 100
polewidth = 10.0
polelen = scale * (2 * self.length)
cartwidth = 50.0
cartheight = 30.0
#View
if self.viewer is None:
from gym.envs.classic_control import rendering
self.viewer = rendering.Viewer(screen_width, screen_height)
l,r,t,b = -cartwidth/2, cartwidth/2, cartheight/2, -cartheight/2
axleoffset =cartheight/4.0
cart = rendering.FilledPolygon([(l,b), (l,t), (r,t), (r,b)])
#Transformation
self.carttrans = rendering.Transform()
cart.add_attr(self.carttrans)
self.viewer.add_geom(cart)
l,r,t,b = -polewidth/2,polewidth/2,polelen-polewidth/2,-polewidth/2
#Pole
pole = rendering.FilledPolygon([(l,b), (l,t), (r,t), (r,b)])
pole.set_color(.8,.6,.4)
self.poletrans = rendering.Transform(translation=(0, axleoffset))
pole.add_attr(self.poletrans)
pole.add_attr(self.carttrans)
self.viewer.add_geom(pole)
#Cart axle
self.axle = rendering.make_circle(polewidth/2)
self.axle.add_attr(self.poletrans)
self.axle.add_attr(self.carttrans)
self.axle.set_color(.5,.5,.8)
self.viewer.add_geom(self.axle)
#Track
self.track = rendering.Line((0,carty), (screen_width,carty))
self.track.set_color(0,0,0)
self.viewer.add_geom(self.track)
self._pole_geom = pole
if self.state is None: return None
#Edit the pole
pole = self._pole_geom
l,r,t,b = -polewidth/2,polewidth/2,polelen-polewidth/2,-polewidth/2
pole.v = [(l,b), (l,t), (r,t), (r,b)]
x = self.state
cartx = x[0]*scale+screen_width/2.0
#Middle of cart
self.carttrans.set_translation(cartx, carty)
self.poletrans.set_rotation(-x[2])
return self.viewer.render(return_rgb_array = mode=='rgb_array')
#Close enviornment
def close(self):
if self.viewer:
self.viewer.close()
self.viewer = None