-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathrf_cartpole_balancing_gemma2
151 lines (109 loc) · 4.35 KB
/
rf_cartpole_balancing_gemma2
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
Generation 0
import numpy as np
def reward(action, obs):
"""Calculates the reward based on pole angle and angular velocity.
Args:
action: A ndarray with shape (1,) representing the force applied to the cart.
obs: A ndarray with shape (4,) representing the current state of the system:
[cart_position, pole_angle, cart_velocity, pole_angular_velocity].
Returns:
The reward as a float.
"""
pole_angle = obs[1]
pole_velocity = obs[3]
# Reward scaling factors for angle and velocity
angle_scale = 100.0
vel_scale = 5.0
# Apply shaping reward to encourage uprightness and slow-down
angle_reward = np.exp(-angle_scale * abs(pole_angle))
velocity_reward = np.exp(-vel_scale * abs(pole_velocity))
return angle_reward * velocity_reward
Generation 1
import numpy as np
def reward(action, obs):
"""Calculates the reward based on pole angle and angular velocity.
Args:
action: A ndarray with shape (1,) representing the force applied to the cart.
obs: A ndarray with shape (4,) representing the current state of the system:
[cart_position, pole_angle, cart_velocity, pole_angular_velocity].
Returns:
The reward as a float.
"""
pole_angle = obs[1]
pole_velocity = obs[3]
# Reward scaling factors for angle and velocity
angle_scale = 100.0
vel_scale = 5.0
# Apply shaping reward to encourage uprightness and slow-down
angle_reward = np.exp(-angle_scale * abs(pole_angle))
velocity_reward = np.exp(-vel_scale * abs(pole_velocity))
return angle_reward * velocity_reward
Generation 2
import numpy as np
def reward(action, obs):
"""Calculates the reward based on pole angle, angular velocity, and cart position.
Args:
action: A ndarray with shape (1,) representing the force applied to the cart.
obs: A ndarray with shape (4,) representing the current state of the system:
[cart_position, pole_angle, cart_velocity, pole_angular_velocity].
Returns:
The reward as a float.
"""
pole_angle = obs[1]
pole_velocity = obs[3]
cart_position = obs[0]
# Reward scaling factors for angle, velocity, and position
angle_scale = 100.0
vel_scale = 5.0
pos_scale = 0.1
# Apply shaping reward to encourage uprightness, slow-down, and staying near the center
angle_reward = np.exp(-angle_scale * abs(pole_angle))
velocity_reward = np.exp(-vel_scale * abs(pole_velocity))
position_reward = np.exp(-pos_scale * abs(cart_position))
return angle_reward * velocity_reward * position_reward
Generation 3
import numpy as np
def reward(action, obs):
"""Calculates the reward based on pole angle, angular velocity, and cart position.
Args:
action: A ndarray with shape (1,) representing the force applied to the cart.
obs: A ndarray with shape (4,) representing the current state of the system:
[cart_position, pole_angle, cart_velocity, pole_angular_velocity].
Returns:
The reward as a float.
"""
pole_angle = obs[1]
pole_velocity = obs[3]
cart_position = obs[0]
# Reward scaling factors for angle, velocity, and position
angle_scale = 100.0
vel_scale = 5.0
pos_scale = 0.1
# Apply shaping reward to encourage uprightness, slow-down, and staying near the center
angle_reward = np.exp(-angle_scale * abs(pole_angle))
velocity_reward = np.exp(-vel_scale * abs(pole_velocity))
position_reward = np.exp(-pos_scale * abs(cart_position))
return angle_reward * velocity_reward * position_reward
Generation 4
import numpy as np
def reward(action, obs):
"""Calculates the reward based on pole angle, angular velocity, and cart position.
Args:
action: A ndarray with shape (1,) representing the force applied to the cart.
obs: A ndarray with shape (4,) representing the current state of the system:
[cart_position, pole_angle, cart_velocity, pole_angular_velocity].
Returns:
The reward as a float.
"""
pole_angle = obs[1]
pole_velocity = obs[3]
cart_position = obs[0]
# Reward scaling factors for angle, velocity, and position
angle_scale = 100.0
vel_scale = 5.0
pos_scale = 0.1
# Apply shaping reward to encourage uprightness, slow-down, and staying near the center
angle_reward = np.exp(-angle_scale * abs(pole_angle))
velocity_reward = np.exp(-vel_scale * abs(pole_velocity))
position_reward = np.exp(-pos_scale * abs(cart_position))
return angle_reward * velocity_reward * position_reward