|
a |
|
b/rdpg/ou_noise.py |
|
|
1 |
# -------------------------------------- |
|
|
2 |
# Ornstein-Uhlenbeck Noise |
|
|
3 |
# Author: Flood Sung |
|
|
4 |
# Date: 2016.5.4 |
|
|
5 |
# Reference: https://github.com/rllab/rllab/blob/master/rllab/exploration_strategies/ou_strategy.py |
|
|
6 |
# -------------------------------------- |
|
|
7 |
|
|
|
8 |
import numpy as np |
|
|
9 |
import numpy.random as nr |
|
|
10 |
|
|
|
11 |
class OUNoise: |
|
|
12 |
"""docstring for OUNoise""" |
|
|
13 |
def __init__(self,action_dimension,mu=0, theta=0.15, sigma=0.2): |
|
|
14 |
self.action_dimension = action_dimension |
|
|
15 |
self.mu = mu |
|
|
16 |
self.theta = theta |
|
|
17 |
self.sigma = sigma |
|
|
18 |
self.state = np.ones(self.action_dimension) * self.mu |
|
|
19 |
self.reset() |
|
|
20 |
|
|
|
21 |
def reset(self): |
|
|
22 |
self.state = np.ones(self.action_dimension) * self.mu |
|
|
23 |
|
|
|
24 |
def noise(self): |
|
|
25 |
x = self.state |
|
|
26 |
dx = self.theta * (self.mu - x) + self.sigma * nr.randn(len(x)) |
|
|
27 |
self.state = x + dx |
|
|
28 |
return self.state |
|
|
29 |
|
|
|
30 |
if __name__ == '__main__': |
|
|
31 |
ou = OUNoise(3) |
|
|
32 |
states = [] |
|
|
33 |
for i in range(1000): |
|
|
34 |
states.append(ou.noise()) |
|
|
35 |
import matplotlib.pyplot as plt |
|
|
36 |
|
|
|
37 |
plt.plot(states) |
|
|
38 |
plt.show() |