-
Notifications
You must be signed in to change notification settings - Fork 0
/
Agent.py
50 lines (40 loc) · 1.54 KB
/
Agent.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
from __future__ import print_function
import numpy as np
import random
import Environment
class AgentQLearn():
def __init__(self, env, curiosity, discount_rate = 0.9):
self.env = env
self.Q = [ [random.random()*0.5 for a in range(env.actions)] for s in range(env.states)]
self.explore_chance = curiosity
self.discount_rate = discount_rate
def learn(self, learning_rate, steps):
state = self.env.observe()
for step in range(steps):
if self.explore() :
action = random.randrange(self.env.actions)
else:
actions = self.Q[state]
action = actions.index(max(actions))
reward = self.env.act(action)
new_state = self.env.observe()
if(self.env.debug > 2):
self.env.print_state(state)
self.env.print_action(action)
print(reward, ' ', new_state)
next_action = self.Q[new_state].index(max(self.Q[new_state]))
self.Q[state][action] = self.Q[state][action] + learning_rate*(reward + self.discount_rate*self.Q[new_state][next_action] - self.Q[state][action])
state = new_state
def explore(self):
return random.random() < self.explore_chance
def print_policy(self, start):
self.env.set_position(start)
s = self.env.observe()
reward = 0
print(["{0:10}".format(i) for i in self.env.actions_name])
for s in range(self.env.states):
self.env.print_state(s)
action = self.Q[s].index(max(self.Q[s]))
self.env.print_action(action)
print('')
print(["{0:10.2f}".format(i) for i in self.Q[s]])