asim-rl/graph
2
0
Fork 0
Code Releases Activity

Plot graph

Browse Source
This commit is contained in:
drothschedl
2022-07-25 11:16:22 +02:00
parent 32209b3cd6
commit 21bcbf49c1
1 changed files with 5 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
DynaQ.py
View File

@@ -40,7 +40,6 @@ class DynaQ:
self.state = self.env.start self.state = self.env.start
self.state_actions = [] self.state_actions = []
self.step_in_episode = 0 self.step_in_episode = 0
self.env.reset()
''' '''
Learning method for agent Learning method for agent
@@ -50,6 +49,7 @@ class DynaQ:
def learn(self, epsilon_decay: float, epsilon_min: float, run: int) -> None: def learn(self, epsilon_decay: float, epsilon_min: float, run: int) -> None:
self.steps_per_episode = [] self.steps_per_episode = []
eps = self.epsilon eps = self.epsilon
self.env.reset()
for episode in range(self.episodes): for episode in range(self.episodes):
done = False done = False
self.reset() self.reset()
@@ -84,6 +84,10 @@ class DynaQ:
self.reset() self.reset()
print("Goal") print("Goal")
eps = max(epsilon_min, self.epsilon * np.exp(-epsilon_decay * episode)) eps = max(epsilon_min, self.epsilon * np.exp(-epsilon_decay * episode))
if run == 9 and self.n_steps == 50 and episode==69:
self.env.print_shortest_path(self.Q)
if run==9 and self.n_steps==50:
self.env.print_shortest_path(self.Q)
''' '''
Returns epsilon-greedy action Returns epsilon-greedy action