2211567
/
MLE
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

good enough i guess

master
Ruben Seitz 2025-11-27 16:09:47 +01:00
parent 85f81e5f23
commit 48a351518d
1 changed files with 28 additions and 30 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

16
04_pacman_rl/reinforcement_learning.py
View File

@ -72,20 +72,18 @@ def q_init():
# print(list(q_table.items())[:5]) # Uncomment to see the first 5 entries # print(list(q_table.items())[:5]) # Uncomment to see the first 5 entries
return q_table return q_table
def epsilon_greedy(q, s, epsilon=0.1): def epsilon_greedy(q, s, epsilon=0.025):
""" """
Return which direction Pacman should move to using epsilon-greedy algorithm Return which direction Pacman should move to using epsilon-greedy algorithm
With probability epsilon, choose a random action. Otherwise choose the greedy action. With probability epsilon, choose a random action. Otherwise choose the greedy action.
Avoids actions that would result in collision with ghost. Avoids actions that would result in collision with ghost.
""" """
# if np.random.random() < epsilon: if np.random.random() < epsilon:
# # Explore: choose random action (excluding blocked actions with Q=0) # Explore: choose random action (excluding blocked actions with Q=0)
# valid_actions = [i for i in range(len(q[s])) if q[s][i] is not None] valid_actions = [i for i in range(len(q[s])) if q[s][i] is not None]
# if valid_actions: return np.random.choice(valid_actions)
# return np.random.choice(valid_actions)
# else: else:
# return np.random.randint(0, len(q[s]))
# else:
# Get all valid (non-blocked) actions with their Q-values # Get all valid (non-blocked) actions with their Q-values
valid_actions = [(i, q[s][i]) for i in range(len(q[s])) if q[s][i] is not None] valid_actions = [(i, q[s][i]) for i in range(len(q[s])) if q[s][i] is not None]