fork download
copy 
  1. S0 = "S0"
  2. S1 = "S1"
  3. S2 = "S2"
  4. BLUE = "blue"
  5. RED = "red"
  6. gamma = 0.5  # TODO MODIFY GAMMA HERE
  7.  
  8. # P(s'|s,a)
  9. P_destination_start_action = \
  10. {(S0,S0, BLUE):0.5,(S0,S0,RED):0.9, (S0,S1,BLUE):0.8,(S0,S1,RED):0,   (S0,S2, BLUE):0,(S0,S2,RED):0,
  11. (S1,S0, BLUE):0.5,(S1,S0,RED):0,   (S1,S1,BLUE):0.2,(S1,S1,RED):0.6, (S1,S2, BLUE):0,(S1,S2,RED):0,
  12. (S2,S0, BLUE):0,  (S2,S0,RED):0.1, (S2,S1,BLUE):0  ,(S2,S1,RED):0.4, (S2,S2, BLUE):1,(S2,S2,RED):1}
  13.  
  14. class MDP:
  15.     def __init__(self):
  16.         self.states = [S0, S1, S2]
  17.         self.actions = [BLUE, RED]
  18.  
  19.  
  20.         self.P_dest_start_action = P_destination_start_action
  21.         self.rewards = {S0: -2, S1: -5, S2: 0}
  22.  
  23. def POLICY_EVALUATION(policy_vec, utility_vec, mdp):
  24.     new_utility_vector = {}
  25.     for s in mdp.states:
  26.         to_sum = [(mdp.P_dest_start_action[(s_tag, s, policy_vec[s])] * utility_vec[s_tag])
  27.                   for s_tag in mdp.states]
  28.         new_utility_vector[s] = mdp.rewards[s] + gamma * sum(to_sum)
  29.     return new_utility_vector
  30.  
  31. def POLICY_ITERATION(mdp):
  32.     utility_vector = {state: 0 for state in mdp.states}
  33.     policy_vector = {S0: BLUE, S1: RED, S2: RED}
  34.     unchanged = False
  35.  
  36.     while not unchanged:
  37.         utility_vector = POLICY_EVALUATION(policy_vector, utility_vector, mdp)
  38.         unchanged = True
  39.         for s in mdp.states:
  40.             BLUE_sum = sum([(mdp.P_dest_start_action[(s_tag, s, BLUE)] * utility_vector[s_tag])
  41.                             for s_tag in mdp.states])
  42.             RED_sum = sum([(mdp.P_dest_start_action[(s_tag, s, RED)] * utility_vector[s_tag])
  43.                            for s_tag in mdp.states])
  44.             if policy_vector[s] == RED and BLUE_sum > RED_sum:
  45.                 policy_vector[s] = BLUE
  46.                 unchanged = False
  47.  
  48.             elif policy_vector[s] == BLUE and RED_sum > BLUE_sum:
  49.                 policy_vector[s] = RED
  50.                 unchanged = False
  51.  
  52.     return policy_vector
  53.  
  54. if __name__ == "__main__":
  55.     Q2_mdp = MDP()
  56.     new_policy_vec = POLICY_ITERATION(Q2_mdp)
  57.     print("===========================END===============================")
  58.     print("S_O policy =", new_policy_vec[S0], " ,S_1 Policy =", new_policy_vec[S1])
  59.  
  60.  
Success #stdin #stdout 0.04s 9512KB
comments (?)
stdin
copy 
Standard input is empty
stdout
copy 
===========================END===============================
S_O policy = red  ,S_1 Policy = blue
https://ideone.com/JLl5e4
language:
Python 3 nbc (python 3.7.3)
created:
7 years ago
visibility:
public

Share or Embed source code

Discover > Sphere Engine API

The brand new service which powers Ideone!

Discover > IDE Widget

Widget for compiling and running the source code in a web browser!