ActionDistribution object fixed

src/DecMCTS.py

@@ -15,30 +15,22 @@ class ActionDistribution:
     Action Distribution
     Working with action sequences and their respective probability
 
-    An Action Sequence here refers to the node_id of the relevant 
-        node in the Tree. The actual action sequence to get to that
-        node is stored inside the node.
-
     To initialise, Inputs:
-    - X: list of node_ids (ints)
-    - n: number of action sequences for comms
+    - X: list of action sequences
+    - q: probability of each action sequence (normalised in intialisation)
 
     """
     
-    def __init__(self, X, n, beta=0.5):
+    def __init__(self, X, q):
         
-        # Starting distribution is uniform
-        self.q = [1/n] * n
-
         # Action sequence as provided
         assert(len(X)==n)
         self.X = X
 
-        # Beta value set to initial value
-        self.beta = beta
-        self.n = n
+        # Normalise 
+        self.q = (np.array(q)/sum(q)).tolist()
+
 
-    
     def best_action(self):
         """
         Most likely action sequence
@@ -53,21 +45,6 @@ class ActionDistribution:
         return np.random.choice(self.X, p=self.q)
 
 
-    def update(self, Xnew):
-        """
-        Update Distribution
-        """
-
-        # If new Action sequences to consider, reset
-        if set(Xnew) != set(self.X):
-            self.X = Xnew
-            self.q = [1/self.n] * self.n
-        
-        else:
-            Eqn = "TODO"
-
-
-
 class Tree:
     """
     DecMCTS tree
@@ -97,7 +74,7 @@ class Tree:
             reward/available actions in coordination with others
     """
 
-    def __init__(self, data, reward_func, avail_actions_func, c_p=1):
+    def __init__(self, data, reward_func, avail_actions_func, comm_n, c_p=1):
 
         self.data = data
         self.graph = nx.DiGraph()
@@ -105,10 +82,12 @@ class Tree:
         self.available_actions = avail_actions_func
         self.c_p = c_p
         self.comms = {} # Plan with no robots initially
-        # TODO CHANGE FOLLOWING LINE
-        self.my_action_distribution = ActionDistribution(X,n)
+        self.comm_n = comm_n # number of action dists to communicate
+
+        # Set Action sequence as nothing for now
+        self.my_act_dist = ActionDistribution([[]],[1])
 
-        # Graph
+        # Graph add root node of tree
         self.graph.add_node(1, 
                 mu=0, 
                 N=0,