decmcts for scheduler, major bug fix

DecMCTS.py

@@ -105,16 +105,17 @@ class Tree:
         self.comms = {} # Plan with no robots initially
         self.comm_n = comm_n # number of action dists to communicate
 
-        # Set Action sequence as nothing for now
-        self.my_act_dist = ActionDistribution([0],[1])
-
         # Graph add root node of tree
         self.graph.add_node(1, 
                 mu=0, 
                 N=0, 
-                state=self.state_store(self.data, None, None)
+                state=self.state_store(self.data, None, None, self.id)
                 )
 
+        # Set Action sequence as nothing for now
+        self.my_act_dist = ActionDistribution([self.graph.node[1]["state"]],[1])
+
+
         self._expansion(1)
 
 
@@ -158,6 +159,7 @@ class Tree:
 
         # For now, just using q = mu**2
         temp=nx.get_node_attributes(self.graph, "mu")
+        temp.pop(1, None)
 
         top_n_nodes = sorted(temp, key=temp.get, reverse=True)[:self.comm_n]
         X = [self.graph.node[n]["state"] for n in top_n_nodes]
@@ -200,13 +202,16 @@ class Tree:
             self.id
         )
 
+        if len(options)==0:
+            return False
+
         # create empty nodes underneath the node being expanded
         for o in options:
 
             self.graph.add_node(len(self.graph)+1,
                     mu = 0,
                     N = 0,
-                    state=self.state_store(self.data, self.graph.node[start_node]["state"], o)
+                    state=self.state_store(self.data, self.graph.node[start_node]["state"], o, self.id)
                     )
 
             self.graph.add_edge(start_node, len(self.graph))
@@ -259,7 +264,7 @@ class Tree:
 #                sim_action = self.sim_selection_func(self.data, options, state)
 #
 #                # add that to the actions of the current robot
-#                temp_state = self.state_store(self.data, temp_state, sim_action)
+#                temp_state = self.state_store(self.data, temp_state, sim_action, self.id)
 #
 #            # calculate the reward at the end of simulation
 #            rew = self.reward(self.data, temp_state) \
@@ -282,8 +287,8 @@ class Tree:
             start_node = self._parent(start_node)
 
             self.graph.node[start_node]["mu"] = \
-                    (gamma * self.graph.node[start_node]["mu"] + avg_reward)\
-                    /(gamma * self.graph.node[start_node]["N"] + 1)
+                    (gamma * self.graph.node[start_node]["mu"]*self.graph.node[start_node]["N"] + avg_reward)\
+                    /(self.graph.node[start_node]["N"] + 1)
 
             self.graph.node[start_node]["N"] = \
                     gamma * self.graph.node[start_node]["N"] + 1