codigo actualizado

Documentos/TFG_Machine_Learning/Reinforce_Learning.py

@@ -7,6 +7,7 @@ Created on Tue Feb  6 19:02:32 2024
 """
 
 import numpy as np
+import math
 import plotly.graph_objects as go
 from tqdm.notebook import tqdm
 import plotly.express as px
@@ -116,8 +117,20 @@ class Swimmer(Agent):
         self.ni = ni
         self.sigma = sigma
 
-        # local position within the periodic box. X = [x, z]^T with 0 <= x < 2 pi and 0 <= z < 2 pi
+        #obstáculos
-        self.X = np.array([np.random.uniform(0, L), np.random.uniform(0, L), 0])
+        self.obstacles= self.generate_obstacles()
+       
+        #Condición inicial para el swimmer
+        valid_initial_position = False
+        while not valid_initial_position:
+            self.X = np.array([np.random.uniform(0, L), np.random.uniform(0, L), 0])
+            valid_initial_position = True
+            # Comprobamos si esta inicialmente esta dentro de un obstáculo
+            for i in range(len(self.obstacles)//2):
+                obstacle_position = np.array([self.obstacles[2*i], self.obstacles[2*i+1], 0])
+                if np.linalg.norm(self.X - obstacle_position) < 0.5*self.sigma:
+                    valid_initial_position = False
+                    break
          
         # absolute position. -inf. <= x_total < inf. and -inf. <= z_total < inf.
         self.X_total = self.X
@@ -148,8 +161,7 @@ class Swimmer(Agent):
         # update coarse-grained state
         self.update_state()
 
-        #obstáculos
+     
-        self.obstacles= self.generate_obstacles()
         
         
         #distancia entre el swimmer y el obstáculo
@@ -394,6 +406,21 @@ class Swimmer(Agent):
             self.W[2] -= 1./8*Wc 
         return action_index
 
+    def periodic_boundaries(self, isxperiodic=True, isyperiodic=True, iszperiodic=True):
+   
+        offset = [math.floor(-self.X[0] * 1/L + 0.5),
+                  math.floor(-self.X[1] * 1/L + 0.5),
+                  0]
+    
+        if isxperiodic:
+            self.X[0] += offset[0] * L
+        if isyperiodic:
+            self.X[1] += offset[1] * L
+        if iszperiodic:
+            self.X[2] += offset[2] * L
+    
+        return self.X
+
  
 #Define Taylor-Green vortex
 
@@ -594,17 +621,22 @@ print(Q)
 Ns = 5000
 spinner = Swimmer(Ns, 1, 1)
 traj = []
+
 obstacles = spinner.generate_obstacles()
 for i in range(Ns):
   spinner.interaction_with_obstacles(obstacles, 2.5, 1, 1., 2.5e-4, 10000, 0.001)
   traj.append(spinner.X[0])
   traj.append(spinner.X[1])
   
+  spinner.periodic_boundaries()
+
   action_index = spinner.take_greedy_action(Q)
   
 
   spinner.update_state()
   
+ 
+  
   #print("Mi estado", spinner.my_state)
   #print("Valor de Wz después de tomar la acción:", spinner.W[2])
 
@@ -616,5 +648,4 @@ print(obstacles[::2])
 plt.show()
 
 
-#comprobación de que W va cambiando