mode

PSO_Lib.py

@@ -31,15 +31,24 @@ def update_Position_centroid_based(p, c1, c2, c3, global_attractor):
     p.position = add_pos_vel(add_pos_vel(d_glob, v_tmp), v_rand)
 
 
-def cost_function(nodes, permutation):
+def cost_function(nodes, permutation, mode = "GEO"):
     cost = 0
-    for i in range(0, len(permutation) - 1):
-        lat1 = nodes['Breitengrad_DEC'][nodes.index[permutation[i]]]
-        lon1 = nodes['Laengengrad_DEC'][nodes.index[permutation[i]]]
-        lat2 = nodes['Breitengrad_DEC'][nodes.index[permutation[i + 1]]]
-        lon2 = nodes['Laengengrad_DEC'][nodes.index[permutation[i + 1]]]
-        cost += math.ceil(geo_distance(lat1, lon1, lat2, lon2))
-    return cost
+    if(mode=="GEO"):
+        for i in range(0, len(permutation) - 1):
+            lat1 = nodes['Breitengrad_DEC'][nodes.index[permutation[i]]]
+            lon1 = nodes['Laengengrad_DEC'][nodes.index[permutation[i]]]
+            lat2 = nodes['Breitengrad_DEC'][nodes.index[permutation[i + 1]]]
+            lon2 = nodes['Laengengrad_DEC'][nodes.index[permutation[i + 1]]]
+            cost += math.ceil(geo_distance(lat1, lon1, lat2, lon2))
+        return cost
+    elif(mode=="EUC_2D"):
+        for i in range(0, len(permutation) - 1):
+            lat1 = nodes['Breitengrad_DEC'][nodes.index[permutation[i]]]
+            lon1 = nodes['Laengengrad_DEC'][nodes.index[permutation[i]]]
+            lat2 = nodes['Breitengrad_DEC'][nodes.index[permutation[i + 1]]]
+            lon2 = nodes['Laengengrad_DEC'][nodes.index[permutation[i + 1]]]
+            cost += math.ceil(euc_distance(lat1, lon1, lat2, lon2))
+        return cost
 
 
 def geo_distance(lat1, lon1, lat2, lon2):
@@ -65,7 +74,7 @@ def geo_distance(lat1, lon1, lat2, lon2):
 
     return (int)(RRR * math.acos(0.5 * ((1.0 + q1) * q2 - (1.0 - q1) * q3)) + 1.0)
 
-def eucl_distance(x1, y1, x2, y2):
+def euc_distance(x1, y1, x2, y2):
     xd = x1 - x2;
     yd = y1 - y2;
     dij = nint(math.sqrt(xd * xd + yd * yd))
@@ -74,15 +83,20 @@ def eucl_distance(x1, y1, x2, y2):
 
 def pre_processing_tsplib(json_file=""):
     ### Takes a filename of a .json file in tsplib and returns a dataframe
+    mode = ""
     nodes = pd.read_json("json/tsplib/" + json_file + ".json")
     solution_tsplib = pd.read_json("json/tsplib/" + json_file + "_tour.json")
+    if(nodes["edge_weight_type"][0]=="GEO"):
+        mode = "GEO"
+    elif(nodes["edge_weight_type"][0]=="EUC_2D"):
+        mode = "EUC_2D"
     known_best_cost = solution_tsplib['tourlength'][0]
     best_route = solution_tsplib['tour']
     nodes.drop(nodes.columns.difference(['node_coordinates']), 1, inplace=True)
     nodes_cleaned = pd.DataFrame(nodes['node_coordinates'].to_list(), columns=['Breitengrad_DEC',
                                                                                'Laengengrad_DEC'])
 
-    return nodes_cleaned, known_best_cost, best_route
+    return nodes_cleaned, known_best_cost, best_route, mode
 
 
 def two_opt(df, route):

Particle.py

@@ -5,10 +5,11 @@ import random
 
 class Particle():
 
-    def __init__(self, nodes):
+    def __init__(self, nodes, mode):
         n = len(nodes)                                              # Länge der Rundreise
+        self.mode = mode
         self.position = get_random_Position(n)                      # Initiale Position eines Partikels
-        self.cost = pl.cost_function(nodes, self.position)          # Initiale Kostenberechnung
+        self.cost = pl.cost_function(nodes, self.position, self.mode)          # Initiale Kostenberechnung
         self.local_attractor = self.position                        # Initialer lokaler Attraktor == initiale Position
         self.local_attractor_cost = self.cost                       # Initialer lokaler Attraktor Kosten == initiale Position Kosten
         self.velocity = list()                                      # leere Transpositionsliste
@@ -17,7 +18,7 @@ class Particle():
         self.position = add_pos_vel(self.position, self.velocity)
 
     def update_cost(self, nodes):                                   # Kostenfunktion neu berechnen
-        self.cost = pl.cost_function(nodes, self.position)
+        self.cost = pl.cost_function(nodes, self.position, self.mode)
 
     def update_velocity(self, c1, c2, c3, globalAttractor):         # Update Methode für velocity
         self.velocity = add_vel(

Population.py

@@ -2,9 +2,9 @@ import Particle
 
 
 class Population():
-    def __init__(self, num_pop, nodes):
+    def __init__(self, num_pop, nodes, mode):
         self.n = len(nodes)
-        self.population = list(Particle.Particle(nodes) for i in range(num_pop))    # Erstelle Population der Größe num_pop
+        self.population = list(Particle.Particle(nodes, mode) for i in range(num_pop))    # Erstelle Population der Größe num_pop
         self.best_solution = self.population[0].position                            # initiale Beste Lösung...
         self.best_cost = self.population[0].cost                                    # ...und deren Kosten
         self.check_for_best_solution()                                              # berechnet die tatsächlich beste Lösung der Startpopulation

Pso.py

@@ -5,8 +5,8 @@ import numpy as np
 import random
 
 
-def pso_two_opt(num_pop, nodes, c1=1/3, c2=1/3, c3=6/10, abort=50):
-    population = pop.Population(num_pop, nodes)                                             # Population erstellen
+def pso_two_opt(num_pop, nodes, c1=1/3, c2=1/3, c3=6/10, abort=50, mode='GEO'):
+    population = pop.Population(num_pop, nodes, mode)                                             # Population erstellen
     not_changed_count = 0
     while (not_changed_count < abort):
         for p in population.population:
@@ -23,8 +23,8 @@ def pso_two_opt(num_pop, nodes, c1=1/3, c2=1/3, c3=6/10, abort=50):
     return(twoOpt, twoOptCost)
 
 
-def pso(num_pop, nodes, c1=1/3, c2=1/3, c3=6/10, abort=50):
-    population = pop.Population(num_pop, nodes)
+def pso(num_pop, nodes, c1=1/3, c2=1/3, c3=6/10, abort=50, mode='GEO'):
+    population = pop.Population(num_pop, nodes, mode)
     not_changed_count = 0
     while (not_changed_count < abort):
         #print(not_changed_count)
@@ -40,8 +40,8 @@ def pso(num_pop, nodes, c1=1/3, c2=1/3, c3=6/10, abort=50):
         #print(population.best_cost)
     return population.best_solution, population.best_cost
 
-def var_pso(num_pop, nodes, c1=1 / 3, c2=1 / 3, c3=1 / 3, abort=20):
-    population = pop.Population(num_pop, nodes)
+def var_pso(num_pop, nodes, c1=1 / 3, c2=1 / 3, c3=1 / 3, abort=20, mode='GEO'):
+    population = pop.Population(num_pop, nodes,mode)
     not_changed_count = 0
 
     while (c1 > 0.05):

main.py

@@ -10,7 +10,7 @@ import PSO_Lib as pslib
 import numpy as np
 import time
 
-df, opt, opt_route = pslib.pre_processing_tsplib("burma14")
+df, opt, opt_route, mode = pslib.pre_processing_tsplib("burma14")
 population = 20
 abbruch = 15
 
@@ -24,7 +24,7 @@ laufzeit = np.array([])
 for i in range(0, 1):
     start = time.time()
 
-    route,kosten = PSO.pso(num_pop=population, nodes=df, c1=9 / 15, c2=3 / 15, c3=3 / 15, abort=abbruch)
+    route,kosten = PSO.pso(num_pop=population, nodes=df, c1=9 / 15, c2=3 / 15, c3=3 / 15, abort=abbruch, mode=mode)
 
     # perm = pslib.get_random_Position(len(df))
     # res = pslib.cost_function(df, pslib.two_opt_with_dist_matrix(distmatrix,perm))