Changed dijkstra_algorithm a little. Added printout of the shortest path and weight in the a_star_algorithm. Added implementation of bidirectional algorithm. Still need good heuristic function for A* and bidirectional algorithms.

2021-11-02 15:17:18 +01:00 · 2021-11-02 15:17:18 +01:00 · 85995b3d16
commit 85995b3d16
parent 602b893f81
4 changed files with 184 additions and 4 deletions
--- a/algorithms/a_star.py
+++ b/algorithms/a_star.py
@ -3,23 +3,43 @@ def heuristic_cost(start, goal, graph):


 def a_star_algorithm(graph, point_set, start, goal, h=heuristic_cost):
+    def return_path_and_weight(c_f, c, s):
+        current_node = c_f[c]
+        shortest_path = [c, current_node]
+        while current_node != s:
+            current_node = c_f[current_node]
+            shortest_path.append(current_node)
+        weight = 0
+        shortest_path.reverse()
+        for k in range(len(shortest_path) - 1):
+            if shortest_path[k] > shortest_path[k+1]:
+                weight += graph[(shortest_path[k], shortest_path[k+1])]
+            else:
+                weight += graph[(shortest_path[k + 1], shortest_path[k])]
+        return shortest_path, weight
+
    for edge in graph.keys():
        point_set[edge[0]].append(edge[1])
        point_set[edge[1]].append(edge[0])
+
    open_set = set()
    open_set.add(start)
+
    came_from = {}
+
    g_score = {k: float('inf') for k in point_set.keys()}
    g_score[start] = 0
+
    f_score = {k: float('inf') for k in point_set.keys()}
    f_score[start] = h(start, goal, graph)
+
    while len(open_set) > 0:
        current = list(open_set)[0]
        for k in open_set:
            if f_score[k] < f_score[current]:
                current = k
        if current == goal:
-            return came_from, current
+            return return_path_and_weight(came_from, current, start)
        open_set.remove(current)
        for neighbor in point_set[current]:
            tentative_g_score = g_score[current]
--- a/algorithms/bidirectional.py
+++ b/algorithms/bidirectional.py
@ -0,0 +1,150 @@
+def heuristic_cost(start, goal, graph):
+    return 0
+
+
+def bidirectional_algorithm(graph, point_set, start, goal, h=heuristic_cost):
+    def return_path_and_weight_front(c_f, c, s):
+        current_node = c_f[c]
+        shortest_path = [c, current_node]
+        while current_node != s:
+            current_node = c_f[current_node]
+            shortest_path.append(current_node)
+        weight = 0
+        for k in range(len(shortest_path) - 1):
+            if shortest_path[k] > shortest_path[k+1]:
+                weight += graph[(shortest_path[k], shortest_path[k+1])]
+            else:
+                weight += graph[(shortest_path[k + 1], shortest_path[k])]
+        return shortest_path, weight
+
+    def return_path_and_weight_back(c_f, c, s):
+        current_node = c_f[c]
+        shortest_path = [c, current_node]
+        while current_node != s:
+            current_node = c_f[current_node]
+            shortest_path.append(current_node)
+        weight = 0
+        shortest_path.reverse()
+        for k in range(len(shortest_path) - 1):
+            if shortest_path[k] > shortest_path[k+1]:
+                weight += graph[(shortest_path[k], shortest_path[k+1])]
+            else:
+                weight += graph[(shortest_path[k + 1], shortest_path[k])]
+        return shortest_path, weight
+
+    def return_path_and_weight_front_meet_back(c_f_f, c_f_b, c_f, s, g):
+        shortest_path_front, weight_front = return_path_and_weight_back(c_f_f, c_f, s)
+        shortest_path_back, weight_back = return_path_and_weight_back(c_f_b, c_f, g)
+        shortest_path_back.reverse()
+        return shortest_path_front + shortest_path_back[1:], weight_front + weight_back
+
+    def return_path_and_weight_back_meet_front(c_f_f, c_f_b, c_b, s, g):
+        shortest_path_front, weight_front = return_path_and_weight_back(c_f_f, c_b, s)
+        shortest_path_back, weight_back = return_path_and_weight_back(c_f_b, c_b, g)
+        shortest_path_back.reverse()
+        return shortest_path_front + shortest_path_back[1:], weight_front + weight_back
+
+    for edge in graph.keys():
+        point_set[edge[0]].append(edge[1])
+        point_set[edge[1]].append(edge[0])
+    open_set_front = set()
+    open_set_front.add(start)
+
+    open_set_back = set()
+    open_set_back.add(goal)
+
+    came_from_front = {}
+    came_from_back = {}
+
+    g_score_front = {k: float('inf') for k in point_set.keys()}
+    g_score_front[start] = 0
+
+    g_score_back = {k: float('inf') for k in point_set.keys()}
+    g_score_back[goal] = 0
+
+    f_score_front = {k: float('inf') for k in point_set.keys()}
+    f_score_front[start] = h(start, goal, graph)
+
+    f_score_back = {k: float('inf') for k in point_set.keys()}
+    f_score_back[goal] = h(goal, start, graph)
+
+    while len(open_set_front) > 0 and len(open_set_back) > 0:
+        current_front = list(open_set_front)[0]
+        current_back = list(open_set_back)[0]
+
+        for k in open_set_front:
+            if f_score_front[k] < f_score_front[current_front]:
+                current_front = k
+
+        for k in open_set_back:
+            if f_score_back[k] < f_score_back[current_back]:
+                current_back = k
+
+        if current_front == goal:
+            return return_path_and_weight_front(came_from_front, current_front, start)
+
+        if current_back == start:
+            return return_path_and_weight_back(came_from_back, current_back, goal)
+
+        if current_front in came_from_back.keys():
+            return return_path_and_weight_front_meet_back(came_from_front, came_from_back, current_front, start, goal)
+
+        if current_back in came_from_front.keys():
+            return return_path_and_weight_back_meet_front(came_from_front, came_from_back, current_back, start, goal)
+
+        open_set_front.remove(current_front)
+        open_set_back.remove(current_back)
+
+        for neighbor in point_set[current_front]:
+            tentative_g_score = g_score_front[current_front]
+
+            if current_front > neighbor:
+                tentative_g_score += graph[(current_front, neighbor)]
+            else:
+                tentative_g_score += graph[(neighbor, current_front)]
+
+            if tentative_g_score < g_score_front[neighbor]:
+                came_from_front[neighbor] = current_front
+                g_score_front[neighbor] = tentative_g_score
+                f_score_front[neighbor] = g_score_front[neighbor] + h(neighbor, goal, graph)
+                if neighbor not in open_set_front:
+                    open_set_front.add(neighbor)
+
+        for neighbor in point_set[current_back]:
+            tentative_g_score = g_score_back[current_back]
+
+            if current_back > neighbor:
+                tentative_g_score += graph[(current_back, neighbor)]
+            else:
+                tentative_g_score += graph[(neighbor, current_back)]
+
+            if tentative_g_score < g_score_back[neighbor]:
+                came_from_back[neighbor] = current_back
+                g_score_back[neighbor] = tentative_g_score
+                f_score_back[neighbor] = g_score_back[neighbor] + h(neighbor, goal, graph)
+                if neighbor not in open_set_back:
+                    open_set_back.add(neighbor)
+
+
+if __name__ == "__main__":
+    g = {
+        (2, 1): 3,
+        (3, 2): 2,
+        (5, 3): 1,
+        (9, 5): 5,
+        (10, 9): 4,
+        (9, 4): 3,
+        (4, 1): 4,
+        (7, 1): 6,
+        (3, 1): 4,
+        (6, 2): 3,
+        (8, 6): 8,
+        (8, 3): 2
+    }
+
+    v = dict()
+    for i in g.keys():
+        v[i[0]] = []
+        v[i[1]] = []
+
+    print(bidirectional_algorithm(g, v, 7, 8, heuristic_cost))
--- a/algorithms/dijkstra.py
+++ b/algorithms/dijkstra.py
@ -26,8 +26,8 @@ def dijkstra_algorithm(graph, point_set, start, goal):
        prev[vertex] = None
        visited[vertex] = False

-    dist[start] = 0
    heap.heappush(queue, (0, start))
+    dist[start] = 0
    visited[start] = True

    while len(queue) > 0:
@ -41,6 +41,10 @@ def dijkstra_algorithm(graph, point_set, start, goal):

            if new_dist < dist[neighbor]:
                dist[neighbor] = new_dist
+                for k in range(len(queue)):
+                    if queue[k][1] == neighbor:
+                        queue[k] = (new_dist, neighbor)
+                        heap.heapify(queue)
                prev[neighbor] = current[1]

            if not visited[neighbor]:
@ -51,6 +55,7 @@ def dijkstra_algorithm(graph, point_set, start, goal):
    while prev[temp] is not None:
        shortest_path.append(prev[temp])
        temp = prev[temp]
+    shortest_path.reverse()
    return shortest_path, dist[goal]


@ -67,7 +72,8 @@ if __name__ == "__main__":
        (3, 1): 4,
        (6, 2): 3,
        (8, 6): 8,
-        (8, 3): 2
+        (8, 3): 2,
+        (9, 1): 8
    }

    v = dict()
--- a/main.py
+++ b/main.py
@ -1,5 +1,7 @@
 import tools.file_service as op_file
 import algorithms.dijkstra as dijkstra
+import algorithms.a_star as a_star
+import algorithms.bidirectional as bidirectional

 if __name__ == '__main__':
    g = op_file.read_graph_from_file("dataset/deezer_clean_data/HR_edgeswith_weight.csv", has_weight=True)
@ -8,4 +10,6 @@ if __name__ == '__main__':
        v[i[0]] = []
        v[i[1]] = []

-    print(dijkstra.dijkstra_algorithm(g, v, 0, 4))
+    # print(dijkstra.dijkstra_algorithm(g, v, 0, 4))
+    # print(a_star.a_star_algorithm(g, v, 0, 4))
+    # print(bidirectional.bidirectional_algorithm(g, v, 0, 4))