A* Algorithm does not find shortest path

Question

I am trying to implement the A* algorithm in python but have hit a problem when trying to find the path of this map:

X X X X X X X     S = Start
0 0 0 X 0 0 0     E = End
0 S 0 X 0 E 0     X = Wall
0 0 0 X 0 0 0
0 0 0 0 0 0 0

I am using the Manhattan method. My implementation does find a path, but not the shortest one. The error starts on it's second move -- after moving right. At this point it can move up and the heuristic cost would be four (three right, one down) or down (three right, one up). Is there a way so it chooses down to get the shortest path?

Code:

class Node:
    def __init__(self, (x, y), g, h, parent):
        self.x = x
        self.y = y
        self.g = g
        self.h = h
        self.f = g+h
        self.parent = parent

    def __eq__(self, other):
        if other != None:
            return self.x == other.x and self.y == other.y
        return False

    def __lt__(self, other):
        if other != None:
            return self.f < other.f
        return False

    def __gt__(self, other):
        if other != None:
            return self.f > other.f
        return True

    def __str__(self):
        return "(" + str(self.x) + "," + str(self.y) + ") " + str(self.f)

def find_path(start, end, open_list, closed_list, map, no_diag=True, i=1):
    closed_list.append(start)
    if start == end or start == None:
         return closed_list
     new_open_list = []
     for x, y in [(-1,1),(-1,-1),(1,-1),(1,1),(0,-1),(0,1),(-1,0),(1,0)]:
        full_x = start.x + x
        full_y = start.y + y
        g = 0
        if x != 0 and y != 0:
            if no_diag:
                continue
            g = 14
        else:
            g = 10
        h = 10 * (abs(full_x - end.x) + abs(full_y - end.y))
        n = Node((full_x,full_y),g,h,start)
        if 0 <= full_y < len(map) and 0 <= full_x < len(map[0]) and map[full_y][full_x] != 1 and n not in closed_list:
            if n in open_list:
                if open_list[open_list.index(n)].g > n.g:
                    new_open_list.append(n)
                else:
                    new_open_list.append(open_list[open_list.index(n)])
            else:
                new_open_list.append(n)
    if new_open_list == None or len(new_open_list) == 0:
        return find_path(start.parent, end, open_list, closed_list, map, no_diag, i-1)
    new_open_list.sort()
    return find_path(new_open_list[0], end, new_open_list, closed_list, map, no_diag)

Answer 1

You seem to be constructing a new open list for each node, which contains only that node's neighbors. This essentially makes your search a form of a depth-first search, while A* should be a best-first search.

You need to use one open list which will be updated with each node's neighbors as you visit that node. The old nodes in the open list must remain there until they are traversed and moved to the closed list.

Regarding what you said in your question, It's OK for the search to try moving up before down (since according to the heuristic, they are the same distance from the goal). What matters is that in the end, the path chosen will be the shortest.