Implementing a tree for a maze to use in DFS, BFS

Question

My program is taking a char array as input from a file. The array looks like this:

"#########",
"# #     #",
"# ##  # #",
"#     # #",
"### # ###",
"#   # # #",
"# # #####",
"#   #   #",
"#########",

I am implementing DFS and BFS to solve this maze with starting from [1,1] and ending in [width - 1,height - 1].

I thought of making a tree representing the maze and then traversing the tree using each algorithm respectively.

I will start at each row and scan for empty cells, at each empty cell every cell that is to its right, left and bottom are going to be children of that cell. It looks something like this:

    for (int i = 0; i < width; i++)
    {
        for (int j = 0; j < height; j++)
        {
            if (isEmpty(maze[i][j]))
                    {
                         putChildren(maze[i-1][j], maze[i][j+1], maze[i+1][j]);
                         //this will check if it's a wall first
                    }       
    }

Is it a viable tactic to implement the tree like this and then use it to traverse the tree with DFS and BFS or I should I go at this another way?

score 3 · Answer 1 · answered Nov 06 '13 at 20:22

3

Nice project, i love things like that. By the way Did you considered directional try algorithm (so called A* algorithm) I think that it will be better for you, especially while working on 2D array. It have better performance in usual cases than other methods, and you don't need to use linked cells. There are also some kind of improvements for this algorithm including memory linked with "try direction first" method. Of course there is no problem with your's method, but consider case when you have really gigantic matrix to process.

answered Nov 06 '13 at 20:22

esavier

423
4
13

I actually plan on implementing A* later but first I want to implement DFS and BFS. I pretty much want to implement all of them to exercise my mind and get familiar with them and in general with artificial intelligence. – Powerbyte Nov 06 '13 at 23:18
I am working a lot in field of AI, and if i can give u a tip, there is no good way of solving problem. What you are need to do is prepare to face possibility that you will need both algorithms to implement and choose better for situation (after problem analysing). Still, to make maze solver best possible, you should allow your implememntation "sensing" surroundings. Also consider making 2D (with "visited" bool flag) array instead of graph. In some cases its easier, in some harder but its worth implementing over usual graph) – esavier Nov 07 '13 at 11:31
next one, even if u are using graph, use visited flag, if you will ind dead end, you can erase them back. Ue greedy algorithm, that will find u the fastest solution and return it. There is also a nice algorithm ( but had to implement) called flood fill room marking. You should assume that rooms have passages between them, and mark only those passages in tree, and also both "start" and "end", flood fill will consider only passable ways, and should discard any paces like open room or dead ends. – esavier Nov 07 '13 at 11:39

score 1 · Answer 2 · answered Nov 24 '13 at 12:35

Your idea is good and pretty straightforward, but I think you misunderstood tree with a graph.

First of all, there is no need to create a tree from a maze graph - you can run BFS, DFS, A* , ... on a general graph.

Moreover, not every maze could be presented as a tree.

Let's look at the example:

"#####",
"#   #",
"# # #",
"#   #",
"#####",

Clearly there is a cycle in the maze, so it cannot be presented as a tree. Your example also has a few cycles.

Implementing a tree for a maze to use in DFS, BFS

2 Answers2