COS 226 Programming Assignment Checklist: 8 Puzzle

Frequently Asked Questions

Can I use different class names, method names, or method signatures from those prescribed in the API? No, as usual, you will receive a serious deduction for violating the API.

Can I assume that the puzzle inputs (arguments to the Board constructor and input to Solver) are valid? Yes, though it never hurts to include some basic error checking.

Do I have to implement my own stack, queue, and priority queue? No, you can use the versions in algs4.jar.

How do I return an Iterable<Board>? Add the items you want to a Stack<Board> or Queue<Board> and return that.

How do I implement equals()? Java has some arcane rules for implementing equals(), discussed on p. 103 of Algorithms, 4th edition. Note that the argument to equals() is required to be Object. You can also inspect Date.java or Transaction.java for online examples.

Must I implement the toString() method for Board exactly as specified? Yes. Be sure to include the board dimension and use 0 for the blank square. Use String.format() to format strings—it works like StdOut.printf(), but returns the string instead of printing it to standard output. For reference, our implementation is below, but yours may vary depending on your choice of instance variables.

public String toString() {
    StringBuilder s = new StringBuilder();
    s.append(N + "\n");
    for (int i = 0; i < N; i++) {
        for (int j = 0; j < N; j++) {
            s.append(String.format("%2d ", tiles[i][j]));
        }
        s.append("\n");
    }
    return s.toString();
}

Should the hamming() and manhattan() methods in Board return the Hamming and Manhattan priority functions? No, you should return the sum of the Hamming and Manhattan distances from the blocks to their goal positions. Also, recall that the blank square is not considered a block. You will compute the priority function in Solver by calling hamming() or manhattan() and adding to it the number of moves.

I'm a bit confused about the purpose of the twin() method. You will use it to determine whether a puzzle is solvable: exactly one of a board and its twin are solvable. A twin is obatined by swapping two adjacent blocks (the blank does not count) in the same row. For example, here is a board and its 5 possible twins. Your solver will use only one twin.

    1  3       3  1       1  3       1  3       1  3       1  3
 4  2  5    4  2  5    2  4  5    4  5  2    4  2  5    4  2  5
 7  8  6    7  8  6    7  8  6    7  8  6    8  7  6    7  6  8

  board      twin       twin       twin       twin       twin

How do I reconstruct the solution once I've dequeued the goal search node? Since each search node records the previous search node to get there, you can chase the pointers all the way back to the initial search node (and consider them in reverse order).

Can I terminate the search as soon as a goal search node is enqueued (instead of dequeued)? No, even though it does lead to a correct solution for the slider puzzle problem using the Hamming and Manhattan priority functions, it's not techincally the A* algorithm (and will not find the correct solution for other problems and other priority functions).

I noticed that the priorities of the search nodes dequeued from the priority queue never decrease. Is this a property of the A* algorithm? Yes. In the language of the A* algorithm, the Hamming and Manhattan distances (before adding in the number of moves so far) are known as heuristics. If a heuristic is both admissible (never overestimates the number of moves to the goal search node) and consistent (satisfies a certain triangle inequality), then this property is guaranteed. The Hamming and Manhattan heuristics are both admissible and consistent. You may use this property as a debugging clue: if it is ever violated, then you know you did something wrong.

I run out of memory when running some of the large sample puzzles. What should I do? Be sure to ask Java for additional memory, e.g., java -Xmx1600m Solver < puzzle36.txt. If your program is unable to solve certain instances, document that in your readme.txt file. You should expect to run out of memory when using the Hamming priority function. Be sure not to put the JVM option in the wrong spot or it will be treated as a command-line argument, e.g., java Solver -Xmx1600m < puzzle36.txt.

My program can't solve puzzle4x4-hard1.txt or puzzle4x4-hard2.txt, even if I give it a huge amount of space. What am I doing wrong? Probably nothing. The A* algorithm will struggle to solve even some 4-by-4 instances.

Testing

Input files.   The ftp directory contains many sample puzzles. The shortest solution to puzzle4x4-hard1.txt and puzzle4x4-hard2.txt are 38 and 47, respectively. The shortest solution to puzzle[T].txt requires exactly T moves. Warning: puzzle36.txt is especially difficult.

Testing. A good way to automatically run your program on our sample puzzles is to use the client Checker.java.

Sample executions.

Possible Progress Steps

These are purely suggestions for how you might make progress. You do not have to follow these steps.

Enrichment

Is there an efficient way to solve the 8-puzzle and its generalizations? Finding a shortest solution to a slider puzzle is NP-hard, so it's unlikely that an efficient solution exists.

What if I'm satisfied with any solution and don't need one that uses the fewest number of moves? Yes, change the priority function to put more weight on the Manhattan distance, e.g., 100 times the Manhattan distance plus the number of moves made already. This paper describes an algorithm that guarantees to perform at most N^3 moves.

Are ther better ways to solve 8- and 15-puzzle instances using the minimum number of moves? Yes, there are a number of approaches.

Can a puzzle have more than one shortest solution? Yes. See puzzle07.txt.

 Solution 1
 ------------------------------------------------------------------------------------
 1  2  3    1  2  3    1  2  3    1  2  3    1  2  3    1  2  3    1  2  3    1  2  3 
    7  6    7     6    7  4  6    7  4  6       4  6    4     6    4  5  6    4  5  6 
 5  4  8    5  4  8    5     8       5  8    7  5  8    7  5  8    7     8    7  8   

 Solution 2
 ------------------------------------------------------------------------------------
 1  2  3    1  2  3    1  2  3    1  2  3    1  2  3    1  2  3    1  2  3    1  2  3
    7  6    5  7  6    5  7  6    5     6       5  6    4  5  6    4  5  6    4  5  6 
 5  4  8       4  8    4     8    4  7  8    4  7  8       7  8    7     8    7  8