Sokoban Solver
Forrest Knight
CS 480 - Artificial Intelligence - Fall 2017
Usage
Compile files
cd src
javac *.java
SokobanMain contains the main method. The program takes command-line arguments with the following options.
java SokobanMain [-option] [Sokoban input file]
Options
-b Breadth-first search
-d Depth-first search
-u Uniform-cost search (move = 1, push = 2)
-gb Greedy best-first search with number of boxes on goal heuristic
-gm Greedy best-first search with Manhattan distances heuristic
-ab AStar with number of boxes on goal heuristic
-am AStar with goals and boxes Manhattan distances heuristic
NO command-line input validation is done. It's a very simple command-line as I was focused more on the implementation of the searches rather than the prettiness of the UI.
Input
The Sokoban files must be in the following format.
[Number of columns]
[Number of rows]
[Rest of the puzzle]
With the following state mappings (note that when the player or box is on the goal, the mapping changes).
0 (hash) Wall (Obstacle)
S (period) Empty goal (Storage)
R (at) Player on floor (Robot)
\+ (plus) Player on goal
B (dollar) Box on floor (Block)
\* (asterisk) Box on goal
Output
The output is in the following format.
1. String representation of initial state
2. String representation of the final state
3. Move solution
4. Number of nodes explored
5. Number of previously seen nodes
6. Number of nodes at the fringe
7. Number of explored nodes
8. Time elapsed in milliseconds