Authors: Kaede Utsunomiya; Yuichi Asahiro
Addresses: Department of Information Science, Kyushu Sangyo University, Japan ' Department of Information Science, Kyushu Sangyo University, Japan
Abstract: The (n2 - 1)-puzzle is a generalisation of the well-known 15-puzzle. Ratner and Warmuth (1990) showed that finding a shortest sequence of moves for the (n2 - 1)-puzzle is NP-hard. Many researches have been devoted to the (n2 - 1)-puzzle so far. Also, the (n2 - 1)-puzzle is chosen as a benchmark problem in many researches developing a new method/strategy. For the (n2 - 1)-puzzle, a real-time algorithm is proposed by Parberry (1995), which is based on a greedy method and completes the puzzle in at most 5n3 - 9n2 / 2 + O(n) moves and needs O(1) computation time per move, although there is no guarantee that the number of moves is optimal (shortest). Following the direction of the research by Parberry (1995), we present an algorithm, which is designed by modifying Parberry's algorithm and giving a tight analysis. The number of moves by the new algorithm is smaller, which needs at most 5n3 - 11n2 + O(n) moves, and computation time per move is also O(1).
Keywords: (n2 - 1)-puzzle; 15-puzzle; greedy algorithm; theoretical analysis.
International Journal of Innovative Computing and Applications, 2017 Vol.8 No.3, pp.133 - 148
Received: 05 Feb 2016
Accepted: 03 Mar 2016
Published online: 15 Sep 2017 *