Design of experiment for tuning parameters of an ant colony optimization method for the constrained shortest Hamiltonian path problem in the grid networks

Abstract

<p style='text-indent:20px;'>In a grid network, the nodes could be traversed either horizontally or vertically. The constrained shortest Hamiltonian path goes over the nodes between a source node and a destination node, and it is constrained to traverse some nodes at least once while others could be traversed several times. There are various applications of the problem, especially in routing problems. It is an NP-complete problem, and the well-known Bellman-Held-Karp algorithm could solve the shortest Hamiltonian circuit problem within <inline-formula><tex-math id="M1">\begin{document}$ {\rm O(}{{\rm 2}}^{{\rm n}}{{\rm n}}^{{\rm 2}}{\rm )} $\end{document}</tex-math></inline-formula> time complexity; however, the shortest Hamiltonian path problem is more complicated. So, a metaheuristic algorithm based on ant colony optimization is applied to obtain the optimal solution. The proposed method applies the rooted shortest path tree structure since in the optimal solution the paths between the restricted nodes are the shortest paths. Then, the shortest path tree is obtained by at most <inline-formula><tex-math id="M2">\begin{document}$ {\rm O(}{{\rm n}}^{{\rm 3}}{\rm )} $\end{document}</tex-math></inline-formula> time complexity at any iteration and the ants begin to improve the solution and the optimal solution is constructed in a reasonable time. The algorithm is verified by some numerical examples and the ant colony parameters are tuned by design of experiment method, and the optimal setting for different size of networks are determined.