A multi-agent based approach to modeling and solving dynamic generalized travelling salesman problem

This paper introduces four different types of Generalized Travelling Salesman Problem (GTSP) which are actually dynamic variants of the well-known logistics problems. For all of these defined types, new cities are added to/deleted from the city domain during the travelling of the salesman. This city addition and deletion during the solution phase of the problem, differentiates the proposed types from the classical GTSP. Since these variants of GTSP are relatively complicated compared to classical forms, an agent-based strategy is proposed in this paper to handle complexity and dynamism. In this respect, proposed agent-based strategy employs a general manager and numerous region agents to control and coordinate the dynamism in their regions and in the central level. Region agents create solutions just for their regions and thereby complexity of obtaining a central solution for each change containing is avoided. Findings of the proposed agent-based strategy confirm that adaptation ability of agent-based strategy against the dynamism is significantly better than classical central solution approach. In this respect, this paper is expected to be novel in two respects. First, those four types of GTSP defined in this paper, are different from the classical GTSP since they have dynamic city domain. Second, the proposed novel agent-based solution strategy is capable to create solutions in a timely manner.