Genetic Algorithms for Solving Shortest Path Problem in Maze-Type Network with Precedence Constraints

Shortest path (SP) problem is a classical combinatorial optimization problem, which has various application domains such as communication network routing and location-based services under cloud environment. However, maze-type networks, sparse networks with many pairs of disconnected nodes, had rarely been studied. A maze-type network is more difficult to analyze than common dense network, since it has rare feasible paths. Moreover, precedence constraints among the nodes further increase the complexity of maze-type network, and this paper aims to develop genetic algorithms for finding the shortest path in maze-type network with precedence constraints. In order to address precedence constrained maze-type shortest path (PCM-SP) problem, the fitness switching genetic algorithm (FSWGA), which has been developed to solve the unconstrained maze-type shortest path (M-SP) problems, is revised by adopting position listing representation as encoding scheme and applying two enhanced decoding procedures. In addition, genetic operator of candidate order based genetic algorithm (COGA) is used to explore the search space effectively, and experiment results demonstrate that the enhanced FSWGA can solve PCM-SP problems more effectively than the previous FSWGA.