Dynamic window search of ant colony optimization for complex multi-stage decision problems

It is difficult to solve complex multi-stage decision problems with strong non-linearity, high dimensional or complex constraints. In contrast to the limitation of dynamic programming techniques and genetic algorithms, ant colony optimization algorithms represent complex constraints naturally and use the local heuristic information to guide search efficiently. In this paper, a dynamic window ant colony optimization algorithm is proposed for the large-scale complex multi-stage decision problems. In the algorithm a subset of the feasible decision set at each stage is selected by real-code genetic optimization and mapped to the nodes in one layer of a layered construction graph. Ants find routes through the layered construction graph, and each route corresponds to a solution candidate. Computation complexity analysis and simulation results demonstrate that, in comparison with basic ant colony optimization algorithms and genetic algorithms, the proposed algorithm greatly improves the computational efficiency.