This paper investigates a novel multi-objective optimization framework for the multi-stage missile target allocation (M-MTA) problem, which also widely exists in other real-world complex systems. Specifically, a constrained model of M-MTA is built with the trade-off between minimizing the survivability of targets and minimizing the cost consumption of missiles. Moreover, a multi-objective optimization algorithm (NSGA-MTA) is proposed for M-MTA, where the hybrid encoding mechanism establishes the expression of the model and algorithm. Furthermore, rule-based initialization is developed to enhance the quality and searchability of feasible solutions. An efficient non-dominated sorting method is introduced into the framework as an effective search strategy. Besides, the genetic operators with the greedy mechanism and random repair strategy are involved in handling the constraints with maintaining diversity. The results of numerical experiments demonstrate that NSGA-MTA performs better in diversity and convergence than the excellent current algorithms in metrics and Pareto front obtained in 15 scenarios. Taguchi method is also adopted to verify the contribution of proposed strategies, and the results show that these strategies are practical and promotive to performance improvement.
Keywords: multi-objective optimization; non-dominated sorting; random repair strategy; rule-based initialization; weapon-target assignment.