Multi-objective genetic algorithms for courses of action planning

Planning military courses of action is a very complex and difficult activity. Planners should take into consideration environmental information, predictions, the end state targeted and resource constraints. Development of courses of action involves solving simultaneously planning and scheduling problems. In this work, a new approach based on genetic algorithms (GA) and multi-objective optimisation is proposed to support resource-constrained courses of action development where both cardinal and ordinal objectives are considered. A vector of fitness evaluations is proposed to control the proportion of the infeasible solutions. Crossover and mutation operators are designed to diversify the search space and improve solutions on all objectives from one generation to another. In the replacement strategy, a selection procedure, based on the dominance concept and a multi-criteria filtering method, is proposed. Such a strategy is applied when the population reaches a critical size. Different GA schemes are compared and their strengths and weaknesses are discussed. The multi-criteria filtering procedure used in the replacement strategy proved very efficient in the diversification of the Pareto front.