Comparison and Evaluation of Multi-Objective Genetic Algorithms for Military Planning and Scheduling Problems: Applied to Course of Action Planning


  1. Guitouni, A.
  2. Belfares, L.
Corporate Authors
Defence R&D Canada - Valcartier, Valcartier QUE (CAN)
Planning detailed military Courses of Action (COAs) is a very complex and time consuming. In this work, we investigate Evolutionary algorithms (EAs) to solve COAs’resource management and scheduling problems. The performance of such algorithms is assessed based on their outcomes quality. Efficient algorithms exhibit a good approximation of the Pareto optimal sets while requiring reasonable computational resources. A good approximation set is the resultant of a trade-off between diversity of solutions and their proximity to the true Pareto front. Such a balance is difficult to achieve with NP-hard problems exhibiting Pareto frontier discontinuity and multimodality although EAs are able to handle such optimization features. In this report, the multicriteria filtering genetic algorithm (MFGA) is proposed to achieve balanced proximitydiversity of the generated solutions. It uses a reproduction procedure based on a multicriteria filtering method and dominance concept to select solutions characterized by at least one bestscored objective or by all objectives achieving minimal threshold values. It is applied together with crossover and mutation operators. We illustrate this multi-objective EA in an enlarged sampling size scheme to solve highly constrained course of action planning problems. Cardinal and ordinal objectives are considered. An empirical comparison with three state-of-the-art multiobjective EAs is done using metrics of performance. The results show that this new approa

Il y a un résumé en français ici.

Report Number
DRDC-VALCARTIER-TR-2003-372 — Technical Report
Date of publication
01 Apr 2008
Number of Pages
Hardcopy;CD ROM

Permanent link

Document 1 of 1

Date modified: