Session:

MSA - Military and Security Applications of Evolutionary Computation

Title:

Application of graph-bound genetic operations to the generation of master air attack plans

Authors:

Darrin Taylor
Paula de Witte
Sherry E. Marcus
Rose Davis

Abstract:

Constructing a Master Air Attack plan requires the negotiations of many constraints. Typical of these constraints are limitations on the munitions that can be delivered to each target, limitations on the platforms that can carry the munitions, as well as limitations on the availability of the platforms from a specified unit. A schedule needs to be constructed that respects the unit contracts, the planner enforced attack waves, and the packaging preferences. In the process, we seek to maximize the targets that can be attained with available resources subject to the priorities of the targets. The initialization of the genetic algorithm is accomplished by randomized greedy operations that quickly find local minima in the solution space. These greedy algorithms are implemented as genetic operators and are used not only to initialize the chromosomes with valid solutions but also to “finish off” the good solutions that result from the genetic algorithm to ensure that any remaining resources are used and that the solution can only be improved by de-allocating some subset of the existing solution.