Multi-level Evolutionary Algorithms resource allocation utilizing model-based systems engineering.

摘要

This research presents an innovative approach to solve the resource allocation problems using Multi-level Evolutionary Algorithms. Evolutionary Algorithms are used to solve resource allocation problems in different domains and their results are then incorporated into a higher level system solution using another Evolutionary Algorithm to solve base camp planning problems currently faced by the U.S. Department of Defense.;Two models are introduced to solve two domain specific models: a logistics model and a power model. The logistic model evaluates routes for logistics vehicles on a daily basis with a goal of reducing fuel usage by delivery trucks. The evaluation includes distance traveled and other constraints such as available resource levels and priority of refilling. The Power model incorporates an open source electrical distribution simulator to evaluate the placement of structures and generators on a map to reduce fuel usage.;These models are used as the fitness function for two separate Evolutionary Algorithms to find solutions that reduce fuel consumption within the individual domains. A multi-level Evolutionary Algorithm is then presented, where the two Evolutionary Algorithms share information with a higher level Evolutionary Algorithm that combines the results to account for problem complexity from the interfacing of these systems. The results of using these methods on 5 different base camp sizes show that the techniques provide a considerable reduction of fuel consumption. While the Evolutionary Algorithms show significant improvement over the current methods, the multi-level Evolutionary Algorithm shows better performance than using individual Evolutionary Algorithms, with the results showing a 19.25% decrease in fuel consumption using the multi-level Evolutionary Algorithm.