Flocking in Distributed Control and Optimization.

摘要

In recent years, the study of biological systems in which flocking and/or swarming phenomena is observed has received increased attention by researchers in control and optimization. In these systems, a great deal of coordination is achieved under relatively minimal communication requirements. In this research project we developed and formally analyzed techniques for control and global optimization that are inspired by flocking and other forms coordinated behavior from biological systems. Most notably we have developed a new approach for simulation-based optimization that makes use of a flocking potential. We show that the flocking-based approach has an important noise reduction property. Under a flocking discipline, simulation noise realizations that induce trajectories differing too much from the group average are likely to be discarded because of each individual's need to maintain cohesion. As result the group trajectories are less affected by such noise.