This paper introduces a hierarchical, decentralized,udand parallelizable method for dealing with optimizationudproblems with many agents. It is theoretically based on a hierarchicaludoptimization theorem that establishes the equivalenceudof two forms of the problem, and this idea is implemented usingudDMOC (Discrete Mechanics and Optimal Control). The resultudis a method that is scalable to certain optimization problemsudfor large numbers of agents, whereas the usual “monolithic”udapproach can only deal with systems with a rather smalludnumber of degrees of freedom. The method is illustrated withudthe example of deployment of spacecraft, motivated by theudDarwin (ESA) and Terrestrial Planet Finder (NASA) missions.
展开▼