Title:   Decentralized Control Using Global Optimization (DCGO) (Preprint)

Conference paper

Flint, Matthew; Khovanova, Tanya; Curry, Michael

Report Date: MAR 2007

Abstract : The coordination of a team of distributed air vehicles requires a complex optimization, balancing limited communication bandwidths, non-instantaneous planning times and network delays, while at the same time trying to allocate limited resources to spatially diverse locations in a near-optimal fashion in a dynamic and uncertain environment. Given that, in this environment, the optimality of a given plan will not last very long when the information state is constantly changing and being updated, a new approach is proposed in this paper. Global-scope plans for the team are generated and distributed using the principle of emergent leadership to provide efficient plan generation and execution with minimal performance degradation compared to a centralized controller under delayed communications. This type of protocol is labeled the Decentralized Control Global Optimization (DCGO) protocol, and is discussed in this paper, along with some simulation results showing that this premise can produce good results in a realistic environment.