Probabilistic Message-Passing Control

There is insufficient current understanding of how to apply fully decentralized control networks sparsely coupled nonlinear dynamical subsystems subject noise track a desired state. As exemplars, this class problem motivated by practical requirements creating power grids robust cascade failures, the digital transformation Industry 4.0 managing IoT connectivity reliably, and controlling transport flow in smart cities computing at edge. We demonstrate that an approach utilizing probability theory characterize exploit uncertainty locally received information, optimized messages passed between neighboring sufficient implicitly infer global knowledge. Thus, state could be realized through signals applied only local using information without any reference Given system can decomposed into set subsystems, we develop theoretical method probabilistic message passing all interacting subsystem level, but which promotes system-wide convergence Our results are corroborated computational experiments on network 10-node partially four separated with inputs determined level. Comparing centralized from nodes achieve validates our hypothesis affected mechanism needing access information. also provide numerical increasing size showing algorithm independent size.