作者: Frederik Deroo , Martin Meinel , Michael Ulbrich , Sandra Hirche
DOI: 10.3182/20140824-6-ZA-1003.00917
关键词: Stability (learning theory) 、 Matrix (mathematics) 、 Gradient descent 、 Control theory 、 Optimal control 、 Term (time) 、 Mathematical optimization 、 Computer science 、 Distributed algorithm 、 Model predictive control
摘要: Abstract Most results on distributed control design of large-scale interconnected systems assume a central designer with global model knowledge. The wish for privacy subsystem data raises the desire to find methods determine an optimal law without centralized knowledge, i.e. in fashion. In this paper we present method guaranteed stability minimize infinite horizon LQ cost functional. introduction adjoint states allows iteratively optimize feedback matrix using gradient descent way, based finite formulation. Inspired by ideas stabilizing predictive control, terminal term is used, which gives bound functional and ensures stability. A presented that are validated numerical experiments.