부산대학교 도서관

This work develops a federated learning-based distributed model predictive control (FL-DMPC) method for nonlinear systems with multiple subsystems to address the privacy-preserving issue of data transmission among subsystems and heterogeneity issue due to non-independent and identically distributed data among subsystems. Specifically, a novel FL framework is proposed to aggregate submodels into a global FL model with a sufficiently small modeling error with provable convergence properties derived based on iteration theory. Subsequently, by incorporating the FL model into a DMPC scheme, an FL-DMPC method is presented to achieve the expected performance of nonlinear systems. Finally, a chemical process network is adopted to demonstrate the effectiveness of the proposed FL-DMPC method.