Distributed identification of stable large-scale isomorphic nonlinear networks using partial observations

Distributed parameter identification in large-scale isomorphic nonlinear multi-agent networks encounters challenges due to inherent nonlinear dynamics and partial observations. Ensuring stability is crucial for stable parameter identification, especially under uncertainties in data and models. To address these challenges, this paper proposes a particle consensus-based expectation maximization (EM) algorithm. The E-step employs a distributed particle filtering approach to achieve global consensus state estimation, approximating the analytically intractable likelihood function arising from unknown dynamic interactions and multiple integrals. The M-step imposes prior contraction-stabilization constraints during likelihood function maximization to ensure stable parameter identification under data and model uncertainties. Performance analysis and simulation results confirm the effectiveness of the proposed method in accurately identifying parameters for nonlinear networks.