Distributed State Estimation Over Sensor Networks With Multi‐Level Quantization

ABSTRACT This paper presents a distributed estimation framework for nonlinear systems over bandwidth‐limited sensor networks. Each node encodes its measurement innovation via multi‐level quantization (MLQ) and transmits a finite‐bit codeword, while the fusion center performs a quantization‐aware approximate minimum mean‐square error (MMSE) update and combines local posteriors using a cross‐covariance–free convex rule. We analyze the asymptotic behavior of quantization terms and derive a sufficient condition ensuring uniform boundedness of the fused error covariance, explicitly linking stability to quantizer resolution and system dynamics. The proposed method achieves low communication cost, local complexity, and scalability with the number of nodes. Simulation studies on maneuvering‐target tracking show consistent improvements over EKF, CI, and CKF across RMSE, IAE, ISE, ITAE, and ITSE metrics under matched bit budgets, and demonstrate robustness to noise and packet loss. The approach is well suited to real‐time control‐oriented applications such as UAV coordination, autonomous vehicles, and smart‐grid monitoring.