Scalable Distributed Estimation for Time-Varying System With Generalized Weighted Try-Once-Discard and Round-Robin Protocols

• Published in: IEEE sensors journal Vol. 25; no. 16; pp. 31653 - 31663
• Main Authors: Song, Yanhua, Zhang, Jinnan, Han, Fei, Shen, Yuxuan, Dong, Hongli, Hu, Zhongrui
• Format: Journal Article
• Language: English
• Published: IEEE 15.08.2025
• Subjects: Distributed estimation; Estimation; Filtering; H∞-consensus; integral measurements; Intelligent sensors; local performance analysis (LPA); Performance analysis; Petroleum; Protocols; round-robin (RR) protocol; Schedules; Sensors; Time-varying systems; Topology; weighted try-once-discard (WTOD) protocol
• ISSN: 1530-437X; 1558-1748
• DOI: 10.1109/JSEN.2025.3582166

This article addresses the distributed <inline-formula> <tex-math notation="LaTeX">{H}_{\infty } </tex-math></inline-formula>-consensus estimation problem for a class of discrete time-varying systems with integral measurements under the schedule of both the generalized Weighted Try-Once-Discard (WTOD) protocol and the Round-Robin (RR) protocol. The system and the integral measurement of each node are first formulated and the lifting technique is then employed to obtain a new delay-free state space model. The WTOD protocol determines multiple elements of measurement to be transmitted to the estimator and the RR protocol chooses multiple neighboring nodes to transmit its information, which are helpful to achieve the desired filtering performance in a flexible manner. By means of the protocol schedule, the distributed estimator is designed by leveraging the available information from the sensor and its neighboring nodes. A finite-horizon <inline-formula> <tex-math notation="LaTeX">{H}_{\infty } </tex-math></inline-formula>-consensus performance constraint is introduced to evaluate both the estimation error of each node and the consensus errors between neighboring nodes. Using the local performance analysis (LPA) method inspired by stochastic vector dissipation, a set of locally sufficient conditions is constructed for each node such that the estimation error dynamics satisfy the <inline-formula> <tex-math notation="LaTeX">{H}_{\infty } </tex-math></inline-formula>-consensus performance constraint over the finite horizon. Furthermore, the desirable estimator parameters are calculated at each node upon receiving the necessary information via the shared channel. For each node, an optimization problem subject to the feasible local conditions is established by minimizing the noise attenuation level, which is implemented in a fully distributed manner. Finally, the effectiveness and applicability of the distributed estimation algorithm developed in this article are demonstrated by an illustrative simulation.