Edge-Dependent Distributed Moving Horizon Estimation With Application to UGV Localization

Le résumé fourni par la source

Safety-critical cyber–physical and Internet of Things systems increasingly operate over time-varying, intermittently disconnected networks, making distributed state estimation essential to sustain accuracy and real-time performance. This work investigates the problem of fast distributed moving horizon estimation (FDMHE) over time-varying sensor networks with disconnected topologies. Within an edge-dependent average dwell time (ADT) framework, we establish input-to-state stability (ISS) for switched nonlinear systems even when all subsystems are non-ISS, with classical ADT and mode-dependent ADT recovered as special cases. Building on this theory, we design an implementable FDMHE scheme based on local Gauss–Newton approximations that performs one local update and one neighbor exchange per sampling period, meeting fixed computation and communication budgets. A suboptimality-based analysis then establishes ISS of the estimation-error dynamics under time-varying and possibly disconnected communication topologies. Experiments on unmanned ground vehicle (UGV) localization confirm that the proposed estimator sustains real-time operation and high estimation accuracy despite switching network topologies that are disconnected, demonstrating practicality for resource-constrained deployments.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.