Nonlinear Fusion Estimation for Networked Sensor Systems
DescriptionA networked sensor system (NSS) consists of spatially distributed autonomous sensors to cooperatively monitor physical or environment conditions. Networked sensor systems (NSSs) can provide a powerful tool to capture the dynamical behavior of many real-world engineering systems such as smart grids, multi-robot localization and internet of things. In most applications, the monitored physical process is nonlinear and the disturbance noise is unknown but bounded (UBB). In fact, it is not easy to realize real-time communication of complete information in NSSs due to the constrains of communication bandwidth and sensor energy. This project aims to develop nonlinear fusion estimation algorithms of NSSs that can be used by a group of sensors to carry out prescribed tasks efficiently under UBB noises and incomplete information.Designing nonlinear estimators under UBB noises and fusion criteria for nonlinear systems are the most fundamental problems in fusion estimation. However, most of the existing nonlinear fusion estimation methods requires knowledge of the statistical property of disturbance noises. The existing linear fusion criteria may not reflect the real correlation between local nonlinear estimates because of the complexity of nonlinear dynamics. Notice that the statistical information of noises is difficult to obtain accurately in most practical systems, while UBB noises can be easily satisfied in practical applications. Hence, learning-based fusion strategies will be studied in this proposal to design efficient nonlinear estimation/fusion criteria under UBB noises.Bandwidth constraint and energy consumption are two of the most concerned problems in NSSs. Incomplete information caused by bandwidth and energy constraints will degrade the system performance seriously. It is necessary to design compensation strategies to minimize the information loss for NSSs. In particular, the complexity of nonlinear systems and unknown statistical information of noises do bring inherent difficulties in designing nonlinear fusion estimation schemes for NSSs. Hence, dynamical quantization fusion methods under distributed dimensionality reduction will be designed in this project to satisfy the limited communication capacity. Also self-triggered schemes for the nonlinear fusion estimation will be developed to save sensor energy. In particular, the developed nonlinear fusion estimation framework in this proposal will be applied to the monitoring and supervision of power grids described by IEEE bus models.The theoretical results generated in this proposal, if well accomplished, will lay a solid foundation for the development of estimation fields of nonlinear fusion, and also provide a fundamental investigation of nonlinear fusion estimation for NSSs.
|Effective start/end date||1/01/20 → …|