Networked Estimation of Relative Sensing Multiagent Systems Using Reconfigurable Sets of Subobservers

Abstract

In this paper, a networked estimation framework for a team of relative sensing multiagent systems is proposed. This framework is constructed based on the developed notion of subobservers (SOs). Within a group of SOs, each SO is estimating certain team states that are conditioned on a given input, output, and other state information. The overall team estimation process is then modeled by a weighted estimation (WE) digraph. By selecting an optimal path in the WE digraph, a high-level supervisor provides a decision on the selection and reconfiguration of the set of SOs to successfully estimate all the states of the multiagent system. In presence of unreliable information due to either large disturbances, noise, and actuator faults, or communication delays certain SOs may become invalid and carry a high cost. In this case, the supervisor reconfigures the set of SOs by selecting a new path in the estimation digraph such that the impacts of these unreliabilities are managed and confined. This will consequently prevent the propagation of unreliabilities to the entire estimation process and avoid performance degradations to the multiagent system. Simulation results are conducted for a five spacecraft formation flight system in deep space where the validity and advantages of our analytical developed schemes are confirmed.