Motivation:

A growth of locally available computing power, coupled with increasingly powerful and flexible communication structures, enables the use of decentralized or distributed processing concept nowadays. This concept makes the inference of the hidden state of complex and geographically widely distributed dynamic systems much more stable and flexible compared to centralized processing. Possible applications are such as distributed diffusion field estimation using wireless sensor networks, distributed cyber-physical systems in industry 4.0, distributed target tracking or localization as well as distributed autonomous spacecraft formation control.

Scenario:

In a general scenario, distributed subsystems/agents/nodes form a network topology and obtain local observations/measurements on the hidden system state (see Figure). Each node in the network will perform the estimation in a distributed and cooperative fashion with information exchange via inter-node communication links. 

Objective:

To estimate the state of a dynamic system in a distributed manner and optimally adapt to dynamic properties of the physical processes, a new design of in-network state estimation algorithms and their extension to nonlinear ones will be studied in order to obtain a comparable robust and stable state estimation result meanwhile considering communication constraints. Consensus-based state estimation will also be enforced in the network on the demand for subsequent processing and control.

Details