Discrete-Time Consensus Filters for Average Tracking of Time-Varying Inputs on Directed Switching Graphs

In this paper, we consider discrete-time distributed estimation on a directed graph with switching topologies. Motivated by a recent PI consensus filter, we modify the protocol and remove the limitation of undirected graph and gain conditions with requirement for global topological information. The protocol is then extended to switching topologies. Convergence analysis is conducted for constant inputs under both balanced directed and general directed graphs with topological switching. Then, the result is extended to all cases for time-varying inputs with an analytical derivation of the error bound. Satisfactory simulation results are shown to validate theoretical claims.