Networks obtained by Implicit-Explicit Method: Discrete-time distributed median solver

TL;DR

This paper introduces a discrete-time distributed median solver using an Implicit-Explicit Method, improving robustness and eliminating chattering issues present in previous continuous-time algorithms, with verified faster convergence in simulations.

Contribution

It proposes a novel discrete-time median solver based on Implicit-Explicit Method that avoids chattering and reduces iteration count, enhancing robustness in distributed consensus.

Findings

Eliminates chattering in median consensus algorithms.

Requires fewer iterations to reach steady-state.

Applicable to undirected, connected networks.

Abstract

In the purpose of making the consensus algorithm robust to outliers, consensus on the median value has recently attracted some attention. It has its applicability in for instance constructing a resilient distributed state estimator. Meanwhile, most of the existing works consider continuous-time algorithms and uses high-gain and discontinuous vector fields. This issues a problem of the need for smaller time steps and yielding chattering when discretizing by explicit method for its practical use. Thus, in this paper, we highlight that these issues vanish when we utilize instead Implicit-Explicit Method, for a broader class of networks designed by the blended dynamics approach. In particular, for undirected and connected graphs, we propose a discrete-time distributed median solver that does not suffer from chattering. We also verify by simulation that it has a smaller iteration number required to arrive at a steady-state.