Average Consensus with Dynamic Compression in Bandwidth-Limited Directed Networks

TL;DR

This paper introduces the PP-ACDC algorithm for average consensus in directed networks with limited bandwidth, using dynamic compression and adaptive quantization to ensure convergence without global information.

Contribution

The paper presents a novel distributed consensus algorithm that combines dynamic compression with adaptive quantization for directed networks, ensuring exact average convergence.

Findings

Converges to the exact average in directed networks

Uses adaptive quantization to handle bandwidth limitations

Validated through numerical analysis and simulations

Abstract

In this paper, the average consensus problem has been considered for directed unbalanced networks under finite bit-rate communication. We propose the Push-Pull Average Consensus algorithm with Dynamic Compression (PP-ACDC) algorithm, a distributed consensus algorithm that deploys an adaptive quantization scheme and achieves convergence to the exact average without the need of global information. A preliminary numerical convergence analysis and simulation results corroborate the performance of PP-ACDC.