Distributed Matrix Pencil Formulations for Prescribed-Time Leader-Following Consensus of MASs with Unknown Sensor Sensitivity

TL;DR

This paper develops a robust distributed control framework for multi-agent systems with unknown sensor sensitivities, enabling prescribed-time leader-following consensus using matrix pencil formulations and novel gain strategies.

Contribution

It introduces a new distributed matrix pencil formulation and a bounded time-varying gain method for prescribed-time consensus in heterogeneous MASs with unknown sensor sensitivities.

Findings

Achieved consensus within prescribed time using the proposed method.

Demonstrated robustness to sensor sensitivity variations.

Validated effectiveness on a group of robot manipulators.

Abstract

In this paper, we address the problem of prescribed-time leader-following consensus of heterogeneous multi-agent systems (MASs) in the presence of unknown sensor sensitivity. Under a connected undirected topology, we propose a time-varying dual observer/controller design framework that makes use of regular local and inaccurate feedback to achieve consensus tracking within a prescribed time. In particular, the developed analysis framework is applicable to MASs equipped with sensors of different sensitivities. One of the design innovations involves constructing a distributed matrix pencil formulation based on worst-case sensors, yielding control parameters with sufficient robustness yet relatively low conservatism. Another novelty is the construction of the control gains, which consists of the product of a proportional coefficient obtained from the matrix pencil formulation and a classic time-varying function that grows to infinity or a novel bounded time-varying function. Furthermore, it is possible to extend the prescribed-time distributed protocol to infinite time domain by introducing the bounded time-varying gain technique without sacrificing the ultimate control accuracy, and the corresponding technical proof is comprehensive. The effectiveness of the method is demonstrated through a group of 5 single-link robot manipulators.