Stability Margin Scaling Laws for Distributed Formation Control as a Function of Network Structure

TL;DR

This paper develops a PDE-based methodology to analyze and improve the stability margins of large-scale distributed vehicle formations, showing how network structure and mistuning enhance stability as the number of vehicles grows.

Contribution

It introduces a PDE approximation for formation stability analysis and demonstrates how higher-dimensional network structures and mistuning improve stability margins.

Findings

Stability margin approaches zero as vehicle number increases.

Higher-dimensional information graphs improve stability scaling.

Mistuning significantly enhances the stability margin.

Abstract

We consider the problem of distributed formation control of a large number of vehicles. An individual vehicle in the formation is assumed to be a fully actuated point mass. A distributed control law is examined: the control action on an individual vehicle depends on (i) its own velocity and (ii) the relative position measurements with a small subset of vehicles (neighbors) in the formation. The neighbors are defined according to an information graph. In this paper we describe a methodology for modeling, analysis, and distributed control design of such vehicular formations whose information graph is a D-dimensional lattice. The modeling relies on an approximation based on a partial differential equation (PDE) that describes the spatio-temporal evolution of position errors in the formation. The analysis and control design is based on the PDE model. We deduce asymptotic formulae for the closed-loop stability margin (absolute value of the real part of the least stable eigenvalue) of the controlled formation. The stability margin is shown to approach 0 as the number of vehicles N goes to infinity. The exponent on the scaling law for the stability margin is influenced by the dimension and the structure of the information graph. We show that the scaling law can be improved by employing a higher dimensional information graph. Apart from analysis, the PDE model is used for a mistuning-based design of control gains to maximize the stability margin. Mistuning here refers to small perturbation of control gains from their nominal symmetric values. We show that the mistuned design can have a significantly better stability margin even with a small amount of perturbation. The results of the analysis with the PDE model are corroborated with numerical computation of eigenvalues with the state-space model of the formation.