Topological Analysis of Vector-Field Guided Path Following on Manifolds

TL;DR

This paper extends vector-field guided path following to Riemannian manifolds, analyzing topological constraints and singularities that limit global convergence, especially for paths homeomorphic to the circle.

Contribution

It generalizes guiding vector fields to manifolds and proves topological limitations on global convergence for certain paths, especially circles, in Euclidean spaces.

Findings

Singular points of guiding vector fields are inherent for paths homeomorphic to the circle in Euclidean space.

Global convergence cannot be guaranteed for such paths due to topological constraints.

Existence of non-path-converging trajectories from boundary regions in higher-dimensional Euclidean spaces.

Abstract

A path-following control algorithm enables a system's trajectories under its guidance to converge to and evolve along a given geometric desired path. There exist various such algorithms, but many of them can only guarantee local convergence to the desired path in its neighborhood. In contrast, the control algorithms using a well-designed guiding vector field can ensure almost global convergence of trajectories to the desired path; here, "almost" means that in some cases, a measure-zero set of trajectories converge to the singular set where the vector field becomes zero (with all other trajectories converging to the desired path). In this paper, we first generalize the guiding vector field from the Euclidean space to a general smooth Riemannian manifold. This generalization can deal with path-following in some abstract configuration space (such as robot arm joint space). Then we show several theoretical results from a topological viewpoint. Specifically, we are motivated by the observation that singular points of the guiding vector field exist in many examples where the desired path is homeomorphic to the unit circle, but it is unknown whether the existence of singular points always holds in general (i.e., is inherent in the topology of the desired path). In the $n$-dimensional Euclidean space, we provide an affirmative answer, and conclude that it is not possible to guarantee global convergence to desired paths that are homeomorphic to the unit circle. Furthermore, we show that there always exist \emph{non-path-converging trajectories} (i.e., trajectories that do not converge to the desired path) starting from the boundary of a ball containing the desired path in an $n$-dimensional Euclidean space where $n \ge 3$. Examples are provided to illustrate the theoretical results.