Some properties of the Riemannian distance function and the position vector X, with applications to the construction of Lyapunov functions

Abstract

The quadratic distance function on a Riemannian manifold can be expressed in terms of the position vector, which in turn can be constructed using geodesic normal coordinates through consideration of the exponential map. The formulas for the derivative of the distance are useful to study Lyapunov stability of dynamical systems, and to build cost functions for optimal control and estimation.Index Terms-Riemannian geometry, geodesic distance, Lyapunov functions. I. NORMAL COORDINATES AND THE POSITION VECTORConsider an n-dimensional manifold X on which we can define a Riemannian metric , , and pick a particular point O as the origin of a system of geodesic normal coordinates. These are obtained by considering all geodesic curves that depart from O. Each geodesic has some initial velocity V , that is, it is a curve γ(t), t ∈ [0, 1] such that(1)Writing the components of the vector V at the origin as {x i }, we then assign the coordinates {x i } to the point γ(1).Repeating the construction for each V in some subset of T O X containing the null vector, we obtain a local system of coordinates around O, which may be extended to any region which does not contain points conjugate to O, that is, where the geodesic curves do not cross or touch. The classic construction of the map from T O X to X thus defined, called the exponential map, is detailed in any good book on Riemannian geometry, such as [1]. With the family of geodesics so defined, we may call the vector field X taking the values X| γ(1) =γ(1) the "position vector." Its expression in the said normal coordinates is X = x i ∂/∂x i . (In contrast, the expression in normal coordinates of the radial geodesic passing through a fixed point {x i } at t = 1 is γ i (t) = tx i .) Keep in mind however than the definition of X is intrinsic, that is to say, independent of the coordinates chosen.In the following sections we shall describe the quadratic distance function and its properties in terms of the vector field X, show how these properties can be useful in the study F Pait is grateful for the generous support of Susanna Stern and the continued interest of readers like you. He is with the Escola Politécnica (which unlike its French counterpart is not called X), of Lyapunov stability, and establish related formulas for the Lie derivatives of the metric tensor and the connection. These formulas will be used in future work concerning algorithms for the "recursive" construction of geometries with appropriate properties. II. PROPERTIES OF THE QUADRATIC DISTANCE FUNCTIONConsider the quadratic distance function defined at each point γ(1) of a manifold bywhere γ(0) = O,γ i (0) = x i , and ∇γγ = 0, that is, γ(t) is a geodesic connecting O and a point γ(1) whose normal coordinates are {x i }. In this paper, we will always consider ∇ as the torsion-free, Levi-Civita connection compatible with the metric , . Functional (2), the integral of the square norm of a curve, is variously called the action integral by physicists and the energy of the curve by mathematicians. We prefer to think of ...

Geometry-aware manipulability learning, tracking, and transfer

“…where is a vector field composed of the initial velocities of all geodesics departing from the origin , and is the inner product (3). As proved by Pait and Colon (2010) , the function (42) is a Lyapunov function for a dynamical system such that if the Lie derivative is negative everywhere except at the origin . To verify this condition, we first express the velocity of the dynamical system (41) in the tangent space of using parallel transport as…”

Geometry-aware Manipulability Learning, Tracking and Transfer

“…where F = Log M (M ) is a vector field composed of the initial velocities of all geodesics departing from the origin M , and •, • M is the inner product (3). As proved in [Pait and Colón (2010)], the function ( 42) is a Lyapunov function for a dynamical system Ṁ = h(M ) such that h( M ) = 0 if the Lie derivative L h V (M ) = 2 h, F M is negative everywhere except at the origin M . To verify this condition, we first express the velocity of the dynamical system (41) in the tangent space of M using parallel transport as…”

A Local Characterization of Lyapunov Functions and Robust Stability of Perturbed Systems on Riemannian Manifolds

“…The following theorem gives the existence of Lyapunov functions and also characterizes their properties for locally asymptotically stable systems evolving on Riemannian manifolds in normal neighborhoods of equilibriums of dynamical systems. In [30,35] the Riemannian distance function is employed as a candidate to construct Lyapunov functions for dynamical systems on Riemannian manifolds. For general discussions on the construction of Lyapunov functions on Riemannian manifolds and metric spaces see [6,7,13,28,30,35].…”

“…In [30,35] the Riemannian distance function is employed as a candidate to construct Lyapunov functions for dynamical systems on Riemannian manifolds. For general discussions on the construction of Lyapunov functions on Riemannian manifolds and metric spaces see [6,7,13,28,30,35].…”