Curvature-Based Optimal Polynomial Geometric Interpolation of Circular Arcs Journal: Computer Aided Geometric Design

Ema Češek, Aleš Vavpetič · 2025

Read Curvature-Based Optimal Polynomial Geometric Interpolation of Circular Arcs Journal: Computer Aided Geometric Design by Ema Češek, Aleš Vavpetič (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2509-01353.txt.

Introduction

Drawing a circular arc with a polynomial curve is a basic task in design software, and the best fit depends on how error is judged. Ema Cesek and Ales Vavpetic study optimal approximations where the error is measured through curvature, how sharply the curve bends. They require geometric continuity, smooth joins of stated order, at the arc's end points.

Three cases are analyzed: a parabolic curve with basic continuity, a cubic with first-order smoothness, and a quartic with second-order smoothness. For each, the authors work out the optimal choice under the curvature standard. They also compare those results with approximations judged by radial error, the distance from the true circle. The study gives designers a clear account of what curvature-based fitting buys, and how it differs from the more common radial standard.

Catalog record

CREATOREma Češek, Aleš Vavpetič
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE51 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2509-01353

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Curvature-Based Optimal Polynomial Geometric Interpolation of Circular Arcs Journal: Computer Aided Geometric Design Ema Češek, Aleš Vavpetič Abstract The problem of the optimal approximation of circular arcs by parametric polynomial curves is considered. The optimality relates to the curvature error. Parametric polynomial curves of low degree are used and a geometric continuity is prescribed at the boundary points of the circular arc. Analysis is done for cases of parabolic G 0 G^{0} , cubic G 1 G^{1} and quartic G 2 G^{2} interpolation. The comparison of the approximation of circular arcs based on curvature with the approximation based on radial error is provided. 2010 MSC 65D05 , 65D07 , 65D17 1 Introduction The approximation of circular arcs is not only interesting from a theoretical perspective but also important for practical applications. Many control systems use parametric polynomials as basic building blocks. Since circular arcs cannot be represented in this way, it is important to find the best possible polynomial approximations. Several error functions can be used to measure the quality of approximants. If we are only interested in the visual image of the approximant, it is probably best to minimise the Hausdorff distance between the circular arc and the approximant. It is well known that the Hausdorff distance between a circular arc and a curve is (under some conditions) equal to the maximum of the radial distance between the arc and the approximant [ 4 ] . This fact makes the analysis of the problem easier. Still, the error function is irrational since part of it is the root of a polynomial function. In the first attempts to find the best approximation, the radial distance was replaced by the corrected polynomial error. The quadratic G^{0} case was studied by Mørken [ 7 ] , and the cubic G^{1} and the quartic G^{2} cases were considered by Hur and Kim [ 3 ] . Later, the problem was also solved with respect to the radial distance and for every inner angle 2\varphi\in(0,\pi] of the circular arc (see [ 10 ] ) and for some other cases with more parameters ( [ 9 ] , [ 8 ] ). Instead of the radial distance, the error function can be used to that measure how the curvature of the approximant matches the curvature of the circular arc. Such an approximation is useful, for example, in cases where the curve is considered as the trajectory of a moving body. The curvature function of the curve itself is more complicated than the distance function, so the analysis of finding the optimal approximant is more difficult and appears less frequently in the literature. Since the curvature of a circular arc is constant, one can determine the minimum value of the integral of the absolute value of the derivative of its curvature to measure the change of curvature of the approximant (see [ 5 ] ). However, more often the error function is taken to be the maximum of the absolute value of the difference between the curvature of the circular arc and of the approximant. Dokken, Dæhlen, Lyche and Mørken [ 2 ] constructed a cubic G^{1} approximant \boldsymbol{p} of a circular arc with the property that \|\boldsymbol{p}(\cdot)\|^{2}-1 is an equioscillating function; i.e. all five local extrema have the same absolute value. In [ 1 ] de Boor, Hölling and Sabin described a parametric cubic spline interpolation scheme for plane curves which works very well for circular arcs. In both papers, the authors also realised that the curvature of the approximant of the circular arc is very close to 1, the curvature of the unit circular arc. The graph of the curvature function of the approximant oscillates five times around the constant 1, which is the same property as for the graph of the radial distance of the approximant. The difference is that the local extrema agree in absolute value in the case of radial distance, whereas in the curvature, they do not. For an optimal approximant, it might be natural to require that the curvature is an equioscillating function. Unfortunately, it turns out that there are no such approximants except for only one inner angle 2\varphi . Kovač and Žagar [ 6 ] impose the condition that the curvature of the approximant coincides with the circular arc’s curvature at the midpoint and at the edges. They proved that their approximants have an optimal asymptotic approximation order. We show that their construction provides the optimal cubic G^{1} approximant only for inner angles smaller than some angle, which is calculated in the paper and at which the curvature is an equioscillating function, and that there are better approximations for circular arcs with an inner angle larger than the above one. We also show that something similar holds for the case of the quadratic G^{0} and the quartic G^{2} approximants. The paper is organized as follows. After the introduction in Section 1 , some basic preliminaries are explained in Section 2 . In Section 3 the quartic G^{0} case is considered and the main theorem Theorem 3.2 describes the optimal approximants. In Section 4 the cubic G^{1} case is considered. In the beginning, two cases motivate the analysis of the error function, and the main theorem Theorem 4.5 , which gives the algorithm for finding the optimal approximant, is stated at the end of the section. Section 5 deals with the quartic G^{2} case, where the algorithm for finding the optimal approximant is stated in Theorem 5.4 . Numerical examples are shown at the end of Section 4 and Section 5 . Some closing remarks are given in the last section. 2 Preliminaries Let us denote the considered circular arc with the inner angle 2\varphi by \boldsymbol{c} . More precisely, \boldsymbol{c} will be parametrized as \boldsymbol{c}\colon[-\varphi,\varphi]\to\mathbb{R}^{2} , 0<\varphi\leq\tfrac{\pi}{2} . It is enough to consider the unit circular arc centered at the origin of a particular coordinate system and symmetric with respect to the first coordinate axis. We can assume that \boldsymbol{c}(\alpha)=(\cos\alpha,\sin\alpha)^{T} . Let the polynomial approximation of degree n of \boldsymbol{c} be denoted by \boldsymbol{p}_{n}\colon[-1,1]\to\mathbb{R}^{2} , where \boldsymbol{p}_{n}=(x_{n},y_{n})^{T} , and x_{n} , y_{n} are scalar polynomials of degree at most n . The Bernstein-Bézier representation of \boldsymbol{p}_{n} will be considered, i.e. \boldsymbol{p}_{n}(t)=\sumop\displaylimits_{j=0}^{n}B_{j}^{n}(t)\,\boldsymbol{b}_{j}, where B_{j}^{n} , j=0,1,\dots,n , are (reparameterized) Bernstein polynomials over [-1,1] , given as B_{j}^{n}(t)=\binom{n}{j}\left(\frac{1+t}{2}\right)^{j}\left(\frac{1-t}{2}\right)^{n-j}, and \boldsymbol{b}_{j}\in\mathbb{R}^{2} , j=0,1,\dots,n , are the control points. Note that since \boldsymbol{c} is symmetric with respect to the first coordinate axis, the best parametric polynomial interpolant must possess the same symmetry, i.e. \boldsymbol{b}_{j}=r(\boldsymbol{b}_{n-j}) , j=0,1,\dots,n , where r\colon\mathbb{R}^{2}\to\mathbb{R}^{2} is the reflection over the first coordinate axis. Circular arc \boldsymbol{c} and the parametric polynomial \boldsymbol{p}_{n} have the geometric contact of order k at the boundary points \boldsymbol{c}(\pm\varphi) , if there exists a smooth regular reparametrisation \rho\colon[-1,1]\to[-\varphi,\varphi] with \rho^{\prime}(\pm 1)>0 , such that \frac{d^{j}\boldsymbol{p}_{n}}{dt^{j}}(\pm 1)=\frac{d^{j}(\boldsymbol{c}\circ\rho)}{dt^{j}}(\pm 1),\quad j=0,1,\ldots,k. In such case, we say that \boldsymbol{p}_{n} is a G^{k} approximation of \boldsymbol{c} . In particular, the G^{0} approximation is equivalent to the interpolation of boundary points, the G^{1} approximation provides that \boldsymbol{c} and \boldsymbol{p}_{n} share the same tangents at the boundary points and the G^{2} approximation additionally requires that \boldsymbol{c} and \boldsymbol{p}_{n} have the same curvature at the boundary points. Since the curvature of

Read the complete text