Directional Mollification for Controlled Smooth Path Generation

Alfredo González-Calvin, Juan F. Jiménez, Héctor García de Marina ([email protected]) · 2026

Read Directional Mollification for Controlled Smooth Path Generation by Alfredo González-Calvin, Juan F. Jiménez, Héctor García de Marina ([email protected]) (2026) 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-2603-21831.txt.

Introduction

González-Calvin and co-authors smooth polygonal chains into curves that still pass exactly through every vertex. Standard mollification rounds a path but loses the vertices, which breaks waypoint fidelity. Their directional mollification builds infinitely smooth curves that stay close to the chain and interpolate its points. Changing one segment affects the output only in a small, known neighborhood. The method also gives closed-form bounds on curvature. A wider family unifies both versions, with uses in robotics, graphics, and CNC machining.

Catalog record

CREATORAlfredo González-Calvin, Juan F. Jiménez, Héctor García de Marina ([email protected])
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE69955
LICENSEhttp://creativecommons.org/licenses/by-sa/4.0/
RIGHTSCC BY SA
SOURCEarXiv

Source: https://arxiv.org/abs/2603.21831

Browse the full library · About the Tower of Babel

Full text

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

Directional Mollification for Controlled Smooth Path Generation Thanks: Submitted to the editors March 19, 2026. Alfredo González-Calvin, Juan F. Jiménez, Héctor García de Marina ([email protected]) Abstract Path generation, the problem of producing smooth, executable paths from discrete planning outputs, such as waypoint sequences, is a fundamental step in the control of autonomous robots, industrial robots, and CNC machines, as path following and trajectory tracking controllers impose strict differentiability requirements on their reference inputs to guarantee stability and convergence, particularly for nonholonomic systems. Mollification has been recently proposed as a computationally efficient and analytically tractable tool for path generation, offering formal smoothness and curvature guarantees with advantages over spline interpolation and optimization-based methods. However, this mollification is subject to a fundamental geometric constraint: the smoothed path is confined within the convex hull of the original path, precluding exact waypoint interpolation, even when explicitly required by mission specifications or upstream planners. We introduce directional mollification, a novel operator that resolves this limitation while retaining the analytical tractability of classical mollification. The proposed operator generates infinitely differentiable paths that strictly interpolate prescribed waypoints, converge to the original non-differentiable input with arbitrary precision, and satisfy explicit curvature bounds given by a closed-form expression, addressing the core requirements of path generation for controlled autonomous systems. We further establish a parametric family of path generation operators that contains both classical and directional mollification as special cases, providing a unifying theoretical framework for the systematic generation of smooth, feasible paths from non-differentiable planning outputs. keywords mollification, path generation, path smoothing, directional mollification, robotics, CNC Funding. This work is especially supported by the FPU program of the Ministry of Science, Innovation and Universities of Spain, and it is supported by iRoboCity2030-CM, Ref TEC-2024/TEC-62, financed by Comunidad Autónoma de Madrid (Spain) and by the ERC Starting Grant iSwarm 101076091 and the RYC2020-030090-I grant from the Spanish Ministry of Science. Corresponding author: [email protected] † † runningheads: Directional Mollification for Path Generation / A. González-Calvin, J. F. Jiménez, and H. G. de Marina MSC 65D10, 65D15, 53A04, 93C85, 44A35 1 Introduction Autonomous robot motion is typically the result of a layered decision-making process. A path planner first computes a collision-free route through the configuration space, usually encoded as a sparse sequence of waypoints via graph-search methods such as A* or sampling-based planners such as Rapidly-exploring Random Trees (RRT) [ 12 ] . This discrete output must then be processed by a path generation step, which produces a smooth geometric path or time-parameterized trajectory that downstream controllers can reliably track. Spline-based approaches, including cubic B-splines [ 11 , 1 ] and Bézier curves [ 19 ] , are among the most widely used tools for this purpose, providing C^{2} continuity and bounded curvature, although they are prone to large curvature excursions when waypoints are irregularly distributed [ 13 ] . Gaussian process regression [ 16 ] and kernel smoothing [ 7 ] offer C^{\infty} smoothness but sacrifice exact waypoint interpolation and introduce O(n^{3}) computational overhead alongside potential numerical instability. Optimization-based formulations [ 14 , 17 ] embed curvature constraints within convex programs, supporting real-time execution, yet their performance is sensitive to parameter selection and requires careful tuning [ 8 ] . In the realm of industrial robots and CNC machines, there have been efforts to smooth linear paths (i.e., paths created by linearly interpolating points in the Euclidean space or \mathrm{SO}(3) ) [ 6 ] . Analytical methods are used in [ 15 ] and [ 18 ] , to create smooth paths for industrial robots, while B-splines are used in [ 9 ] to control the approximation error and the curvature of the smoothed curve in CNC machining, thus generating feasible paths. Nevertheless, the proposed methods trade-off computational simplicity for smooth and curvature guarantees. Recently, mollification has been introduced as an efficient path generation technique [ 5 ] . The method operates by convolving the original path with mollifiers—smooth, well-behaved functions with compact support—producing a C^{\infty} approximation of any path, even discontinuous ones, with arbitrary precision. As a particular application, it can be used to generate smooth paths from a linearly interpolated sequence of waypoints, while providing a simple closed-form curvature upper bound that enables straightforward feasibility assessment. Despite these advantages, mollification has a fundamental geometric limitation: the smoothed path is confined within the convex hull of the original one, meaning that the defining points, such as waypoints, of the original path cannot be enforced to be points of the generated curve. The main contribution of this paper is to address the inherent limitation of conventional mollification [ 5 ] of not being waypoint-preserving . To this end, we introduce a new term, which we call the directional derivative term, that is incorporated into the conventional mollification. This term computes a weighted mean of the directional derivatives at each point of the original path, using the same mollifier as weight. The resulting technique, which we call directional mollification , ensures that if the original function is convex (resp. concave)—as is the case for the components of a curve generated by linearly interpolated points—the mollified function remains below (resp. above) the original, effectively circumventing the convex hull constraint. The smoothness and convergence properties of conventional mollification are inherited to some extent. Furthermore, by appropriately selecting the parameters, the directionally mollified curve is guaranteed to be C^{\infty} while passing exactly through all points defining the original piecewise structure, e.g., making waypoints invariant under the directional mollification process. As an example, see fig. 1 , where the original piecewise linear function— the origin could be considered a waypoint—is shown in black, and the conventional and directional mollifications in green and red, respectively. As can be seen, the conventional mollification is restricted to the convex hull (of the image of) the original function, and does not intersect the origin, while the directional mollification does, and, as we will show in this work, preserves the smoothness and convergence properties of conventional mollifiers. Finally, as a last result, introducing a weight on the directional derivative term yields a parametric family of C^{\infty} functions sharing the same convergence properties, of which both conventional mollification and directional mollification are special instances. The remainder of this paper is organized as follows. Section 2 introduces the notation and preliminaries. Section 3 reviews the concept of mollifiers, the main results of conventional mollification , and the directional derivative of convex functionals. Section 4 presents the novel concept of directional mollification and discusses its properties and relationship to conventional mollification. Section 5 applies these results to linearly interpolated points and other curves, covering the key practical contributions of this work, including the waypoint-preserving property and the possibility of switching continuously between conventional and directional mollification. Section 6 unifies both approaches into a single parametric family, and section 7 concludes the paper. 2 Notation We define as \mathbb{N}=\{1,2,\dots,\} the set of positive integers, and as \mathbb{N}_{0}=\mathbb{N}\cup\{0\} . All integrals must be thought with respect to the Lebesgue measure and the Borel sigma algebra. For a function f\in L^{p}(X):=L^{p}(X,\mathbb{R}) with p\in[1,\infty] , we denote the L^{p} norm as ||f||_{p} while for a vector x\in\mathbb{R}^{n} its \ell_{p} norm as ||x||_{p} . It will be clear from the context their difference. We say that f:X\to\mathbb{R} is locally

Read the complete text