Parallelobox: Improved Decomposition for Optimized Parallel Printing using Axis-Aligned Bounding Boxes

Hayley Hatton, Muhammed Khalid, Umar Manzoor, John Murray · 2026

Read Parallelobox: Improved Decomposition for Optimized Parallel Printing using Axis-Aligned Bounding Boxes by Hayley Hatton, Muhammed Khalid, Umar Manzoor, John Murray (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-2603-29579.txt.

Introduction

Hatton and co-authors present Parallelobox, a method for splitting 3D prints across several printers at once. Splitting models helps fit large prints or cut support needs. Their method uses axis-aligned bounding boxes as height fields. Boxes grow to suit a parallel printing goal, tuned by a metaheuristic search. Simple mesh clipping then cuts the model along the boxes. That yields fast decomposition with good print traits. Tests against two rival methods used simulated printing. Parallelobox won on the parallel print time metric.

Catalog record

CREATORHayley Hatton, Muhammed Khalid, Umar Manzoor, John Murray
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE61851
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2603.29579

Browse the full library · About the Tower of Babel

Full text

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

Parallelobox: Improved Decomposition for Optimized Parallel Printing using Axis-Aligned Bounding Boxes Hayley Hatton, Muhammed Khalid, Umar Manzoor, John Murray Abstract Much contemporary research in additive manufacturing focuses on breaking down models into constituent parts in the pursuit of various factors, such as printability of large models in smaller printing volumes, or reduction of support structures. Newer research has begun to focus on using these decomposition processes for printing models across multiple printers in parallel. We present a novel approach to this that incorporates axis-aligned bounding boxes as height fields to improve the characteristics of decomposition, including printing time, feasibility, and aesthetics. By expanding these bounding boxes according to a parallel printing objective, with additional improved efficiency from a metaheuristic process, these boxes can then be used for rapid decomposition using simple out-of-the-box mesh clipping operations. This algorithm is experimentally evaluated across a range of models against two other contemporary approaches to parallel printing that use more rudimentary techniques, such as recursive symmetry and cube skeletonization. Parallelobox outperformed each of these across a range of sample models on the basis of a ”parallel printing time” metric using simulated 3D printing to compute the results. Fig. 1 : Overview of the Parallelobox process, from model to printing and assembly. I Introduction In the field of additive manufacturing, ”decomposition” (also known as ”partitioning”) is often used to break apart a model into smaller pieces for various purposes, such as, for example, enabling the model to be printed in a smaller printing volume than the full size of the model, or reducing the support structures required to print the model. A more novel application of decomposition that has emerged in recent years involves utilizing multiple sets of 3D printers to print parts of models simultaneously in parallel to reduce the real time required to print them [ 2 ] [ 9 ] . This application imposes different constraints and requirements on the decomposition process. Where other algorithms may freely tolerate the production of decomposed parts with large disparities in size and surface area (should this be acceptable in pursuit of their own objectives), and may instead focus on reducing the aggregate printing time, in the case of parallel printing, printer utilization is most important. What this means is that a longer aggregate printing time is perfectly acceptable if it is translated into a shorter parallel printing time [ 6 ] . In an ideal algorithm, given any number of printers, a perfectly even distribution of printing time across the decomposed parts can be rapidly computed without any other side-effects (for example, impairments in the final assembly’s aesthetic). This would effectively mean that all available printers are utilized perfectly, with no redundant printer time whatsoever. However, there are many different forms this can take, as there are potentially infinite ways to decompose models subject to this singular constraint. This opens up the potential for approaches that can achieve this as best as possible given certain desired parameters, while potentially improving other aspects of the process as well. In this paper, an alternative approach to decomposition is proposed. The concept of ”axis-aligned bounding boxes” (AABBs) is utilized as a core tool for guiding the partitioning process. These axis-aligned bounding boxes can effectively become quadratic ”height field” blocks that represent the surface of a model when arranged properly across it. When represented this way, height field blocks supply desirable properties to the algorithm, such as reduced overhang and the possibility of eliminating significant quantities of internal volume. This can primarily reduce printing time and material usage while also improving aesthetic quality. Their axis-aligned nature constrains the complexity of their implementation and makes their application considerably more feasible than oriented boxes, which are much more complicated to arrange. This paper presents a novel algorithm that combines the distribution of axis-aligned bounding boxes across input models with an objective that produces a decomposition set with minimal parallel printing time, while secondarily reaping the additional benefits from using AABBs in aesthetic quality. This algorithm is compared to two other contemporary papers in the field of geometry decomposition for parallel printing, using simulated data across a disparate range of valid models. The results demonstrate the temporal superiority of this algorithm in comparison, where this algorithm was considerably more efficient in general and more flexible in how it tolerates a range of different geometries. This is achieved through the efficient positioning and expansion of height blocks that can be aware of the surface properties of the geometry, as compared to the more top-down approaches of the other two papers. This paper offers the following technical contributions: • An algorithm for decomposing models for printing simultaneously across a given number of printers that, at its core, uses axis-aligned blocks as height fields for a flexible and even distribution of mass while reducing overhanging areas. • Further refinements to the performance of the algorithm using a symmetry-based decomposition step, k-means++ clustering for initial block placement, and a metaheuristic for printer quantity determination. • An experiment comparing this algorithm with two pre-existing ”parallel partitioning” algorithms, demonstrating the superiority of this approach in reducing the ”parallel printing time”. II Related Work II- 1 Decomposition for Additive Manufacturing This paper branches from the higher domain of research that involves the optimization of additive manufacturing processes, specifically the subdomain that involves decomposition or partitioning models into constituent parts. One of the defining papers for this is Chopper [ 11 ] , which approaches the familiar problem of having to fit models into smaller printing volumes than themselves by decomposing them into smaller parts. To this end, it employs binary space partitioning (BSP) strategies to decompose models through planar cuts. The substance of the algorithm is in arranging these BSPs in a way that best fits the objective of printability, while optimizing other qualities such as structural stability, connection strength, and aesthetics (e.g. hiding the seam). Prioritization is achieved through the evaluation of a set of objective functions which rank the BSPs in terms of these qualities. Another, more recent paper showing evolution from this archetype, is Optimizing Object Decomposition to Reduce Visual Artifacts in 3D Printing [ 4 ] , which tackles the same issue of breaking models down into smaller parts to be printed, but iterates on the limitations of planar cuts by using ”non-planar cuts” instead. This process begins with oversampling patches of the model surface, whose boundaries demarcate potential cut regions. A labeling ”Integer Linear Program” is applied to refine this initial segmentation, allowing cuts to deviate in such a way as to preserve appearance from seams by directing them through regions identified as ”occluded” through the use of an ambient occlusion mapping function, while also incentivizing against overhanging regions to reduce the need for supports. The quantity of cuts required is ultimately selected based on the size of the model. Their reasoning is that a greater number of cuts only makes sense for larger models, as smaller models are better printed and assembled in fewer pieces. Their results demonstrate this using both quantitative and qualitative assessments on a range of largely complex 3D models, but lacked in ensuring printability due to the lack of maximum size limitations of produced parts, and ensuring the feasibility

Read the complete text