The Community for Technology Leaders

Guest Editor's Introduction: Computer Graphics in Textiles and Apparel Modeling

David E. Breen, California Institute of Technology

Pages: pp. 26-27

Woven cloth has been part of the human experience for thousands of years. The oldest known relic of human industry is a piece of hemp fabric dating back approximately 8,000 years. Given the scope of human exposure to cloth, it's surprising how little is known about its properties. This "old-fashioned" material, which we take for granted every day, is actually quite complex. Its complexity brings a versatility that makes it useful not only for clothing and sails, but also as an industrial material. Woven cloth is an integral component of certain composite materials used to reinforce structural elements and provide a sturdy "skin" for many advanced aircraft and vehicles.

The reasons for modeling textiles and apparel are numerous. In the computer animation/entertainment field, the demand for more realistic virtual actors increases the need for faster, better, and easier tools to clothe them. The tools are being developed. It is now time to move them out of the research labs into the entertainment studios for use with the computer-generated characters we see on television and in the movies.

The need for textile modeling technologies is even greater within the CAD/CAM field. Certainly, many of the clothing modeling tools developed for the entertainment industry can also be used to design real clothing. Future apparel CAD systems should allow fashion designers to experiment easily with a variety of fabrics and patterns on a 3D dynamic virtual mannequin before the actual garment is manufactured. Once the design is complete, it can be sent to an automated loom that weaves the fabric and then to a computer-controlled cutting table that nests and cuts the appropriate patterns. Customers might even try on a virtual garment in an augmented reality environment before having it custom sewn.

The textile modeling requirements for industrial applications are quite different from the requirements for entertainment and apparel design applications. Woven cloth is often used within composite materials because it can be formed into highly complex shapes while still maintaining its outstanding mechanical and structural properties. In the composites domain, cloth is not allowed to freely drape over another object. Instead it is forced into a specific predetermined shape. Composite material designers need to know if a cloth can be deformed into the desired shape and, if so, what outline is required to cover a specific area. Additionally, they need a detailed structural analysis of the material's strength.

Computer-aided manufacturing places additional demands on textile and apparel modeling. Industry wants to automate the manufacture of clothing and other textile-based products, which is extremely labor-intensive. Integration of robotics into the manufacturing process is expected to save significant time and money. While robot manipulation of rigid objects, such as auto parts, has been thoroughly studied and is well understood, the automated manipulation of a highly flexible and complex material such as cloth still presents great challenges to the robotics community. Controlling a robot to properly handle and place a limp object is a difficult problem that requires extensive study. Modeling a cloth's dynamic properties and predicting its motion in reaction to applied forces are critical to solving the general control problem.

Two distinct groups have studied textile and apparel modeling, each with different goals and priorities. The computer graphics community's interest has grown significantly over the past 10 years, driven by a desire to include realistic, "physically based" cloth objects in images and animations. Most of this work has focused on developing computationally tractable models that produce "cloth-like" behaviors, and most of it is concerned with qualitative results that produce a particular visual effect and "look good."

Textile engineers have focused more on scientific correctness. Tracing the roots of modern textile modeling back more than 60 years to research performed by F.T. Peirce in the 1930s, this community has emphasized the low-level mechanical properties of cloth, relating its behavior to traditional mechanical parameters, such as Young's modulus, bending modulus, and Poisson's ratio. Much effort has been devoted to calculating stress-strain curves, load-extension relationships, and bending-moment dependence on curvature.

Another significant focus of textile engineering has been on modeling the micromechanical relationships of a single thread crossing. Until very recently, surprisingly little effort has gone into modeling the macroscopic deformations of cloth.

A main goal of this special issue of IEEE Computer Graphics and Applications is to bring these two research communities together. The textile engineering community has decades of experience measuring, studying, and analyzing cloth to achieve a detailed understanding of the low-level behavior of cloth. The result is a substantial body of empirical data describing the behavior of real cloth, which is essential for verifying the validity of any textile model.

The computer graphics community brings a broad background in computational geometry and physically based modeling to the problem of textile and apparel modeling, as well as numerous compelling applications. As these two groups learn more about each other's work, the cross-fertilization should significantly advance the technology and applications of these models.

About the articles

With articles from both computer graphics and traditional textile research groups based in North America, Europe, and Asia, IEEE CG&A has achieved the goals of this special issue. Four articles were selected from those submitted to the theme issue. The fifth article, Aono et al., had been previously submitted, reviewed, and accepted. It was included because of the relevance of its subject matter. The articles span a wide range of topics within textile and apparel modeling, including computer animation, dynamic cloth modeling, CAD for composite materials, and CAM for fabric manipulation.

Ng and Grimsdale present an overview of the current state of cloth modeling in computer graphics. They classify modeling techniques into three categories according to the nature of the main theory or method used in each one. The three categories are geometric, physical, and hybrid. Geometric techniques use a high-level geometric description to model the shape of cloth without considering its mechanical properties. The physical techniques produce fabric models by first performing a physical simulation. Hybrid techniques combine aspects of both the other two.

Volino, Magnenat Thalmann, Jianhua, and Thalmann survey the research being conducted at MIRAlab, University of Geneva, and the Swiss Federal Institute of Technology on clothing animation. Their article presents the evolution of their work toward the development of deformable clothes for virtual actors. It also describes their current general system for creating autonomous clothing for any synthetic human. This is a complete summary of the world's leading effort in clothing modeling and simulation.

Eberhardt, Weber, and Strasser present a significant advancement over previous particle-based cloth models by extending them to be dynamic and to include hysteresis—features that make these types of models more effective and practical. The authors have achieved substantial computational speed-ups through various optimizations that combine methods from computer algebra, theoretical physics, numerical mathematics, and ray tracing.

Aono, Denti, Breen, and Wozny detail algorithms that can be used within a woven cloth composite CAD system. Once a designer can simulate the forming of a flat woven sheet into a complex 3D shape, the CAD system often reveals unwanted "anomalous" events within the sheet. Given the limitations of the sheet's deformation properties, these anomalies can cause buckles and tears during the manufacturing process. Such cases require cutting pieces, or "darts," out of the cloth to accomplish the fitting. This article first defines dart models, then describes how appropriately inserted darts can suppress anomalies in 3D composite plies.

Eischen, Deng, and Clapp describe a model suitable for simulating the motion of a cloth sample being moved during a manufacturing process. The specific processes considered are fabric pick-up and lay-down, fabric draping on a stacker, and fabric folding. The model is based on stress-resultant, geometrically exact, nonlinear shell theory including nonlinear material response and an adaptive arc-length control algorithm. Their article also includes an excellent summary of the relevant textile engineering literature.


Textile and apparel modeling will become ever more important as clothing and composites producers demand better design tools. As the drive to automate textile manufacturing increases, control algorithms will require the detailed knowledge about the flexible material being manipulated that only simulation can provide. As the entertainment industry further expands its use of digital techniques, we can expect the animated characters on movie screens to soon be wearing computer-generated clothing. Even with decades of textile and apparel modeling research behind us, numerous challenges and unsolved problems lie ahead in this field. The articles in this issue demonstrate that we are well positioned to face these challenges.

About the Authors

David E. Breenis the assistant director of the Computer Graphics Laboratory at the California Institute of Technology. His research interests include physically based modeling, augmented reality, object-oriented computer animation, and geometric CAD. Breen received his BA in physics from Colgate University in 1982. He received his MS and PhD degrees in computer and systems engineering from Rensselaer Polytechnic Institute in 1985 and 1993. He is a member of the IEEE Computer Society, ACM Siggraph, Eurographics, and the Fiber Society.
67 ms
(Ver 3.x)