Technical Report No. VSR-04.02

End-of-Year Technical Report

For Project

Digital Human Modeling and Virtual Reality for FCS

By

The Virtual Soldier Research (VSR) Program

Center for Computer-Aided Design

College of Engineering

The University of Iowa

116 Engineering Research Facility

Iowa City, IA 52242-1000

VSR TEAM

K. Abdel-Malek, J. Arora, S. Beck, M. Bhatti, J. Carroll (Clarkson University), T. Cook, S. Dasgupta, N. Grosland, R. Han, H. Kim, J. Lu, C. Swan, A. Williams, J. Yang

K. Farrell, R. Vignes, T. Sinokrot, A. Mathai, T. Marler, J. Muhs, Q. Wang, X. Zhou, J. Lee, J. Kim, X. Man, S. Rahmatala, S. Dandach, R. Fetter, E. Horn, A. Patrick, Z. Mi,

Dated: October 25, 2004

CONTRACT/PR NO. DAAE07-03-D-L003/0001

1. Motivation for modeling NBC clothing for the soldier

Over the past two decades, significant advances in computer-based methods have permitted engineers to work increasingly with digital models of mechanical systems as opposed to more costly physical prototypes. Inevitably, however, when it is time to investigate performance of humans interacting with the mechanical system, a physical prototype is created and then tested with real humans. A major challenge in engineering design today is the creation of realistic digital human models that can interact with digital models of mechanical systems, further reducing the need for physical prototypes in the design process.

Digital human models are software-based algorithms that capture and predict how humans perform a variety of mechanical and perceptual tasks within different spatial environments. Such digital human models are frequently displayed in 3D virtual reality environments. There is a significant spectrum of research issues associated with creating realistic digital human models, including: human motion prediction; human dynamics; physiology with muscle activation and fatigue; control; and artificial intelligence, to name just a few. One research area of definite interest to the textile engineering community is that of mathematical modeling of clothing and its interaction with digital human models. A previous effort at modeling of clothing on “virtual actors” was undertaken by Volino et al [1], where the interest was primarily in the visual appearance and effects of the clothing models.

Many types of clothing can affect human mobility and comfort in performing physical tasks. For example, if the clothing binds on the wearer’s joints, it can restrict motion or make the performance of necessary tasks more difficult. Alternatively, if the clothing does not permit heat to be conducted or convected away from the body, the wearer can suffer heat-induced fatigue or stroke. To quantify the range of effects that different clothing prototypes can have on human performance in a digital human modeling framework, it is necessary that the clothing be characterized and mathematically modeled in a way that permits interaction with digital human models.

In modeling clothing systems, there are fabric characteristics and garment characteristics. Fabric characteristics derive from the constituent properties of the comprising fiber material, the manner in which fibers are spun into yarns, and the textiles into which the yarns are interwoven to form the fabric. Once these basic fabric characteristics are selected, the effective mechanical and thermal conductivity properties of the fabric can be computed using micromechanical analysis techniques such as unit-cell analysis and homogenization.

While garment characteristics are determined in part by the fabric properties, they are also determined by the cut of the fabric, the manner in which different garment pieces are sewn or stitched together, the fit of the garment to the body of the human being modeled, and the manner of fastening used to achieve closure of the garment.

In modeling clothing effects on digital human models, it is envisioned that numerical garment models will be draped onto digital human models of realistic anthropometric proportions, and baseline interaction forces necessary to achieve compatibility between the human models and the garments will be quantified. Then, as the digital human models perform simulated physical tasks that induce local shape and/or dimension changes, the additional interaction forces between the garment and human models will be quantified. These additive interaction forces can then be used to predict the ability and/or efficiency with which the human models can complete the tasks being considered.

2. Literature review of current cloth/textiles modeling techniques

Fabric and clothing modeling has been studied by both the computer graphics community and textile engineering community with different interests. Computer graphics researchers focus on the realistic visual representation of fabric deformations while textile engineers are more interested in accurate modeling and prediction of the mechanical properties of fabrics. Roughly speaking, research efforts from the computer graphics community are more related to macro-scale simulation of clothing while those in textile engineering usually involve more detail on the small-scale structure of fabrics.

Particle-based fabric models, which reproduce visually convincing draping configurations with relatively low computation cost, have been widely adopted in computer graphics and represent the state-of-the-art in macro-level clothing simulations. In this section, a review of particle-based models is conducted. In order to evaluate their physical accuracy, the constitutive relationships, i.e. the internal force formulations, employed in these models are emphasized.

2.1. Breen and House’s Model

A particle-based fabric model was first proposed by Breen and House [2]. In their work, fabric was discretized as an array of particles located along warp and weft yarns. Strain energy functions were defined to capture four basic mechanical interactions of the particles, repulsion, stretching, bending and trellising (in-plane shear). Static draping configurations were obtained by minimizing the total energy of all particles and a stochastic gradient descent method was developed by Breen et al. to search for the minima.

The repulsion energy was designed to keep particles at a minimum distance, preventing cloth self-intersection. It is given as follows

(1)

where is the distance between two particles and denotes the minimum distance. The stretching energy was defined as

. (2)

is a stiffness parameter, which is tuned to obtain desired effect. According to Breen, the combination of these two energy functions should “constrain each particle tightly to the nominal distance from each of its four-connected neighbors” and a “separation force” between neighboring particles is obtained as

. (3)

From this equation, one may notice that in addition to its original purpose, the repulsion energy also describes the compression behavior of fabric yarns. However, this may lead to an ambiguity regarding . For compression is the free distance between two neighboring particles, which depends on the grid density applied for discretization, while for repulsion is the minimum distance keeping clothing from self-contact. These two quantities are different in magnitude and need to be distinguished.

Bending about warp and weft directions was considered at each particle. For either direction, a bending angle (Figure 1a) formed by two line segments connecting a particle and its two neighbors was used to represent the bending deformation of the fabric patch centered at the particle. Experimental data produced by Kawabata Evaluation System [3] was used to construct the bending energy function. By assuming constant moment and curvature in this patch, the energy was given as

, (4)

where M and K denote moment and curvature in the bending direction and the relationship is readily available from corresponding Kawabata tests. Fitting an arc to a particle and its two opposite neighbors (Figure 1a), the curvature was related to the bending angle at the particle as

, (5)

and the bending energy in terms of is thus obtained.

Trellising deformation at a particle is measured by the shear angle shown in Figure 1b. The trellising energy function was defined by evaluating the work performed by external force in a Kawabata shear test. It is given as

, (6)

where and denote the measured shear force and angle respectively while is the width of the Kawabata shear test specimen. Since each particle represents a fabric patch, scaling Equation (6) by a ratio of the area of the patch to the area of actual Kawabata shear test specimen the trellising energy function is derived.

Breen’s model uses experimental data to construct the bending and shearing energies and applies strong springs to constrain the stretching in fabrics. These treatments represent the general constitutive relationships of fabrics and satisfactory draping results of various fabric types were obtained.

2.2. Eberhardt’s Model

Motivated by Breen’s work, Eberhardt et al. [4] proposed a simulation approach for dynamic fabric draping, which is governed by the equation

(7)

where and denote the location and the velocity of a particle respectively. is the Lagrange function of the system and it was written as

, (8)

where and are the kinetic energy and gravitational potential of particle i respectively and , and are strain energies corresponding to three types of internal forces, tension/compression, shearing and bending. Symbolic forms of the internal forces were derived and Equation (7) was reduced to a system of ordinary differential equations. A Runge-Kutta method with adaptive step-size control was suggested to solve the system.

To construct accurate energy functions, Kawabata experimental data was used. Piecewise linear functions were used to approximate the original Kawabata curves and two parameters, , the slope and , the intercept with x-axis, were retrieved for each linear approximation segment. Based on the two parameters, a quadratic form of strain energy for the approximated neighborhood was constructed. The bending energy at a particle was assumed to be a function of the two bending angles about the weft and the warp directions and it was given as

, (9)

The shearing energy at a particle was considered as a function of the four shear angles formed by the gridlines connecting itself and its four directly connected neighbors and it was defined as

. (10)

Likewise, the tension/compression energy was implemented as follows

(11)

where and denote the locations of a particle and one of its four neighbors respectively and is the free distance between and .