«12-1986 Modeling and Animating Human Figures in a CAD Environment Norman I. Badler University of Pennsylvania, badler Follow this and ...»
University of Pennsylvania
Technical Reports (CIS) Department of Computer & Information Science
Modeling and Animating Human Figures in a
Norman I. Badler
University of Pennsylvania, email@example.com
Follow this and additional works at: http://repository.upenn.edu/cis_reports
Part of the Computer Engineering Commons, and the Computer Sciences Commons
Norman I. Badler, "Modeling and Animating Human Figures in a CAD Environment",. December 1986.
University of Pennsylvania Department of Computer and Information Science Technical Report No. MS-CIS-86-88.
This paper is posted at ScholarlyCommons. http://repository.upenn.edu/cis_reports/995 For more information, please contact firstname.lastname@example.org.
Modeling and Animating Human Figures in a CAD Environment Abstract With the widespread acceptance of three-dimensional modeling techniques, high-speed hardware, and relatively low-cost computation, modeling and animating one or more human figures for the purposes of design assessment, human factors, task simulation, and human movement understanding has become feasible outside the animation production house environment. This tutorial will address the state-of-the-art in human figure geometric modeling, figure positioning, figure animation, and task simulation.
Disciplines Computer Engineering | Computer Sciences Comments University of Pennsylvania Department of Computer and Information Science Technical Report No. MSCIS-86-88.
This technical report is available at ScholarlyCommons: http://repository.upenn.edu/cis_reports/995
MODELING ·AND ANIMATING HUMAN
FIGURES IN A CAD ENVIRONMENTDr. Norman Badler MS-CIS-86-88 GRAPHICS LAB 14 Department Of Computer and Information Science School of Engineering and Applied Science University of Pennsylvania Philadelphia, PA 19104-6389 December 1986 · :Acknowledgements: The research reported here from the University of Pennsylvania is partially supported NASA Contract NAS9-17239, NSF CER Grant MCS 8219196, Lockheed Engineering and Management Services, and ARO Grant DAA6-29-84-K-0061, DAA29-84-9-0027.
MODELING AND ANIMATING HUMAN FIGURES
IN A CAD ENVIRONMENTDr. Norman I. Badler Computer and Information Science University of Pennsylvania Philadelphia, P A 19104-6389
ABSTRACTWith the widespread acceptance of three-dimensional modeling techniques, high-speed hardware, and relatively low-cost computation, modeling and animating one or more human figures for the purposes of design assessment, human factors, task simulation, and human movement understanding has become feasible outside the animation production house environment. This tutorial will address the state-of-the-art in human figure geometric modeling, figure positioning, figure animation, and task simulation.
INCORPORATING HUMAN FIGURE MODELS INTO 3-D CADAs more complex environmental 3-D objects are designed with computer-aided systems, incorporating a human figure into the design or the evaluation of a design is becoming crucial. Though not new, the graphical basis for creating, modeling, and controling one or more human figures in a 3-D world is expanding and the application base is growing. Human figure models have long been used in cockpit and automobile occupant studies [14, 12, 28]; now they are finding application in space station desig n, maintainence assessment, and product safety studies.
There are several methodologies for integrating human figure models into a CAD system model:
• The self-contained system: the human figure modeling system contains an internal CAD system with which to build the object models.
• The CAD integral: the CAD system contains a figure model which can be manipulated in the same fashion as other designed objects. Often the figure model is selected from a fixed set of size, shape, and position choices.
• The CAD adjunct: the CAD system produces output files which may be used by the human figure modeler to establish the environment; in return, the figure modeler can return the body as an object to the CAD system.
For example, historical human factors systems tended toward the first option, such as Combiman  and Sammie . With newer CAD modeling systems, the tendency swung to the second option whereby a specific model was built into the system, such as BUFORD  and PLAID . Most recently, the third option has proven viable with the appearance of TEMPUS  as an adjunct to an existing CAD system (in fact, PLAID). Most research-oriented efforts find this route the path of leas t resistance as it separates the more difficult human factors problems from the reasonably well understood geometric design system.
With the expansion of the 3-D CAD field to include a large number of competent and competitive products, the future of human figure modeling must be to fit into this environment rather than suppl ant it. That is, the human figure must become just one other object to the design system, albeit one with very special capabilities, requirements, and size variability. Thus the first two options have been the historical route to integrating people into the designed environment, while our own efforts have focused on the third approach as the more adaptable and viable way of managing existing software technology. This tutorial therefore intends to address the issues surrounding the incorporation of human figure models into an existing (or contemplated) 3-D, solid modeling, CAD environment.
Given that the third option is the one which most cleanly separates the issues of human figure modeling and CAD, there are a number of capabilities which must be addressed by a body modeling
• Creating and selecting individual or statistical human figure models, body sizes, and clothing.
• Providing interfaces to the CAD object information: object files, representation formats, and solid models.
• Providing graphics output: there must be suitable graphics for both bodies and objects, with the possibility of real-time viewing.
• Describing and animating the tasks to be performed.
• Offering appropriate analysis tasks: reach assessment, view assessment, collision detection, and population selection.
We will discuss all these items in the subsequent presentation. For a consistent on-going context, however, the discussion will be based on the TEMPUS system developed at the University of Pennsylvania. TEMPUS has been a relatively recent human factors system, was carefully designed and implemented from a Computer Science perspective, is undergoing active and continual development, and extends into some novel areas we feel are essential to human factors analysis.
ANTHROPOMETRYHuman bodies come in a wide variety of sizes. Anthropometry is concerned with measurement of a body to establish segment lengths (which cannot themselves be directly measured) and statistics on the distributiqn of body sizes in some population. A set of standard anthropometric measurements may be taken from a subject and segment lengths computed by regression equations to size all body segments.
A standard system for this computation has been developed by Analytics, Inc.: the CAR system .
Unfortunately their data seems to have been derived from male subjects, as the female data is just scaled. Though male/female differences may have been unimportant to the early pilot simulators, n broader application base has now made this essential. Segment length data is essential for the proper computation of body position during reach and view assessment and the establishment of postures such as sitting.
A distinction must be made between specific individuals in a database and generic individuals computed by regression formulas or statistical data. In fact, one should have the ability to customize bodies to examine typically worse case situations. For example, TEMPUS lets the user construct segment lengths by a combination of percentiles and actual lengths. Thus a user can create a generic individual, for example, based on a soth percentile body but with 90th percentile arms, lOth percentile legs, and a given specific torso length.
Besides segment lengths, the body somatotype is important in determining the girth of various body segments. Though not as important for reach assessment, girth information is essential for proper clearance assessment and impacts joint limits. Torso and pelvic width may be the limiting factors in sitting comfort. Arm girth may make reaching into tight spaces difficult. During normal breathing, the torso may expand about 4%.
In the case of space shuttle and space station design it is important to know that the body height increases up to two inches due to the lack of gravity and the consequent expansion of the spine and joints. Thus the segments are not evenly scaled; rather, the length difference must be made up in the torso.
The body is represented as a tree structure of joints and segments. Each joint is characterized by a rotation angle or angles, and each segment by its length between the proximal and distal joints.
(Proximal means closer to the body center and distal, further.) One distinguished joint is made the body root: sometimes the center of the pelvis (where the spine connects to the pelvis) and sometimes the center of mass. In the latter case, the location of the body root is not fixed in relation to another body joint; rather, there is a prismatic (sliding, variable length) "invisible" segment between the body center of mass and some actual body joint such as the center pelvis.
The location of the body root is important from a computational standpoint since it is from there that the nested transformations will be computed. Changing the body root to some other body joint is a useful and desirable capability. It permits some body joint to be considered to bej'txed relative to the world coordinate system. While the center of mass is the proper such point for the human figure in free-fall, the body root can be established elsewhere for other effects. For example, during a walk, the body pivots around the ankle; during that phase of motion setting the body root at the ankle will allow simple expression of the rotation of the whole body relative to it [7, 35]. Of course, there may be countermotions in other joints, such as the hips.
The actual motion of the joints is not purely rotational. The geometry of some joints is in fact quite complex (for example, consider shoulders and knees) and may be only loosely approximated by pure rotations. Most human factors simulations appear to ignore these differences, though they are important for other biomechanical, medical, and prosthetic applications. We shall assume pure rotations, however, since the impact on human factors assessments appears to be reasonably small.
Where it matters most is in the shoulder and this can be managed by using a clavicle joint to model the shoulder lift as the arm is raised .
For the most part, the segments are considered rigid. The one major exception is the spine which may be approximated by multiple segments or, better, by a spatial curve . The difficulty with the former case is the coordination of the individual segments given desired global motions; for example, it is hard to prevent "jittering" of adjacent segments of a vertebral model . The advantage to the latter is that the curve may be controlled by natural tangent vectors at each end. The length of the curve is tricky to control, but for limited torso motions it can be assumed relatively conect .
Without the curved spine the body shape is clearly unacceptable for extreme sideways or forward positions.
Body movements are restricted by joint limits and the presence of the rest of the body. While joint limits for simple hinges (one degree-of-freedom, such as the elbow) are reasonably easy to assess, limits for complex spherical joints are more difficult. Korein  used spherical polygons to describe the possible configurations of the distal segment to a joint. In general, accurate joint motion and joint limit models are difficult to obtain since joint geometry is never exactly spherical and limits are somewhat dependent on the individual. Body models often have a comfortable and an extreme value for the joint limits. Again, hard data is difficult to obtain.
The problem of avoiding collisions with the body itself is more work. Joint limits describe the relationship between adjacent segments, but for non-adjacent segments other comparisons must be used. Geometric comparison of the segment shape is required. With the use of bounding boxes, the computation cost may be kept to a minimum . On the other hand, one can sometimes ask the user to visually assess the viability of a pose. Using the front and back clipping planes to delimit a "slice" of the scene through the figure and the object surfaces in question, the view can be established approximately parallel to the object surface to visually check collision or contact relationships. This scheme obviously benefits from fast hardware.
One of the foremost tasks of a human figure modeling system is the determination of the reachable space of the figure in some pose. This space has been determined empirically for particular individual s and populations and synthetically from joint limits . Given the reach space, the objects in and out of the space may be simply dete1mined by distance measurement or point-in-polyhedron tests.
All the anthropometry data is affected by external factors such as various suits or clothing.