A Statistical Method for Predicting Automobile Driving Posture
Reed, Matthew P., Manary, Miriam A., Flannagan, Carol A. C., Schneider, Lawrence W., Human Factors
The design of passenger car interiors is now commonly facilitated by the use of three-dimensional (3-D) human representations that can be manipulated in a computer environment (Porter, Case, Freer, & Bonney, 1993). These computer-aided-design (CAD) human models have increased in sophistication in recent years with advances in computer hardware and software, but their effective use is hampered by the lack of valid methods to set the posture of the models in the simulated vehicle interior.
In the mid-1950s, Dempster (1955) introduced an approach to ergonomic assessment for seated vehicle occupants using an articulated, two-dimensional (2-D) template. A similar template design and a weighted 3-D manikin for measurements in actual vehicles were standardized in the mid-1960s for passenger car interior design by the Society of Automotive Engineers in Recommended Practice J826. In this text SAE Jn refers to a Society of Automotive Engineers (SAE) Recommended Practice, published in the Automotive Engineering Handbook (SAE, 2001). These two tools, the 2-D template and the 3-D H-point machine, are still widely used for designing vehicle interiors but are supplemented by statistical tools that predict the distributions of particular posture characteristics for the U.S. population. These task-oriented percentile models, based on posture data from a number of studies, are available for driver-selected seat position (SAE J15 17), eye position (J94 1), reach (1287), and head location (J1052). See Roe (1993) for a thorough review of the use of these tools in contemporary occupant packaging.
Although the existing task-oriented percentile models are very useful for vehicle design, they are not directly applicable to the posturing of human figure models because they address the population distribution of particular posture characteristics, rather than predict the posture for any particular anthropometric category. For example, the SAE "eyellipse" provides a prediction of the mean and distribution of driver eye locations but does not predict the eye location for women 1550 mm tall or men 1800 mm tall. This more detailed information is necessary to establish an accurate posture for a particular instance of a CAD human model, which necessarily represents a single set of anthropometric variable values.
As computer technology has developed, CAD models have been created to simulate the 2- and 3-D physical manikins, supplemented by more complete 3-D human representations. Porter et al. (1993) briefly reviewed the features of 13 human-modeling systems in use prior to 1993 with potential application to vehicle design. Software development moves rapidly, however, and the two models most widely used for vehicle interior design at the time of this writing--Jack (EDS-PLM Solutions, Cypress, CA) and Tecmath's RAMSIS (Human Solutions, GmbH, Kaiserslautern, Germany)--were not mentioned in the Porter et al. review.
One impediment to more widespread use of human models for vehicle design has been a lack of valid posture prediction. Without posture-prediction capability built into the model or available through other external data or statistical models, many of the most useful applications of the CAD human models are unreliable. For example, vision and reach assessments require an accurate starting posture for the particular manikin dimensions being used. In the absence of accurate posture prediction, CAD human models are valuable primarily for visualization rather than for ergonomic assessment.
Few published studies are applicable to posture prediction for vehicle occupants. In many early studies data were presented only in the aggregate or in terms of a population distribution, so the findings are not applicable to human model posture (Hammond & Roe, 1972; Meldrum, 1965; Phillipart, Roe, Arnold, & Kuechenmeister, 1984). Seidl (1994) presented the most complete approach to whole-body driving posture prediction to date. …