Performance-Based Classification of Occupant Posture to Reduce the Risk of Injury in a Collision

This study numerically investigates the development of an adaptive restraint system based on precrash classification of occupant posture. A catalog of restraint laws optimized for nine postures uniformly distributed in posture space is employed. First, the performance of each restraint law is globally assessed by performing crash simulations in a parametric fashion throughout the entire posture space. Then, restraint systems with catalogs (RSCs) with various numbers of restraint laws are evaluated in terms of injury cost with respect to a restraint system optimized with respect to a nominal posture (RSN). Parametric and nonparametric supervised classifiers are developed for each catalog, and their performances are analyzed. A catalog with the optimized laws of two out-of-position postures (central and leaning left) showed high performance in terms of reduced injury cost with respect to optimum performance for two distinct validation sets (25.3%/21.6% with statistical classifiers versus 26%/23.8% optimum performance). The percent injury reduction increased as the number of classes was increased but had diminishing returns going from five to nine restraint laws (28%/24.2% with statistical classifiers versus 30.4%/29.1% optimum reduction). The results of this study indicated that restraint systems with performance-based classes perform better than restraint systems with region-based classes. Expanding the number of restraint laws and developing new classification algorithms may further improve the performance of adaptive restraint systems.