Posture prediction models in digital human modeling for ergonomic design: A systematic review

Abstract

Posture prediction models have been widely used to support ergonomic design. This systematic review critically assessed the development, validation, and applications of posture prediction models in Digital Human Modeling (DHM). Following PRISMA guidelines, 24 studies were included from a search across nine academic databases, categorized into data-driven models (n = 12) and optimization-based models (n = 12). Data-driven models, particularly those employing neural network regression and artificial neural networks, demonstrated strong predictive accuracy and adaptability, but often lacked generalizability due to data imbalance and limited participant/task diversity. Optimization-based models, using algorithms such as gradient descent and genetic algorithms, showed high biomechanical fidelity but computational challenges and limited computer-aided design (CAD) integration. While a few models have been integrated with existing CAD software such as JACK and Santos™, most lacked ergonomic evaluation and real-time usability. Limitations identified include insufficient diverse datasets, computational inefficiencies, and limited validation in real-world conditions. Future research should prioritize model development supported by scalable motion data using computer vision-based technologies and hybrid strategies that combine learning-based inference with biomechanical simulation, offering a promising path toward achieving both accuracy and physiological realism in posture prediction.