A novel approach to predicting human ingress motion using an artificial neural network

Area of Science:

• Ergonomics and Human Factors
• Computer Science
• Biomechanical Engineering

Background:

• Digital human models are increasingly available, highlighting the need for accurate human movement data in product design.
• Current methods for acquiring human motion data are time-consuming and often yield unrealistic results, hindering efficient product development.
• Optimizing product design, such as automobile door panels for ease of ingress/egress, requires rapid prediction of human motion.

Purpose of the Study:

• To propose a novel algorithm for rapid and accurate prediction of full-body human motion.
• To enhance the efficiency and accuracy of product design processes utilizing digital human models.
• To overcome limitations in existing optimization-based research by addressing the ambiguity of human motion objective functions.

Main Methods:

• Construction of a motion capture database using an optical motion capture system across diverse environments.
• Application of a generalized regression neural network (GRNN) to generate full-body human motion from the database.
• Algorithm design inspired by human motor learning principles, where new motions are generated based on past experiences.

Main Results:

• The proposed algorithm successfully predicts full-body human motion rapidly and accurately.
• Generated motions demonstrate statistical similarity to actual human movements.
• The method effectively handles variations in design parameters and environmental variables.

Conclusions:

• The research provides a foundational algorithm for rapid human motion prediction adaptable to various environmental factors.
• This approach significantly increases the utility of digital human models in product design and related fields.
• The algorithm offers a viable solution to the inefficiencies and inaccuracies associated with traditional motion data acquisition methods.