Predicting Human Upper Extremity Reaching Motions: Comparison of Optimization-Based Method and Heuristic Method

Predicting human upper extremity reaching motion in 3D space can support adaptive interactions with computer-controlled systems (robots and virtual avatars), and applications in ergonomics and rehabilitation. This study compares two predictive approaches: an optimization-based method (OPM) and a proposed heuristic method (SFM) that integrates steering dynamics path planning, an adaptive velocity model, and inverse kinematics. Both methods were validated against motion capture data from ten participants performing four reach tasks. Predictions and inter-subject variability were evaluated for path, velocity, and upper extremity joint configuration using root mean square error and dynamic time warping. Results show that SFM more accurately predicts spatial path and velocity, whereas OPM achieves greater precision in joint angle estimation. As input, OPM requires the initial and end posturesand the task duration, while SFM needs the initial posture, initial and target end-effector positions, and initial and estimated peak velocity. These results highlight trade-offs between accuracy and behavioral variability when selecting motion prediction methods.