Predicting Individualized Joint Kinematics Over Continuous Variations of Walking, Running, and Stair Climbing

Area of Science:

• Biomechanics
• Robotics
• Human-Computer Interaction

Background:

• Personalized gait adaptation is essential for effective wearable robot assistance.
• Manual tuning of multi-activity gait models is clinically impractical due to time constraints.

Purpose of the Study:

• To investigate the generalizability of kinematic gait individuality across different ambulation modes and tasks.
• To determine if a single-task prediction of gait individuality is effective within a mode or across modes.

Main Methods:

• Quantified kinematic individuality for each subject, joint, and ambulation mode (walking, running, stair ascent/descent) using an open-access dataset.
• Employed N-way ANOVAs to assess prediction method efficacy.
• Tested cross-modal generalization of walking individuality versus mode-specific predictions.

Main Results:

• Kinematic individualization significantly improved model fit across joints and tasks when modes were considered separately.
• Modal individualization enhanced fit in 81% of trials by an average of 4.3% across the gait cycle.
• Statistical significance for improved fit was observed across all joints for walking and running, and for half the joints during stair negotiation.

Conclusions:

• Kinematic gait individuality is effectively generalizable within walking and running modes.
• Generalization of gait individuality on stairs is joint-dependent and shows mixed trends.
• Mode-specific prediction of gait individuality offers significant improvements, particularly for walking and running.