Predictive Gait Generation Based on Passive Morphology: Kinematic and Kinetic Optimization

This study investigates whether the inherent morphology of a passive dynamic walking system can accurately reproduce human gait kinematics and kinetics and assesses its potential for predictive gait generation.

A multistage optimization algorithm was implemented to tune monoarticular and biarticular elastic and viscoelastic components representing the quasi-stiffness of lower-limb muscles. The resulting model was evaluated against an open source gait dataset from six healthy adults with comparable body mass and walking speeds. Resemblance was quantified using the coefficient of determination (\boldmath $R^{2}$) for joint kinematics and kinetics at three gait speeds of approximately 0.83 m/s, 0.96 m/s, and 1.12 m/s corresponding to the lower, intermediate, and upper stable slope angle limits.

Purely passive gaits replicated human hip and knee joint angles with average \boldmath $R^{2}$ of 0.98 and 0.99 respectively. When optimized for joint moments, the model maintained strong kinematic agreement and produced ankle and hip joint moments with average $R^{2}$ of 0.94 and 0.60 across corresponding speeds. Ankle joint angle and knee joint moment, however, had room for improvement with average \boldmath $R^{2}$ of 0.45 and 0.01 respectively.

A passive dynamic walking model with a novel optimization algorithm and elastic elements replicating lower-limb muscles, can reproduce human gait kinematics and kinetics with high accuracy, demonstrating that morphology alone can account for essential features of locomotor dynamics.

This work highlights the role of morphology in human locomotion and introduces a simple, accessible framework for predictive gait generation to evaluate the expected biomechanical impact of assistive and rehabilitative devices.