This paper investigates how human walking mechanics can be represented using a small number of sequentially composed multivariable linear controllers. By solving a mixed integer optimization problem, the authors identify four controllers—each corresponding to a distinct phase of the gait cycle—that capture over 96% of the variance in joint torques across subjects. The approach generalizes well to new subjects, offering a compact and accurate representation of the relationship between joint torques, angles, and velocities.
Why it matters: Understanding human walking in terms of a few simple controllers provides a foundation for designing assistive and augmentation devices that coordinate multiple joints. This research supports the development of more effective exoskeletons and prostheses that can adapt across users and walking conditions.