Computational modeling of neuromuscular response to swing-phase robotic knee extension assistance in cerebral palsy.

Predicting subject-specific responses to exoskeleton assistance may aid in maximizing functional gait outcomes, such as achieving full knee-extension at foot contact in individuals with crouch gait from cerebral palsy (CP). The purpose of this study was to investigate the role of volitional and non-volitional muscle activity in subject-specific responses to knee extension assistance during walking with an exoskeleton. We developed a simulation framework to predict responses to exoskeleton torque by applying a stretch-reflex spasticity model with muscle excitations computed during unassisted walking. The framework was validated with data collected from six individuals with CP. Framework-predicted knee angle at terminal swing was within 4 ± 4° (mean ± sd) of the knee angle measured experimentally without the addition of spasticity. Kinematic responses in two-thirds of the participants could be accurately modeled using only underlying muscle activity and the applied exoskeleton torque; incorporating hamstring spasticity was necessary to recreate the measured kinematics to within 1 ± 1° in the remaining participants. We observed strong positive linear relationships between knee extension and exoskeleton assistance, and strong negative quadratic relationships between knee extension and spasticity. We utilized our framework to identify optimal torque profiles necessary to achieve full knee-extension at foot contact. An angular impulse of 0.061 ± 0.025 Nm·s·kg