A Bionic Foot Controlled by a Synergy-Driven Neuromechanical Model Enables Walking at Various Speeds in Socket-Suspended and Bone-Anchored Prosthesis Users.

Human locomotion adapts to different conditions, resulting in changes in gait parameters like speed, stride time, and length. Bionic limbs strive to mimic natural walking patterns, with speed adaptation being a key feature. Research shows that myoelectric bionic legs allow individuals with agonist-antagonist myoneural interface (AMI) amputations to control speed-adaptive walking. However, those with non-AMI amputations show difficulty generating consistent electromyography (EMG) signals. Therefore, we aim to create a human-machine interface that provides speed-adaptive biomimetic behavior without relying on EMGs. Steady-state locomotion can be modeled as the sequential recruitment of muscle groups during the gait cycle. To replicate this motor control, we created a control framework for a bionic foot using a neuromechanical model driven by synthetic muscle activations, replacing EMG recordings. We tested the controller on two individuals with transtibial amputations-one with a socket-suspended prosthesis and the other with a bone-anchored prosthesis. Muscle activation peaks fell within target ranges, leading to peak plantar-flexion torques at 49% of the gait cycle. The averaged model torques aligned with those from inverse dynamics on the intact side (RMSE $=$ 0.52 $ ~\pm ~$ 0.3 (Nm/Kg), r $=$ 0.52 $~\pm ~$ 0.4). The results show that the control system effectively modulates joint torques in timing and amplitude for two subjects across three walking speeds (0.55 to 1.1 m/s). Designed for steady-state walking, it can modulate torque during speed transitions. This first investigation aims to prove the feasibility of a personalized biomimetic control framework for bionic limbs without relying on EMGs, supporting walking at various speeds.