Real-Time Gait Phase Estimation and Speed-Adaptive Trajectory Generation for a Robotic Transfemoral Prosthesis

This study proposes an effective method for generating gait trajectories for robotic transfemoral prostheses used by impaired persons with above-knee amputations. The gait phase variable was robustly estimated in real-time using data from an Inertial Measurement Unit (IMU) attached to the affected thigh. A linear Kalman filter based on a Simple Harmonic Oscillator (SHO) model was used, which inherently avoids phase saturation and minimizes nonlinear distortions. This ensures smooth phase progression across varying walking speeds. Additionally, a Bounded Periodic Trajectory Generator (BPTG) was introduced to generate periodic joint trajectories constrained within specific angular ranges. This was achieved using an Artificial Neural Network (ANN) model trained on human gait data. The proposed estimation algorithm for the gait phase variable was evaluated through treadmill walking experiments involving four healthy subjects, where all four subjects assumed their right leg to be the impaired limb. The performance of BPTG was successfully evaluated through experiments with a subject walking on a treadmill while wearing a robotic transfemoral prosthesis attached to the right leg, which was also assumed to be the impaired limb.