両足首エクソスケルトンを用いた歩行性能推定のためのリアルタイム制御器の開発

This arm employs a within-subject design with two methods of estimating metabolic cost versus the gold standard measure of metabolic cost, wherein a single participant is subjected to two distinct measurements. This design allows for a direct comparison of the effects of each method (i.e., estimation versus gold standard) within the same individual, minimizing intersubject variability and enhancing the statistical power of the analysis.

This intervention uses a robotic ankle exoskeleton equipped with a real-time adaptive controller that adjusts plantarflexion torque based on each participant's walking mechanics. Unlike standard exoskeleton controllers that use fixed or pre-programmed assistance levels, this system employs human-in-the-loop optimization to continuously update torque magnitude and timing during treadmill walking. The controller integrates metabolic estimations, kinematic data, and musculoskeletal modeling to identify individualized assistance patterns that reduce walking effort and improve muscle activation efficiency. Participants complete multiple walking trials while the controller automatically modifies assistance to determine the optimal personalized settings.

Device feasibility will be evaluated by the successful real-time operation of the robotic ankle exoskeleton and adaptive controller during treadmill walking. Feasibility is defined as the controller's ability to continuously generate, update, and apply assistive torque in real time based on incoming biomechanical and physiological data without system failure, interruption, or safety-related termination. Successful operation will be confirmed by continuous controller function and synchronized data acquisition across walking trials.

Net oxygen consumption (VO₂) and carbon dioxide production (VCO₂) will be measured during treadmill walking using indirect calorimetry (Cosmed K5, Cosmed USA Inc., Chicago, IL). Metabolic rate will be calculated using standard equations during steady-state walking conditions. Measurements will be collected at regular intervals to characterize metabolic demand under different exoskeleton assistance configurations.

Estimated metabolic rate will be derived from joint-space musculoskeletal models using kinematic and kinetic data collected during treadmill walking. Model-based estimates will be computed on a stride-by-stride basis to provide an indirect estimate of metabolic demand that can be compared with direct measurements from indirect calorimetry.

Lower-limb muscle activation patterns will be estimated using joint-space musculoskeletal models based on motion capture and ground reaction force data collected during treadmill walking. Estimated muscle activation values will be computed on a stride-by-stride basis to characterize neuromuscular engagement during exoskeleton-assisted gait.

Muscle activation of lower-limb muscles (e.g., tibialis anterior, gastrocnemius medialis, gastrocnemius lateralis, soleus) will be measured during treadmill walking using surface electromyography (Delsys). EMG signals will be collected continuously and processed to quantify muscle activation patterns during exoskeleton-assisted gait.

Controller parameter convergence will be assessed during human-in-the-loop optimization trials by evaluating changes in controller gain and timing parameters across successive walking bouts. Convergence is defined as stabilization of controller parameters within a predefined range during the optimization process.