Human-in-the-Loop Optimization of Knee Exoskeleton Assistance for Minimizing User's Metabolic and Muscular Effort

Area of Science:

• Robotics
• Biomechanics
• Human-Computer Interaction

Background:

• Lower limb exoskeletons can help prevent musculoskeletal disorders but often lack adaptive control.
• User-oriented control strategies are crucial for optimizing exoskeleton assistance in real-time.

Purpose of the Study:

• To develop and evaluate a human-in-the-loop (HITL) control for a knee exoskeleton.
• To minimize user physical effort by reducing interaction torque and metabolic cost.
• To innovate by estimating metabolic cost in real-time using a machine learning model within the HITL control.

Main Methods:

• Implemented a HITL control strategy for a knee exoskeleton utilizing a CMA-ES algorithm.
• Evaluated user physical effort using interaction torque and estimated metabolic cost.
• Developed a machine learning regression model to estimate metabolic cost in real-time.
• Compared HITL control against zero-torque and no-device conditions.

Main Results:

• The machine learning model estimated metabolic cost with a root mean squared error of 0.66 W/kg and a mean absolute percentage error of 26% (n=5), providing faster and less noisy estimations than a respirometer.
• The HITL control reduced user metabolic cost by 7.3% (vs. zero-torque) and 5.9% (vs. no-device).
• Interaction torque was reduced by 32.3% compared to a zero-torque control (n=1).
• User-specific HITL control demonstrated lower metabolic cost than non-user-specific assistance.

Conclusions:

• The developed HITL control effectively reduced physical effort and metabolic cost during assisted walking, surpassing non-exoskeleton and zero-torque conditions.
• Real-time metabolic cost estimation via machine learning is feasible and beneficial for HITL exoskeleton control.
• This proof-of-concept highlights the potential of adaptive HITL controls for enhancing exoskeleton-assisted locomotion.