Wearer-Prosthesis Interaction for Symmetrical Gait: A Study Enabled by Reinforcement Learning Prosthesis Control.

Area of Science:

• Biomechanics
• Robotics
• Rehabilitation Engineering

Background:

• Advancements in robotic prostheses aim to enhance amputee gait performance beyond basic restoration.
• Understanding the influence of prosthesis mechanics and control on wearer gait symmetry is crucial but limited.

Purpose of the Study:

• To investigate how robotic transfemoral prosthesis mechanics impact wearer gait symmetry.
• To explore the potential of reinforcement learning (RL) for personalizing prosthesis control.

Main Methods:

• Utilized a previously designed RL supplementary control system to tune 12 prosthesis control parameters.
• Recruited subjects who walked with a robotic transfemoral prosthesis on a treadmill while the RL controller adapted parameters.
• Measured stance time symmetry, step length symmetry, and bilateral anteroposterior (AP) impulses.

Main Results:

• Changes in robotic knee mechanics influenced gait temporal-spatial symmetry measures in both lower limbs.
• Stance time symmetry correlated with net inter-limb AP impulse; step length symmetry correlated with braking/propulsive impulse symmetry.
• Personalizing transfemoral prosthesis control can improve gait symmetry, but requires coordination between wearer and prosthetic control.

Conclusions:

• Robotic prosthesis mechanics significantly influence gait symmetry.
• Achieving maximal gait symmetry necessitates coordination between the wearer's intact limb motor control and adaptive prosthetic joint control.
• The RL-based system shows promise as a tool for studying wearer-prosthesis interactions.