Sub-optimally Solving Actuator Redundancy in a Hybrid Neuroprosthetic System with a Multi-layer Neural Network...

Area of Science:

• Biomedical Engineering
• Neuroscience
• Robotics

Background:

• Functional electrical stimulation (FES) is explored as a supplementary torque assist for lower-limb powered exoskeletons in individuals with paraplegia.
• Hybrid neuroprostheses integrate FES-assist and exoskeleton torques for standing and walking.
• Actuator redundancy necessitates optimal allocation of FES and exoskeleton torque to balance muscle fatigue and tracking errors.

Purpose of the Study:

• To develop and validate a novel control strategy for hybrid neuroprostheses that optimally allocates FES and exoskeleton torque.
• To address the challenge of accurately modeling musculoskeletal responses to FES for traditional optimal control.
• To minimize muscle fatigue and tracking errors in FES-assisted exoskeleton systems.

Main Methods:

• A novel identification and control structure utilizing a recurrent neural network (RNN) for system dynamics identification and feedforward neural networks (FNNs) for control.
• Supervised learning trains the RNN, while reinforcement learning trains the FNNs for sub-optimal control actions.
• FNNs incorporate unique output layer activation functions to manage asymmetric FES constraints and symmetric exoskeleton motor control inputs.

Main Results:

• The RNN successfully identified system dynamics, enabling effective control.
• The FNNs provided sub-optimal control actions, demonstrating the feasibility of the reinforcement learning approach.
• Experimental validation on a seated participant with a single-joint hybrid neuroprosthesis confirmed the system's efficacy.

Conclusions:

• The proposed RNN and FNN-based control structure offers a viable solution for optimal torque allocation in hybrid neuroprostheses.
• This approach effectively addresses the challenges of modeling complex musculoskeletal dynamics and actuator redundancy.
• The validated system shows promise for enhancing mobility and reducing muscle fatigue in individuals with paraplegia using FES-assisted exoskeletons.