Human-Machine Co-Adaptation for Robot-Assisted Rehabilitation via Dual-Agent Multiple Model Reinforcement Learning...

Area of Science:

• Robotics
• Rehabilitation Engineering
• Artificial Intelligence

Background:

• Modeling human behavior in robot-assisted rehabilitation is challenging due to complex human cognition and physiology.
• Single-model approaches struggle to capture human-machine interaction dynamics.
• Adaptive, patient-centered interventions are crucial for effective rehabilitation.

Purpose of the Study:

• To introduce a novel Dual-Agent Multiple Model Reinforcement Learning (DAMMRL) framework for robot-assisted ankle rehabilitation.
• To address the limitations of traditional single-model approaches in capturing human-machine dynamics.
• To develop a versatile and adaptive system for personalized patient therapy.

Main Methods:

• Utilized a multiple model strategy with simple submodels to approximate complex human responses.
• Integrated Multiple Model Adaptive Control (MMAC) and co-adaptive control strategies.
• Demonstrated system versatility in both simulated and real experimental environments.

Main Results:

• Evaluated feasibility and potential with 13 healthy subjects and nine patients with lower-limb motor disorders.
• Achieved promising results, affirming the anticipated benefits of the DAMMRL approach.
• Showcased the system's ability to adapt to varying levels of patient incapacity.

Conclusions:

• The DAMMRL framework offers a new paradigm for robot-assisted ankle rehabilitation.
• The study highlights the potential for adaptive, patient-centered therapeutic interventions.
• This research paves the way for future advancements in intelligent rehabilitation systems.