AI enhanced collaborative human-machine interactions for home-based telerehabilitation.

Area of Science:

• Robotics
• Rehabilitation Engineering
• Computer Science

Background:

• Telerehabilitation offers on-demand therapy, reducing patient travel time and costs.
• System robustness is critical, as network issues like latency and jitter can degrade user interaction quality.
• Adapting robotic systems to user behavior is key for effective remote rehabilitation.

Purpose of the Study:

• To propose a data loss compensation method for robotic telerehabilitation systems.
• To enhance the robustness and quality of user-system interaction despite network instability.
• To train a robotic system to adapt to user behavior using collected task data.

Main Methods:

• Utilized a virtual reality (VR) environment for data collection during a collaborative task.
• Employed nonlinear autoregressive models with exogenous input (NARX) and long-short term memory (LSTM) neural networks.
• Trained a robotic system to predict and smooth user movements, compensating for data loss.

Main Results:

• The proposed approach effectively compensates for data loss, maintaining interaction quality.
• Long-short term memory (LSTM) networks demonstrated an ability to learn human-like interaction patterns.
• The trained predictor completed the task nearly as fast as human execution (25s vs. 23s).

Conclusions:

• The developed data loss compensation method enhances robotic telerehabilitation system performance.
• Machine learning, particularly LSTM, shows promise in creating adaptive and human-like robotic rehabilitation systems.
• This approach facilitates more reliable and effective remote physical therapy delivery.