Probabilistic Model-Based Learning Control of a Soft Pneumatic Glove for Hand Rehabilitation

Area of Science:

• Rehabilitation Engineering
• Robotics
• Neuroscience

Background:

• Stroke survivors often experience significant hand dysfunction, limiting their ability to perform activities of daily living (ADL).
• Soft pneumatic gloves offer a potential solution for assisting stroke patients, but effective control strategies for human-robot interaction remain a challenge.
• Challenges include the inherent nonlinearities of soft robots and the difficulty in interpreting human intentions during rehabilitation.

Purpose of the Study:

• To develop advanced control approaches for a soft robotic glove system to enhance hand function in stroke survivors.
• To create an effective human-soft robot integrated system for post-stroke rehabilitation.
• To improve the independence of stroke survivors in performing ADL.

Main Methods:

• A soft pneumatic glove was employed to assist individuals with stroke-impaired hands.
• A probabilistic model-based learning control strategy was implemented to address system complexities.
• A training modality was designed to be task-oriented and intention-driven, with evaluation on able-bodied and stroke survivor participants.

Main Results:

• The soft robotic glove provided adaptive assistance, enabling participants to perform various tasks.
• Significant improvements were observed, including decreasing tracking error and muscle co-contraction, alongside an increasing hand gesture index over training sessions.
• Stroke survivors demonstrated enhanced hand functions and improved muscle coordination post-training.

Conclusions:

• A novel learning-based soft robotic glove training system was developed for post-stroke hand rehabilitation.
• The system shows significant potential for improving hand function and promoting the application of soft robotic solutions in stroke recovery.