Research on stiffness adaptation control of medical assistive robots based on stiffness prediction

Predicting human stiffness, especially at the distal end of the human arm, holds significant potential for various applications. It facilitates the realization of humanoid stiffness regulation in robots, improves the adaptability and human-likeness of interactive robots, and addresses critical issues in human control of medical assistive robots. Recognizing that surface electromyographic (EMG) signals not only contain rich information but are also easy to collect and process, they serve as an optimal choice for predicting human stiffness. To establish a mapping relationship between surface EMG signals and stiffness information, we constructed a stiffness acquisition system to collect signals such as EMG, angular, force, and displacement signals. Additionally, considering the influence of different angles (configurations) of the human arm on the stiffness at the distal end, we researched a stiffness prediction model for the distal end of the human arm using a multilayer perceptron. Experimental results demonstrate that our proposed stiffness prediction model, utilizing EMG information provided by the EMG armband along with angular information, can predict the stiffness at the distal end of the human arm in various scenarios. This provides ample reference for achieving humanoid stiffness regulation in medical assistive robots.