NSF Award
2205205
This Smart Health (SCH) award will contribute to the advancement of the national health and welfare by developing technology to help the millions of people who have suffered an amputation or who have a congenital deficiency of an upper limb. These problems create a devastating impact on the quality of the victims’ lives and pose a significant financial burden to people in the U.S.A. Unfortunately, the state of the art in prosthetic hand replacements is not comparable to their popular portrayal in sci-fi movies, leaving upper limb-absent people with rudimentary control of basic grasp functions. This work will explore methods to help upper limb-absent people learn advanced control of sophisticated prosthetic hands with an automated training regimen that can be used at home. Automating this aspect of healthcare with remote learning functionality can help disabled people access treatment more quickly and at a lower cost. Furthermore, research from this grant will be used to create learning experiences for high school students from low-income households to help educate the next generation of engineers and scientists. Thus, this research can benefit the society and economy of the USA. <br/><br/>Currently, there are numerous options for dexterous wearable co-robot assistants, such as prosthetic hands, with future devices becoming even more sophisticated. In fact, the dexterity of these devices is rapidly outpacing the ability for people to intuitively control them. One main source of this problem stems from the inability to reliably interpret the intentions of the human operator over the course of months and years. Another problem is that the science behind customizable training programs to empower disabled people to harness the full potential of dexterous assistive robots such as prosthetic hands has not been deeply explored. This project will address these deficiencies through three main endeavors: (1) exploration of a novel bimodal skin sensor to overcome limitations with current sensors, (2) investigation of the utility of machine learning algorithms to classify the intention of human operators, and (3) creation of a new reinforcement learning paradigm to sequentially introduce customized, subject-specific muscle training exercises over time to maximize biocontrol signal classification accuracy and the number of independent biocontrol channels.<br/><br/>This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.
