Collaborative Research: An Integrated, Proactive, and Ubiquitous Prosthetic Care Robot for People with Lower Limb Amputation: Sensing, Device Designing, and Control

Individuals with lower limb amputation face multiple disadvantages while walking, such as increased strains on the lower back, increased metabolic cost, and gait asymmetry. Existing prosthetic robot prescriptions suffer from several deficiencies: they either cannot provide accurate and prompt device control or can only be used under a constrained laboratory environment. The key challenge for lower limb amputees is that there is no longer any feedback control to the device from the visual information obtained through the eyes. As a result, the rest of the body requires extra musculature recruitment to compensate for the increased instability, posing a greater risk of falling and negative joint issues. This award aims to develop a novel prosthetic robot that would use the reconstruction of the neuromuscular feedback from the visual information to provide proactive and optimal control in different walking conditions. The outcome of this project will help clinicians provide enhanced care when prescribing lower limb prostheses and positively affect the lives of amputees living within the United States. In addition, this project will create a multidisciplinary education program comprising knowledge of design and control, human-machine interface, sensors, and machine learning for undergraduate and graduate students from engineering, computer science, and medicine. Students with disabilities will also participate through data collection, meetings, and mentorship. The objective of this project is to develop a new prosthetic robot that enables proactive and user-specific prosthetic control to improve walking function in a variety of conditions found in daily life. The approach relies on interrelated advances in sensing strategy, device hardware, and control framework as follows. 1) A motion capture method using only wearable inertial motion unit (IMU) sensors will be developed to reconstruct and predict human gait in daily living, while enabling assessment and clinical diagnosis beyond lab settings. This will be achieved through a knowledge distillation method leveraging a complex teacher network using multiple sensing modalities to train a simpler student network that only uses IMU sensors. 2) A lightweight, energy-efficient, and semi-active pneumatic and hydraulic hybrid device will be developed to enable the user to tune settings as per the operating environment. 3) A reinforcement-learning-based adaptive algorithm will be developed to control and sync user-device cooperation. In addition, the mixed-criticality design paradigm will be borrowed from the real-time systems field to ensure the safety and stability of the robot. This project is supported by the cross-directorate Foundational Research in Robotics program, jointly managed and funded by the Directorates for Engineering (ENG) and Computer and Information Science and Engineering (CISE). 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.