CPS Large: Cybernetic Interfaces for the Restoration of Human Movement through Functional Electrical Stimulation

Abstract:

Functional electrical stimulation (FES) is a promising technology for activating muscles in spinal cord injured (SCI) patients. The objective of our project has been to develop an intuitive user interface and control system for FES that allows high-level tetraplegic patients to regain the use of their own arm. This work has had two primary outcomes: contributions to the development of a technology that benefits those with high-level SCI, and the development of biologically-inspired design principles for cyber-physical systems. The project has had two main components: decoder development for determining how the subjects wish to move their arm, and controller development for getting the arm to the desired location. We have been using human and animal models for each of these project components, allowing us to investigate both the practical issues relevant to our current human subjects, and longer term questions dependent on the development of more robust cyber-physical interfaces for FES control. This poster summarizes progress during the course of the project with an emphasis on the past year. During the past year, we have focused on the development of our control architecture. Due to the complexity of the human neuromuscular system and the individual characteristics of each patient, we implemented a semi-parametric framework for system identification, modeling and control that allows us to efficiently obtain models usable for control. We have augmented this approach with a supervisory controller that allows us to regulate limb stability in addition to motions and forces. This is particularly important for an FES limb, which may be under actuated. Having completed the design of our human decoder, we also expanded our explorations into cortical control, demonstrating that we could estimate not only intended reach directions, but also whether subjects were intending to reach or stabilize posture. This allowed us to create a mixture model of reaching intent with improved performance.