Demand for technology to support gait training after neurological injury is increasing due to population aging. Due to recent advances in sensing, actuation, and computation, robots are ideal tools to deliver gait training, but their potential in gait neurorehabilitation has not yet been fully realized. In this context, crucial difficulties are identified in the employed control schemes, which are required to accommodate inter-individual gait variations, while promoting stable and energetically efficient gait patterns. The proposed project combines experiments with a lower limb exoskeleton with biomechanical modeling to determine subject-specific assistance strategies that enable a new approach to robot-aided gait neurorehabilitation, named GOALL (Goal-Oriented, subject Adaptive, robot-assisted Locomotor Learning). The conducted research activities have relevant applications both in rehabilitation and in human augmentation, while improving our basic understanding of gait biomechanics. The planned education and outreach components will be targeted to engage a community of graduate, undergraduate and K-12 students in topics at the intersection of robotics and biomechanics. The dissemination of the research methods and results in an open source format will benefit the robotics and biomechanics communities.
The proposed project formalizes new control methods to modulate discrete kinematic variables of gait, achieving controllability of such variables without fully constraining the gait cycle kinematics, thus promoting inter-individual variability in gait kinematics. To this aim, we pursue a systematic approach to the design of gait assistance primitives, i.e. multi-joint coordination patterns capable of modulating a chosen gait parameter, at different gait speeds. The proposed approach is based on inverse dynamics and pulsed torque approximation, and is followed by human-in-the-loop experiments to test the efficacy of assistance primitives to modulate a selected gait parameter during motor adaptation. The experimental investigation is paralleled by neuromechanical modeling of the response to robotic intervention, with the ultimate goal of generalizing the results to other gait parameters of interest for various patient populations.
