Toward Muscle-Driven Control of Wearable Robots: A Real-Time Framework for the Estimation of Neuromuscular States During Human-Exoskeleton Locomotion Tasks

Abstract

The ability to efficiently assist human movement via wearable robotic exoskeletons requires a deep understanding of human-exoskeleton physical interaction. That is, how the exoskeleton affects human movement and how the human body reacts to robotic assistance. In this context, it is central to gain access to human neuromuscular states, i.e. neural activation to muscle, muscle fibers short-stretch cycle, tendon strain, musculotendon viscoelasticity. This would enable the personalized design of assistive devices and human-exoskeleton interfaces with respect to a specific subject's anatomy and force-generating capacity. Here we present a real-time electromyography-driven framework interfaced to a robotic bilateral ankle exoskeleton. This framework provides real-time information about joint and underlying muscle mechanics. We provide a quantitative evaluation of the real-time framework across a repertoire of human-exoskeleton locomotion tasks. We also show how this enables understanding how robotic exoskeletons in parallel to human limbs contribute to alter normative musculoskeletal mechanics. This will open new avenues for the creation of symbiotic exoskeleton technologies that operate as an extension of the own body.