Neuromechanical adaptation to robotic exoskeletons during human locomotion.

Abstract

Humans are highly adept at modifying muscle activation patterns within a step to walk at various speeds on different terrains. On longer timescales, humans also adapt neuromuscular control over years to accommodate body transformations due to growth and aging. Currently we have little understanding of the processes and mechanisms for human locomotor adaptation during intermediate to long term time scales (minutes to days). In this dissertation I tested a novel tool for studying human locomotor adaptation; a pneumatically powered robotic exoskeleton. I constructed lightweight exoskeletons that allowed free rotation about the ankle joint. Artificial pneumatic muscles attached to the exoskeleton created torque about the ankle. Observing the immediate neuromuscular responses to exoskeleton mechanics can provide insight into the process and control of locomotor adaptation. In the first experiment, I demonstrate that the exoskeleton can generate approximately 60% of the ankle plantar flexor torque and 70% of the positive plantar flexor work performed during normal walking. These results were consistent across walking speeds and with one or two artificial muscles assisting in parallel. In the next experiment, the exoskeleton provided plantar flexion assistance proportional to the subjects' soleus electromyography amplitude, effectively increasing soleus strength. The added mechanical power greatly perturbed ankle joint movements at first, but subjects quickly learned the new system dynamics. Subjects rapidly decreased soleus and gastrocnemius recruitment to walk with a normal gait. When subjects were tested again three days later, they demonstrated a lasting motor memory of exoskeleton dynamics. In the final experiment, the exoskeleton was reconfigured to simulate antagonistic coactivation about the ankle joint. An artificial dorsiflexor muscle produced ankle torque proportional to soleus electromoyography amplitude. In contrast to the previous experiment, rather than walking with normal kinematics, subjects adopted a new gait pattern with almost no ankle push off. After initially fighting the exoskeleton torque, subjects decreased both soleus and gastrocnemius activation. These findings suggest that altering individual muscle patterns within a synergistic group is inherently more difficult then making collective gain adjustments to entire groups of synergistic muscles. Overall, these studies demonstrate that humans form a motor memory of musculoskeletal dynamics for controlling locomotion.