Human locomotor control, adaptation, and rehabilitation

Motor skill acquisition is a necessity for any person who desires to improve their ability to interact with the world. A fundamental process in motor skill acquisition is the formation of internal models that represent the dynamic environments we interact with. Internal models are neural representations relating the spatio-temporal patterns of muscle activation to desired limb trajectories.The goal of this dissertation is to develop robotic training algorithms that aid in motor skill acquisition, based on a neuro-computational understanding of the internal model formation process.The specific task studied was adaptation to a robot-applied force field during treadmill walking. Studying motor system adaptation to novel locomotor perturbations is directly relevant to gait rehabilitation following neurological injury. The paradigm captures the essence of movement tasks in which the motor system must learn to compensate for a dynamic perturbation, and thus the results will likely generalize to other movement and perturbation paradigms.To study locomotor adaptation, we developed a robot capable of measuring and manipulating leg movement. This work quantified the robot's performance capabilities. Using this robot, we confirmed that humans utilize internal model formation when adapting to novel dynamic environments during locomotion and that this process can be modeled computationally using a difference equation that can be viewed as arising from an optimization of kinematic error and effort. Within this computational framework, we then designed and experimentally tested two novel robotic training algorithms. The first uses a robotic environment to accelerate the learning process by amplifying the trajectory errors that drive motor adaptation. The second algorithm uses a robotic trainer to limit performance errors while systematically reducing assistance. Finally, we tested the feasibility of two, related, novel control algorithms that represent a step toward applying this technology and control algorithms in patients with spinal cord injury.Thus, the contributions of this dissertation are: (1) robotic technology and mathematical models for understanding the neural computations underlying internal model formation during locomotor skill acquisition (2) two novel, computationally based, robotic training algorithms and (3) a pilot study that applies this technology and algorithms to retraining patients with spinal cord injury to walk.