Human sensorimotor control is extremely complex. Because of this, it remains an open question how best to control lower-limb devices, such as exoskeletons and robotic prostheses, that seek to assist in human movement. A common approach is to attempt to replicate average biological joint torques, but this does not incorporate individualized differences in movement or the sensorimotor control loop that allows humans to adjust motor commands in response to sensory feedback. This dissertation presents three different approaches to device control, as well as a novel form of sensory feedback, with each successive approach providing greater incorporation of user input.

First, we develop a passive assistive device for running that successfully reduces human energetic cost. Previous approaches to develop assistive devices for running have focused on the more costly stance phase, which requires more massive devices providing high torque output. We instead focus on the swing phase of running, which only comprises a small component of the total cost of running but can be more easily assisted. We develop a lightweight, passive assistive device, validate its ecacy in reducing metabolic cost in a user study, and characterize human biomechanical gait adaptation that allows users to take advantage of the device.

Second, we present an approach for customizing the control parameters of robotic prostheses for individuals with lower-limb amputation. Prior work has shown success using a technique called human-in-the-loop optimization to choose exoskeleton parameters that best reduce energetic cost, but this technique had not been examined in robotic prosthetic devices. We develop four different parameterized control strategies and optimize parameters of an ankle-foot prostheses in case studies for users with transtibial amputation. Although some participants were able to reduce their energetic cost with optimized compared to generic control parameters, the success of human-in-the-loop optimization was more more limited in prostheses than past work with exoskeletons.

One key difference between individuals with amputations using robotic prostheses and able-bodied users of robotic exoskeletons is the quality and type of sensory feedback from the device being optimized, with individuals with amputation receiving only limited sensory feedback from interactions at the socket and whole-body proprioception of intact joint positions. To augment this sensory feedback, we build and characterize a wrist exoskeleton capable of presenting continuous wrist flexion and extension torques, as well as sensing user wrist angle. Using the wrist exoskeleton, we perform a user study characterizing how human wrist torque perception differs while seated compared to walking, as well as moving the wrist compared to keeping it still. We also perform simulations demonstrating the effect of applied wrist torque error on resulting perception data to ensure that the performance of the wrist exoskeleton does not significantly bias our results.

Finally, we present a system that uses the wrist exoskeleton to closes the sensorimotor control loop of robotic prostheses, in which an individual with amputation teleoperates their own prosthetic ankle and receives sensory feedback regarding its behavior. We develop two control schemes that give the user full control or partial control of a semi-autonomous robotic prosthesis. We perform static and dynamic benchtop testing of all system components to ensure sucient accuracy. We also demonstrate proof-of-concept of the system by conducting a pilot study with one participant with transtibial amputation and show that after only one day of training, the participant was able to use the wrist exoskeleton to control desired trajectories within human wrist perception error after just one day of training. This novel system design can be used in the future to address scientific questions regarding human motor control, as well as test translational outcomes for people with amputation.