The Runbot: engineering control applied to rehabilitation in spinal cord injury patients

Human walking is a complicated interaction among the musculoskeletal system, nervous system and the environment. An injury affecting the neurological system, such as a spinal cord injury (SCI) can cause sensor and motor deficits, and can result in a partial or complete loss of their ambulatory functions. Functional electrical stimulation (FES), a technique to generate artificial muscle contractions with the application of electrical current, has been shown to improve the ambulatory ability of patients with an SCI. FES walking systems have been used as a neural prosthesis to assist patients walking, but further work is needed to establish a system with reduced engineering complexity which more closely resembles the pattern of natural walking. The aim of this thesis was to develop a new FES gait assistance system with a simple and efficient FES control based on insights from robotic walking models, which can be used in patients with neuromuscular dysfunction, for example in SCI. The understanding of human walking is fundamental to develop suitable control strategies. Limit cycle walkers are capable of walking with reduced mechanical complexity and simple control. Walking robots based on this principle allow bio-inspired mechanisms to be analysed and validated in a real environment. The Runbot is a bipedal walker which has been developed based on models of reflexes in the human central nervous system, without the need for a precise trajectory algorithm. Instead, the timing of the control pattern is based on ground contact information. Taking the inspiration of bio-inspired robotic control, two primary objectives were addressed. Firstly, the development of a new reflexive controller with the addition of ankle control. Secondly, the development of a new FES walking system with an FES control model derived from the principles of the robotic control system. The control model of the original Runbot utilized a model of neuronal firing processes based on the complexity of the central neural system. As a causal relationship between foot contact information and muscle activity during human walking has been established, the control model was simplified using filter functions that transfer the sensory inputs into motor outputs, based on experimental observations in humans. The transfer functions were applied to the RunBot II to generate a stable walking pattern. A control system for walking was created, based on linear transfer functions and ground reaction information. The new control system also includes ankle control, which has not been considered before. The controller was validated in experiments with the new RunBot III. The successful generation of stable walking with the implementation of the novel reflexive robotic controller indicates that the control system has the potential to be used in controlling the strategies in neural prosthesis for the retraining of an efficient and effective gait. To aid of the development of the FES walking system, a reliable and practical gait phase detection system was firstly developed to provide correct ground contact information and trigger timing for the control. The reliability of the system was investigated in experiments with ten able-bodied subjects. Secondly, an automatic FES walking system was implemented, which can apply stimulation to eight muscles (four in each leg) in synchrony with the user’s walking activity. The feasibility and effectiveness of this system for gait assistance was demonstrated with an experiment in seven able-bodied participants. This thesis addresses the feasibility and effectiveness of applying biomimetic robotic control principles to FES control. The interaction among robotic control, biology and FES control in assistive neural prosthesis provides a novel framework to developing an efficient and effective control system that can be applied in various control applications