Walking is difficult or impossible for many spinal cord injured (SCI) individuals, due to lost voluntary control of the leg musculature. Devices developed to restore lost function and provide ambulation offer the potential to improve mobility, health, and overall quality of life of SCI individuals. Efforts in this area have focused on two areas:​ mechanical orthoses and functional neuromuscular stimulation (FNS).

One mechanical orthosis, the reciprocating gait orthosis (RGO), can stabilize and support joints and reduce muscle force requirements. The RGO locks the ankle and knee, and incorporates a linkage mechanism to facilitate reciprocal leg movement by kinematically coupling the two hip joints. Though it can provide stable, unsupported standing with minimal effort and crutch-assisted walking for short distances, the RGO does not provide sufficient gait performance to meet the needs of the SCI population.

Existing FNS systems attempt to generate walking movements by electrically stimulating many leg muscles with simple control strategies. These systems have failed to gain clinical acceptance because they require high levels of stimulation, resulting in rapid fatigue. Thus, FNS-only systems have provided only low-speed walking for short distances. Furthermore, existing technology for selectively stimulating large numbers of muscles has not provided adequate ease-of-use or long-term reliability.

The hybrid RGO/FNS system offers a near-term solution for functional ambulation with currently available technology. However, dynamic models are required, (i) to enhance current understanding of human musculoskeletal movement in the RGO;​ (ii) to provide the tools needed to evaluate RGO/FNS gait strategies;​ and (iii) to design nonlinear FNS control systems.

This thesis presents the development and evaluation of a model for the swing phase of human paraplegic ambulation in an RGO. First, experimental and inverse-dynamics analyses of a SCI individual walking in an RGO (without FNS) were conducted to determine the kinematics and kinetics of RGO gait. Next, a dynamic model of RGO gait was developed. Finally, the model was used to generate simulations to improve understanding of the dynamics of RGO ambulation and to investigate strategies for how to improve RGO gait performance with FNS.

Three-dimensional movement, foot/ground reaction force, and crutch/ground reaction force data were acquired from a single SCI subject (complete T11/T12 lesion) walking in an RGO. A 3D, eight degree-of-freedom biomechanical model for the swing phase of RGO gait was then developed. The model consists of four segments:​ the right and left legs, the pelvis, and the upper trunk. Upper-body assistance (from crutches) is accounted for by a set of control forces/torques acting at the shoulders, and stance foot/ground interaction is represented by rolling of a 3D ellipsoid surface on a plane.

Forces and torques required to generate a dynamic simulation tracking experimentally measured joint kinematics and ground reaction forces during swing-phase of RGO gait were determined with an optimal control algorithm. Good agreement between simulation and experiment was obtained.

The swing-phase dynamic model was used to investigate efficient strategies for RGO/FNS gait. Experimental measurement and simulation suggested that much of the inefficiency associated with RGO gait may be caused by the deceleration of the trunk and body center-of-mass (COM) during swing phase, requiring a great deal of muscular effort during double support to restore lost momentum. An optimal control problem was formulated to provide a smoother COM progression and minimize swing-phase muscular effort, while maintaining the same walking speed attained in RGO gait. The body segment velocities required to initiate swing phase were also optimized to reduce energy requirements of the preceding double-support phase.

Simulations demonstrated that swing-phase movement for moderate-speed RGO gait requires very little muscular effort, if the proper initial conditions can be established. The. gait patterns generated via optimal control provided smooth forward progression of the trunk and body COM, with muscular effort required only to generate swing-foot clearance and maintain balance in the coronal plane. These findings are consistent with previous studies of dynamic mechanisms for efficient stiff-legged ambulation, which have shown that purely ballistic (passive) sv/ing-phase is possible for two- and three-segment systems.

The initial segment velocities required to produce nearly ballistic swing-phase movement are less than those observed during RGO gait, so energy requirements of double support should be reduced. However, FNS would be required during double support so that swing can be initiated without the pelvic "thrust" that would otherwise be required.

Most existing RGO/FNS systems utilize FNS just prior to and during swing, leaving the user to rely upon upper-body strength during the remainder of double support. In contrast, the work presented here suggests that efficient RGO/FNS ambulation may require precise control and significant FNS-generated torques during the entire double-support phase, in order to establish the proper initial conditions for nearly ballistic swing-phase movement.