Modeling and Optimal Control of Human Walking for Motion Understanding and Usage in Lower-Limb Exoskeletons

Abstract

Wearable robotics, also known as exoskeletons, are devices typically worn over the human limbs to provide support and help with movement. Either by assisting the user to increase their capabilities, to help them regain movement abilities like in rehabilitation or to be used as an everyday device for restoring mobility. Exoskeletons have shown promising results in helping individuals regain their mobility by assisting them in walking tasks. However, there are still some challenges we need to overcome in order for the exoskeletons to be a significant part of the daily life. These challenges include but are not limited to: improving the control strategies to provide a better, more comfortable walking motion and to be able to adapt to different walking styles based on the individual needs and reject any perturbations to the motion. Exoskeletons have the potential to change the lives of millions of people suffering from mobility impairments around the world. The current problem with exoskeleton control in gait is mainly based on the idea of using pre-defined trajectories to control the motion, this causes the system to be rigidly imposing motion on the user which is uncomfortable and feels unnatural. This thesis aimed to implement a modified version of the Geyer/Song model to create a more flexible trajectory generation method and add optimal control as a method to control the model gait parameters. The model was verified against the original model from [1]. Given the lack of dependency on any libraries to model the neuromuscular system, the model is compatible with C++ based libraries for the dynamics. A simplified mechanical walking problem was implemented with optimal control to serve as a skeleton problem and allow for the addition of the muscle and reflex systems. A comparison was done on the muscle and joint layers of the modified model. The re- sults have shown that the modified model follows the referenced model behaviour in terms of the motion generated despite some modifications and software differences. The optimal control problem for mechanical walking converged with an objective function of torque minimization and angular momentum minimization and gave a cyclic solution successfully. The results from the torque minimization showed greater promise since they were smoother and close to the expected human walking motion. The muscle-based optimal control walk- ing was formulated but did not converge and thus the further complexity of the reflex system was not added in this thesis. To test the new trajectories, a mechanical frame was designed to hold the TWIN exoskeleton and allow the testing of any modifications with minimum human interaction. The frame is able to hold the exoskeleton and motion is played on it in two positions, one with the exoskeleton hanging with sufficient ground clearance to avoid any ground contact and two with the exoskeleton able to make ground contact and push off to move forward stably. An experiment was designed to evaluate how individuals feel and react to the different exoskeleton trajectories and modes of operation. Initial results showed that the overall EMG signals from the lower limb were lower in the use of the exoskeleton and that the difference between grip forces and loads on crutches showed up the most in periods of instability. These results can help us identify how comfortable the person feels with the motion being imposed on them.