Design and Control of Quasi-direct Drive Actuation for Lightweight and Versatile Wearable Robots

Wearable robots have shown great potential for augmenting the physical capabilities of humans in lab settings. However, wearable robots for augmenting the physical capabilities of humans under community-based conditions are the new frontier of robotics. Furthermore, the design and control are still considered to be grand challenges for providing physical augmentation for humans. In terms of design, the state-of-the-art exoskeletons are typically rigid, bulky, and limited to lab settings. In terms of control, most of the rhythmic controllers are not versatile and are focused only on steady-state walking assistance. The motivation behind my research is to improve both the design and control performance to develop lightweight, compliant, and versatile wearable robots. Our design and control enable a paradigm shift from lab-based to community-based wearable robots. We envision that our work will enable a paradigm shift of wearable robots from lab-bounded rehabilitation machines to ubiquitous personal robots that can reduce mechanical loading for both the able-bodied wearers and disabilities, such as help with workplace injury prevention, pediatric and elderly rehabilitation, home care, and power augmentation. My research focuses on 1) Developing an innovative actuation paradigm to design compliant AND high bandwidth lightweight exoskeleton 2) Analyzing compliance and control bandwidth performance of quasi-direct drive actuation using a unified human-robot interaction model 3) Developing walking and squatting controllers for non-rhythmic versatile assistance My research pioneered the quasi-direct drive actuator paradigm for wearable robots by designing the first lightweight and compliant quasi-direct drive exoskeleton. The dissertation details the development of high-performance wearable robots, including actuation systems, modeling, and non-gait-cycle-based robust controllers for walking and squatting assistance. The working principle, design, and evaluation of several exoskeletons are presented; from tethered to portable, cable-driven actuation to quasi-direct drive actuation. In addition, a unified human-robot interaction model is presented that compares three actuation paradigms (i.e., conventional, series elastic, and quasi-direct drive actuator) with comprehensive benchmark results in terms of compliance and control bandwidth. Finally, two non-rhythmic non-gait-cycle-based torque controllers leveraging only kinematic data are developed and evaluated on a knee exoskeleton. One is a stiffness model-based controller for walking, and another is a biomechanics model-based controller for squatting. Rigorous human subject experiments were conducted to evaluate the efficacy of the designed exoskeleton and the novel controllers. The experimental results show that our designed lightweight and compliant knee exoskeleton can reduce 7.45 - 15.22% extensor muscles’ activation under walking assistance and reduce 70% - 87.5% extensor muscles’ activation under squatting assistance. The metabolic rate during squatting was reduced by 9% compared to squatting without an exoskeleton. This reduction in metabolic rate is comparable to the effects of removing 7.2 kg weight from an 80 kg subject. It shows that the proposed exoskeleton design and control improvement can significantly augment the physical capabilities of humans during both walking and squatting, which is a big step toward wearable robots providing superhuman augmentation