Lightweight, Non-Restrictive, and Adaptive Back Exosuit for Use in Dynamic Lifting Environments

Abstract

For decades, academia and industry have explored opportunities for back exoskeletons and exosuits to reduce low back injury, which is a major public health issue around the world. Even though back exoskeletons and exosuits have shown great potential by reducing lumbar moments and back extensor muscle activities, their widespread adoption in the workplace has been limited in large part due to discomfort and usability challenges such as movement restriction and the weight of the device. This dissertation details the design, control algorithms, and evaluation of an active back exosuit for use in dynamic lifting environments. This work explored a new hardware design that is lightweight, non-restrictive by utilizing flexible ribbon cable actuation and external muscle architecture. The overall device weighed only 2.7 kg, similar to regular backpacks, even including batteries. All the actuation unit and motion sensors are integrated into the function apparel which makes it easy to don and comfortable to wear for extended periods of time. An adaptive impedance controller was developed to address sagittal plane restrictions of passive systems while maintaining biomechanical benefits during lifting. It applies low impedance during lowering and high impedance during lifting and introduces the transition phase with velocity-based interpolation to make smooth transition between lowering assistance and lifting assistance. The back exosuit was evaluated in a series of experiments. The first experiment evaluated its adaptability compared to passive systems with different levels of assistance. The results demonstrated that an active back exosuit with adaptive impedance controller could achieve back extensor muscle activity reduction similar to a high stiffness passive system, while maintaining maximum sagittal plane range of motion as much as a low stiffness passive system. The second experiment evaluated the back exosuit for an hour-long work simulation task involving lifting, carrying, walking, and picking up boxes under a shelf. The adaptive impedance controller worked robustly and mis-triggered only for several times out of 9,600 lifts (0.1%). It reduced back extensor muscle activities by 18%, which is the highest documented reduction during the unconstrained dynamic lifting task for more than 30 minutes. The final part of this dissertation explored the possibility of scaling the assistance based on task demands, i.e., whether the user holds an object or not, and the weight of the object. The task adaptive controller was developed by using a camera mounted on the exosuit and the machine learning algorithm trained to recognize objects. A proof-of-concept study demonstrated that the controller successfully scales the forces based on the image data and improved the biomechanical efficacy during lowering down the object to the ground compared to the previous controller. Additionally, it improved the user perception by eliminating the negative aspects of the previous controller such as jerkiness or too much assistance while handling a lightweight object. This dissertation presents a comprehensive investigation of a soft active back exosuit aimed at mitigating low back injuries in the workplace. Through the integration of a lightweight and non-restrictive design, as well as adaptive control algorithms, this back exosuit exhibits high levels of device usability, significant reductions in back extensor muscle activity, and favorable perceptual feedback. These results demonstrate the exosuit's potential for practical application in real-world environments in the future.