A design framework for user-specific multi-DoF wearable robots

Abstract

A considerable fraction of the world population is plagued by neuromuscular disorders and ageing-onset weakness that have no cure other than symptomatic management. The mechanical management of the manifestations of diseases ranging from Parkinson’s disease to stroke has been an active research topic in robotics since the 1990s. Advances in sensing, actuation, energy and computing technologies together with breakthroughs in material sciences have facilitated an increased penetration of soft wearable robotic technologies resulting in applications from healthcare to space. The compliant nature inherent to the soft design paradigm makes soft-bodied wearable robots ideal for assisting complex joints such as the shoulder and wrist. However, most exosuit research focus on single-DoF suits, with the few suits capable of multi-DoF assistance all adopting a biomimicry-based design approach. The biomimicry-based approach is sub-optimal and requires an excessive number of actuators/sensors, negating the inherent size and bulk advantages of soft-bodied systems over their rigid-bodied equivalents.

To develop a better alternative, we took (bio)inspiration from the concept of muscle synergies and the physiology behind the sense of proprioception to propose a more efficient multi-DoF exosuit design framework. While an exosuit prototype with actuation was not realised, the sensing method of the proposed architecture was validated in simulation and a physical prototype, which confirmed conformance to the concepts that inspired the suit architecture. The mapping to derive joint kinematics was done using artificial neural networks, working analogously to the brain's processing of sensory information to give the sense of proprioception. More sophisticated neural networks were then investigated to derive a mapping that also compensated for the suit's inherent nonlinearities resulting in joint tracking accuracy comparable to state-of-the-art wearable kinematic sensing systems, and this was demonstrated in simulation. Additionally, the research in this thesis also proposes a method to move away from a qualitative designer-centric approach to a novel mathematical-modelling-based approach to exosuit design. The proposed framework was grounded in 3D limb-shape reconstruction, a continuum-mechanics-based study of skin deformation, and a model to understand the behaviour of lines of actuation under tension across the limb surface. This data-driven method was undertaken to develop a more systematic approach to exosuit design and provide a platform for wearer-specific design optimisation.