Preview

PDF, English - main document Download (30MB) | Terms of use

Abstract

The 21st century witnesses significant advancements in wearable assistive robotics, shifting from rigid exoskeletons to flexible, soft robotic suits. These suits, being clothing-like devices, better align with the body’s natural biomechanics, offering enhanced comfort, efficiency, and adaptability. However, challenges remain in balancing support, energy efficiency, and real-world readiness.

This thesis aims to address these challenges by developing a wearable soft robotic suit designed to assist walking for both healthy individuals and those with mobility decline, with older adults as a key use case. The research has two objectives: (1) Technological development: designing a lightweight, portable device that minimizes sensory input and calibration, providing a simple, plug-and-play solution. The focus is on hip flexion assistance, a critical movement for walking, especially for frail individuals with weakened hip flexors due to aging; and (2) Real-world assessment: evaluating the device’s effectiveness in healthy individuals and later in frail older adults, to assess its potential in dynamic, everyday environments. The thesis hypothesizes that a tendon-driven, lightweight design can balance efficiency, portability, and comfort across diverse users. Two actuation configurations are explored: an underactuated single-motor system and a fully actuated dual motor system for bilateral assistance. The research also posits that a control system using minimal sensory input, such as inertial sensors, can adapt dynamically to gait cadence and, as an add-on, leveraging computer vision for context-aware control can effectively modulate assistance across environments. Ultimately, it is hypothesized that this soft robotic suit can enhance metabolic efficiency, maintain natural gait patterns, and support user autonomy.

Beyond the hardware design, a significant contribution of this thesis is the development of advanced control strategies to ensure synchronization with the user’s natural gait.

Initial development focuses on laboratory testing, exploring dynamic gait phase estimation using an Adaptive Oscillator (AOs)-based controller relying on inertial sensors. This approach works well in controlled settings but faces limitations related to delayed responses during abrupt gait transitions, such as cadence changes or starting from a standstill. To improve performance, a machine learning (ML)-augmented AOs controller is introduced in a sub-study, enabling dynamic adjustments during transitions.

A key innovation of this thesis, beyond gait cadence adaptation, is the integration of computer vision for context-aware modulation. While most current assistive devices are optimized for level walking, the controller of the device can integrate computer vision to analyze environmental features, such as terrain variations (e.g., slopes, stairs, or flat ground), to dynamically adjust the level of assistance. This contextual awareness enables the system to more effectively respond to real-world challenges, significantly enhancing its adaptability, versatility, and efficiency, marking a major advancement in wearable assistive robotics.

The transition from laboratory testing to real-world applications highlights, however, the need for a more streamlined control strategy. The complexities of dynamic environments informs the development of a simplified, high-performance control system that eliminates the need for dynamic models or machine learning. The improved controller employs a phase portrait-based approach, estimating gait phases from raw motion data, computing the angular separation between position and velocity vectors in the hip phase portrait. This refinement enhances the system’s effectiveness in outdoor settings and sets the stage for trials with both fit and frail individuals.

The real-world applicability of the soft robotic suit is assessed through field tests, starting with young, fit individuals hiking outdoor. The suit demonstrates an average 17.79% reduction in metabolic cost of transport, improving efficiency without disrupting natural gait. These results are indicative of the suit’s ability to augment walking in physically demanding activities in unstructured environments, a challenge for many systems designed for controlled settings. The suit also preserves the user’s sense of agency, being both effective and intuitive to use.

Finally, the soft robotic suit is tested in older adults, showing an average 10.48% reduction in metabolic cost of transport during outdoor walking tasks on flat ground. This is particularly significant for frail individuals, who often experience reduced walking efficiency and increased fatigue with age.

The overall results demonstrate the suit’s potential to effectively assist walking while maintaining natural gait patterns. These findings suggest the possibility of improving mobility and reducing fatigue for both fit and frail users.