Advancing wearable soft robotics: development, characterization, and modeling of soft sensors and Shape Memory Alloy (SMA) actuators for closed-loop active on-body compression

Abstract

Unlike traditional rigid robotics, soft robotics utilizes flexible, compliant materials that allow for safer interactions with humans and better adaptability to dynamic, irregular environments. This inherent compliance and flexibility make soft robotics particularly suitable for wearable applications, where direct and conformable interaction with the human body is necessary. This adaptability is especially beneficial in fields such as rehabilitation, human-robot interaction, and wearable haptic devices, where gentle force application and the ability to conform to the body’s shape are critical. While advancements in soft sensors and actuators have significantly driven wearable technologies forward, challenges remain in the physical interface between these systems and the human body. Soft sensors can provide objective, quantitative measurements of mechanical interactions, but accurate on-body sensing is complicated by factors such as body movement, tissue compliance, and forces introduced by the garment-body interaction, including stretching and folding. These issues are particularly problematic for textile-based sensors, which are prone to nonlinearity and hysteresis due to their material properties. On the actuator side, significant progress has been made with smart materials like Shape Memory Alloys (SMAs), which have garnered attention for their ability to provide dynamic compression in wearable systems. These systems have potential applications ranging from medical therapy to aerospace, particularly for astronauts countering the challenges of microgravity. This research provides a comprehensive investigation into soft sensing and actuation for wearable applications, leveraging advanced materials and modeling techniques. First, we develop and evaluate soft force sensors for on-body sensing, addressing the complexities introduced by non-flat, dynamic body surfaces. Traditional models, as well as artificial neural networks such as LSTM, are employed to improve force estimation accuracy under these challenging conditions. A textile-based sensing matrix, composed of 72 individual sensors, is introduced to measure distributed forces across complex body geometries. Among the explored modeling approaches, the LSTM model demonstrated superior performance in estimating forces, achieving low modeling errors and showing promise for real-time wearable applications. In parallel, we advanced the use of SMA actuators in wearable systems, particularly in compression devices. This research focuses on the characterization and modeling of both SMA spring and knitted wire actuators, addressing key challenges such as material fatigue, force degradation over repeated cycles, and dynamic force control. The integration of these SMA actuators into wearable compression systems, combined with adaptive control strategies, enabled precise force regulation across varied body surfaces. The results show substantial improvements in force stabilization, making these systems suitable for long-term applications in medical therapy, athletic performance enhancement, and even aerospace environments. Overall, this work contributes to the development of a fully integrated soft sensing-actuation system capable of achieving real-time dynamic force control in wearable devices. By incorporating advanced soft sensors and SMA actuators with data-driven modeling techniques, this research lays the groundwork for future innovations in wearable robotics, with potential impacts across healthcare, sports, and aerospace applications.