Multi-modal sensing array for soft robots

Abstract

Soft robots, known for their flexibility and adaptability, are poised to revolutionize applications requiring gentle interaction with humans and delicate environments. Despite their potential, the development of soft robots has been limited by the disparity in adaptability and dexterity compared to natural organisms, due to a lack of advanced sensing technologies. Draws inspiration from the sensing mechanisms of biological entities, this thesis presents advancements in the field of sensing in soft robotics, addressing key challenges related to multi-modal sensory integration, high-density tactile sensing, and edge perception. First of all, a novel approach is introduced through the development of a comprehensive functional framework that integrates sensing and action within a closed-loop sensorimotor control system. This framework is based on a multi-modal artificial 3D mechanoreceptor, inspired by the mechanoreceptors in human skin, which enhances the robot's ability to perform dynamic haptic exploration. The sensor features a multi-layered design with a tensor-based nonlinear theoretical model to characterize its deformation modes and optimize its sensing properties. Crack-based strain sensors arranged in a centrosymmetric pattern enable high-sensitivity detection of mechanical stimuli such as normal and shear forces, skin stretch, and vibrations. This design facilitates the extraction of diverse information critical for effective haptic exploration and object recognition, achieving an impressive object recognition accuracy of 96%. Addressing issues of sensor density, the thesis details the design, fabrication, and integration of two generations of large-area, high-density tactile sensor arrays into a robotic hand. The first-generation sensorized skin, featuring 46 sensing pixels, demonstrated effective dynamic performance and sensory feedback. The second-generation e-skin, with an improved density of 110 pixels and the use of advanced piezoresistive materials, achieved over 90% accuracy in object recognition through machine learning, validating its potential for enhancing robotic autonomy and dexterity. The thesis also explores advancements in mechanically encoded materials for edge perception, presenting a groundbreaking approach to improving proprioceptive and tactile sensing in soft robots. By employing sophisticated transduction principles, these materials enhance the detection and processing of environmental stimuli, such as pressure, temperature, and surface roughness. Mechanical models provide valuable design guidelines, and future innovations in this technology are expected to bridge the gap between biological sensing mechanisms and engineered systems. This transformative technology promises to advance human-robot interaction and expand the applications of soft robotics in healthcare, prosthetics, and industrial automation, paving the way for more capable and intelligent robotic systems. Collectively, the advancements presented in this thesis promise to significantly enhance the future of soft robotics. Inspired by the working mechanism of human skin, which can detect multiple mechanical stimuli (multimodality), embeds a number of mechanoreceptors in the skin (high-density), and process sensing information in the skin level (edge perception), this thesis aims to address these three sensing challenges face by soft robots. By integrating a novel sensorimotor control system inspired by human skin, the development of an advanced 3D mechanoreceptor with precise deformation modeling and multi-modal capabilities marks a major leap in haptic exploration. The progress in tactile sensor arrays, especially the second-generation e-skin with increased pixel density and superior materials, enhances robotic dexterity and object recognition accuracy. Furthermore, the breakthrough in mechanically encoded materials offers a new approach to edge perception, improving the efficiency and adaptability of soft robots by minimizing electronic processing needs. Collectively, these innovations bridge the gap between biological sensing mechanisms and engineered systems, setting the stage for more capable, intelligent, and versatile soft robotic technologies in various applications.