UC San Diego

As robots grow increasingly prevalent in real-world environments, sensory systems capable of sensing complex deformations and environmental interactions are needed for robust control. Soft robotics has emerged as a field of study that seeks to replace rigid components in traditional robots with materials that are compliant. It has garnered interest for real-world applications due to intrinsic safety embedded at the material level, deformable materials capable of shape and behavioral changes, and conformable physical contact for manipulation. Yet, with the introduction of soft and stretchable materials to robotic systems comes a myriad of challenges for sensor integration, including multi-modal sensing capable of stretching, embedment of high-resolution but large-area sensor arrays, and sensor fusion with an increasing volume of data. This dissertation describes the design, fabrication, and data processing of soft, tactile sensor skins, with the ultimate goal of enhancing future collaborative robots that will work alongside and physically interact with people with a human-like sense of touch.

This thesis focuses on how the integration of soft sensor skins and machine learning enables soft robots to perceive physical interaction for complex haptic tasks. Part 1 on Soft Sensors (Chapters 2, 3, and 4) presents the design and fabrication of various types of soft strain sensors, using compliant materials such as silicones and polymers. Fabrication methods include soft lithography and 3D printing. Performance of the sensors are characterized and modeled. Part 2 on Soft Robot Perception (Chapters 5 and 6) describes how machine learning can be used to augment the performance of soft sensors and actuators. The method demonstrates how recurrent neural networks can be used for graceful degradation and learned perception of external contacts and forces despite not having a priori information about the individual sensors. Part 3 on Social Touch for Physical Human-Robot Interaction (Chapters 7 and 8) analyzes the use of the liquid metal sensors as robotic skins for the classification of affective (social) touch and builds towards the development of a framework for representing physical contact, for use at the human-robot interaction layer of abstraction.

In recent years, the concept of soft-bodied robots has rapidly grown in popularity. Researchers have developed many interesting forms of actuation that more closely mimic the functionality and capabilities found in nature. The next step for the field is to develop biologically-inspired tactile sensing for soft-bodied robots that can safely interact with and explore their environments. In the short term, the field can focus on deployable, high-resolution sensor skins, algorithms for processing the dense sensor information, and reliable feedback control for soft robots. Building upon the fundamental work presented in this dissertation, the future consists of robots that can touch and feel with the sensitivity and perception of natural systems.