I. Motivation
Although tactile cues play an integral role in human interaction with the physical world [1], development of tactile perception systems for robots is less advanced and more poorly understood than visual and auditory sensing modalities. Improving haptic intelligence in robotic systems requires advances in tactile sensors, generation and collection of data from physical interactions using these sensors, and new algorithms to interpret and process the data. In many modern haptic experiments, probes or robots equipped with tactile sensors are used to manually or autonomously explore objects or surfaces. The goal is then to learn relationships between the tactile data and human descriptions or labels of these objects or surfaces. Increasingly, haptics researchers are using machine learning algorithms to discover these relationships.