Biology to Technology in Active Touch Sensing - Introduction to the Special Section

RECENT years have seen increased study of the role of active touch sensing in animal behavior. Touch is an important means through which animals acquire information about their surroundings. The behavioral tasks enabled by active touch, and the goals that these tasks subserve, are extraordinarily diverse. They may range from locomotion (walking, climbing, and crawling), to foraging, feeding, grooming, and tool use. In most cases, these tasks would be impossible to complete without touch sensation. This notion should be familiar to anyone who has dismissed the possibility of eating after a visit to the dentist that included local anesthesia. The close interplay betweenmovement and somatosensation is a distinctive hallmark of the hapticmodality. Our understanding of haptic perception has built on research that has unfolded over the course of a century. This pace stands in stark contrast to the rapid advances in robotic systems, engineering design, and control that have occurred over a much shorter time span, measured in years or, at most, decades. Even as the operational capabilities of such systems has expanded, it has become increasingly clear that touch sensing is indispensable for robotic systems that are required to function effectively in real-world environments. Many tactile sensing technologies have been developed during the last few decades [14], but the important role played by motor behavior in touch sensing is only beginning to receive wide attention in the robotics community. This has served as further motivation for the present special section on active touch sensing in humans, robots, and other animals. Active touch sensing is recovering information about the world by “touching” rather than “being touched” [6]. There are differences of opinion, however, about the extent to which touching motions must be “purposeful” in order to qualify as active touch. While “passive touch” refers unambiguously to the activation of mechanoreceptors by an external force, the term “active touch” embeds both physiological and psychological components [4], [5], [6], [8], [9], [10]. The physiological component is the activation of mechanoreceptors via self–generated movement, while the psychological component is the intent of the movement: either to displace a limb to a different position in space (e.g., reaching or locomotion), or to acquire sensory information (e.g., when determining fabric texture or when manipulating a tool) [10]. For example, walkingwould not typically be used as an example of “active touch,” but tapping one’s foot against mud to sense compliance would be – even though these two behaviors might actively stimulate the foot in similar manners. In engineering language, active touch refers to the idea of interpreting touch–elicited signals that are captured by sensors whose motion is deliberately controlled to facilitate information gain or to otherwise achieve behavioral goals that depend on it [3]. Thus, the term “active” is applied to suggest an essential role of motor behavior in eliciting or shaping sensory signals, at least when the movement is intended to capture information via touch. This can be contrasted with familiar situations such as stationary listening to an auditory scene, which might involve cognitive activity on the part of the listener, but which are otherwise largely independent of the movements of the listener. The movements involved in active sensing may be deliberate, or they may be reflexive, driven by hardwired behavioral patterns or activity controlled at earlier stages in neural processing. In an engineered system, the former might refer to a reactive algorithm for exploration that adapts motor behavior to what is felt in a continuous fashion, while the latter would suggest a pre–programmed strategy in which a touched surface is systematically probed in a way that elicits sensor signals that are to be interpreted at a later stage. As such, in animals or robots, the behavior involved may depend to a greater or lesser degree on the sensory signals that are felt. This has led to some ambiguity, and some authors [3], [5] have preferred the term “active perception” as opposed to “active sensation.” The former emphasizes that the process of controlling the sensor is indeed understood to capture sensory information that can refine integrated percepts or achieve related goals of the organism. In animals, the process of active touch sensing can engage multiple body parts and employ highly evolved specializations of the sensory and motor systems. In humans, the Y. Visell is with the Department of Electrical and Computer Engineering, Media Arts and Technology Graduate Program, University of California, Santa Barbara, CA 93106. E-mail: [email protected]. N. Lepora is with the Department of Engineering Mathematics and Bristol Robotics Laboratory, University of Bristol, Bristol, BS8 1TH, United Kingdom. E-mail: [email protected]. M.J.Z. Hartmann is with the Departments of Mechanical Engineering and Biomedical Engineering, Northwestern University, Evanston, IL 60208. E-mail: [email protected]. V. Hayward is with the Institute of Intelligent Systems and Robotics, Sorbonne Universits, UPMC Univ Paris 06, UMR 7222, ISIR, F-75005, Paris, France. E-mail: [email protected].