Design Considerations for Multimodal "Sensitive Skins" for Robotic Companions

Abstract

The sense of touch is the earliest sensory system to become fully functional in all species. Even as early as nine weeks in the womb, the fetus will close its fingers in a gripping motion if its palm is touched (Montagu 1986). The skin, our largest sensory organ, is capable of detecting skin indentations as small as 1μm (Nolte 2002). It is our sense of touch that helps to protect us by allowing us to feel pain so we do not damage our bodies. We manipulate objects using the tactile sensors of our hands. Additionally, touch has an affective and social component – expressed through hugs, handshakes, and pats on the back. One only needs to look at the interactions between humans and their pets to see these expressions of affective touch in action. To help motivate this discussion of the importance of touch let’s conduct a thought experiment. Imagine you wake up in bed and have to start your day without using your sense of touch. How far would you get just relying on vision and sound processing alone? Most likely, you would have trouble leaving your room. Now replace yourself with an autonomous robot and you can see why tactile sensing systems are important. One major goal of robotics research is to produce robotic systems which can exist in the very complex human world. Touch sensing systems can add a great deal to allowing the robot to function autonomously in this world. First, vision alone is not sufficient to function in unstructured environments due to problems of occlusion (Lumelsky et al. 2001). Second, tactile information can be combined with vision and auditory information to form stronger multi-modal percepts. Third, a full body tactile system can help to convey the “illusion of life” in the robot – as no matter how lifelike the motion of the robot may appear, if the robot is touched and it does not respond, that illusion is instantly broken. Fourth, in the context of human-robot interaction (HRI), the sense of touch can carry social, affective, or other information. In addition to these qualities, a well designed skin should feel pleasant to touch and not distract from the interaction. In many cases, the field of robotics today has focused much of its attention on vision and auditory sensing. If tactile sensors are used, they have primarily been confined to discrete locations, such as those of Sony’s AIBO robotic dog (Sony Product Literature) or have been primarily used in grippers for manipulation tasks, such as NASA’s Robonaut (Martin, et al. 2004). While these uses are important, there exists much potential for full-body, multimodal, tactile sensing systems – or “sensitive skin” as defined by (Lumelsky, et al. 2001) . O pe n A cc es s D at ab as e w w w .in te ch w eb .o rg