Sensing Touch

Author: Jeremy Fishel

Design Goals: Since the project's inception, the goal of the BioTac project was to produce an artificial tactile sensor that mimicked the size, shape and function of the human fingertip.

Human finger compared to the BioTac conceptual schematic.

Particular focus was paid to develop a robust tactile sensor that has the sensitivity and response to all tactile sensing modalities found in the human fingertip. While human fingers have mechanisms to repair themselves while damaged, we do not have these luxuries in the artificial world. Instead the design goal of the BioTac was to develop a sensor that is low cost and easy to repair when it inevitably wears out or becomes damaged.

The Human Sense of Touch: Physiologists divide touch into two different categories. Proprioceptive touch consists of various neural transducers found in muscles that allow us to detect the position of our muscular-skeletal system and how much force is being applied by our muscles. Cutaneous touch consists of sensory receptors found within the skin that respond to various forms of mechanical stimuli.

The two categories of sensory transducers inside the human body that give rise to the sense of "touch": Prorprioceptive Touch and Cutaneous Touch.

The Artificial Sense of Touch: Current goals in robotics frequently strive to emulate human performance in dexterous manipulation by mimicking the mechanical capabilities of the human hand. If these applications are to be successful, they will require similar sensory capabilities of the human hand. Robotics offers many solutions for the sense of proprioceptive touch.

The Proprioceptive sense of touch is well covered in the artificial domain with many technologies such as strain gages and position encoders to mimic these biological capabilities.

Unfortunately, the Cutaneous sense of touch has not enjoyed this level of progress. Most tactile sensors have developed along what we consider to be the wrong direction. Many engineers have mistakingly concluded that the biological sense of touch is merely a high-resolution sense of normal forces to create an high resolution image of a surface, or as we mockingly refer to them at SynTouch: "taxels".

Most competing tactile sensors are designed along the line of creating high-resolution tactile pixels or "taxels" which only focus on a small portion of what the true sense of touch is. They miss out on some very important elements of touch such as vibration sensing and thermal sensing which are essential for many tasks such as grip control and object identification

This approach of tactile pixel sensing appears to be inspired by the the great progress made in artificial vision which has experienced a substantial success. However, for the biological sense of touch, this is just not how things work. To demonstrate this concept try holding a penny on your finger without moving, you'll find that it is impossible to distinguish which side is touching your finger when the penny is resting on it.

Humans aren't very good at high-resolution tactile sensing... To realize this, try the following example: place a penny on your finger and without looking try to guess which side you are touching.

The reason this task is difficult is because humans are not good at this type of high-resolution pixel imaging, in fact they are incredibly bad at it compared to our visual capabilities to do such discrimination.

So what can we learn from this? We've concluded that if nature hasn't endowed us with sensors that have incredibly high-resolution to do such tasks, clearly it is not that important to do things humans do. When presenting this test, I typically wait for someone to interject with the idea that they can tell which side they are touching, a perfect segue to what the sense of touch really is. Clearly when reaching into our pockets it is rather trivial to determine which side of the coin you are touching. This feature extraction is however completely unrelated to the sense of high-resolution imaging, and instead completely dependent on vibrations. Looking closely at the penny you'll notice that the tail's side has some high-contrast features: the text of "ONE CENT" and "UNITED STATES OF AMERICA" as well as the columns and edges of Lincoln's memorial all elicit very powerful vibrations when sliding your finger over them, especially compared to the other side of the penny which is relatively smooth. This is an excellent example of why multi-modal touch with the full suite of sensory capabilities of the human finger are necessary, without the ability to detect these vibrations we would be completely helpless in this task. Developing a sensor that has such multi-modal sensor capabilities is one of the core design principles of the BioTac.

The BioTac includes all three cutaneous sensory mechanisms found in skin in a simple and robust package.

The BioTac Design: The BioTac consists of a rigid core that contains all of the electronics for it to function. It is covered with a low-cost and easy to replace silicone skin and the space between the skin and the core is filled with an electrically conductive fluid, giving it the size, shape and compliance of the human finger. This design is very robust as there are no wires or sensors inside the skin or fluid, and if the skin ever wears out or becomes damaged it is easy and low-cost to replace.

The three sensory modalities of the BioTac are made possible by three separate sets of transducers. As forces are applied to the skin, the skin and fluid deform and the changes in fluid impedance as it deforms can be detected by an array of electrodes on the surface of the core. As objects are slid across the surface of the BioTac they generate vibrations that can be detected by a hydroacoustic pressure transducer inside the core. As objects of different thermal conductivity come into contact with the core, the heat that flows from the BioTac into the object produces thermal gradients that can be detected as a change in temperature of the thermistor in the BioTac tip.

The three sensing modalities of the BioTac.

In designing a sensor that follows these biological capabilities our own research team and collaborators have been able to do things that humans are able to do with touch, such as:

SynTouch, does not currently provide software to directly extract these features, but will be developing it in the near future. Examples of such applications can be found in the publications.