The BioTac - Multimodal Tactile Sensor

This page describes the SynTouch BioTac sensor. For software and downloads please see the SynTouch ROS Package.

Overview

Humans require the sense of touch to identify objects based on their tactile properties and to dexterously interact with them. Until recently, humanoid robots have needed to rely on machine vision as a substitute for this important sensory function. The BioTac (SynTouch LLC) is a unique tactile sensor capable of acquiring sensory modalities that mimic the full range of capabilities in the human fingertip. It consists of a rigid core that houses all of the sensory electronics and an elastomeric skin made of low-cost silicone, making the design uniquely robust and easy to repair if the skin becomes worn or damaged. The space between the skin and core is inflated with a liquid giving the sensor a compliance very similar to the human fingertip.

The sensory modalities of the BioTac were modeled after the tactile function of the human fingertip to robustly sense:

• Forces: The BioTac is capable of sensing both normal and shear forces in addition to point of contact. Most other tactile sensors provide what we refer to as taxels (tactile pixels) capable of sensing only normal forces at different locations. In the BioTac, as contact forces distort the elastic skin and underlying conductive fluid, changes in electrical impedance between 19 electrodes distributed over the core can be detected and used to extract force and point of contact through machine learning (Wettels & Loeb 2011) or straightforward analytical approaches (Lin et al.). Like the human fingertip, the BioTac captures the ability to sense shear forces through the use of a fingernail. Shear forces applied to the sensor cause stretching and bulging of the skin around the nail that are readily detected by electrodes in this region. The electrode impedances can also be used to discriminate objects based on compliance (Su et al., 2012).

• Vibrations - Humans have exquisite sensitivity to small vibrations less than a micrometer near the natural frequency of the human fingertip. When vibrations arise on the surface of the BioTac's skin (due to slip or contact events) they propagate through the skin and fluid with very little attenuation where they can be sensed by a pressure sensor inside the core. In our research we have demonstrated that the BioTac is even more sensitive to contact than the human fingertip and capable of sensing vibrations down to a few nanometers (Fishel & Loeb, 2012a). With this high sensitivity we have been able to record vibrations to discriminate textures with better performance than humans (Fishel & Loeb 2012b).

• Heat Flow - Many materials we interact in the world are attributed properties of being warm (blankets, foams, etc.) or cool (metals, concrete, etc.). Given that both a blanket and a piece of metal that have been sitting on a shelf for a long time are both in fact the same temperature (room temperature) this is an interesting association to make. It turns out this phenomenon has to do not with temperature, but with heat flow. As the human body is typically warmer than the environment, when your finger makes contact with an object heat flows out of it and into the object. Thermally conductive objects such as metals cause heat to flow more readily, making your fingertip feel cooler than when touching a thermally insulative object. Using these principles the BioTac has been designed with a heater and thermistor to capture the same phenomenon. The sensor has been reliably used to discriminate objects based on their thermal properties (Xu et al., 2013)

See the two videos below for a demonstration of the BioTac's capabilities:

Using the BioTac

Note: Detailed descriptions of the BioTac sensory modalities and function are covered in its Product Manual.

All sensory modalities of the BioTac are sampled and digitized at 12 bits by a microprocessor inside the sensor. Communication between the BioTac and other devices is done using SPI protocol as explained in more detail in the Product Manual. Through use of chip selects, multiple BioTacs can share the same sampling sequence and timing. In most applications, the SPI host sends a command to sample a particular channel to all BioTacs and then reads their responses back individually. The following channels of information are available from the sensor:

• Electrode Voltages (E1 - E19) are measured across a voltage divider with reference to a load resistor. When pressing down over an electrode the measured voltage will decrease.

• DC Pressure (Pdc) or static pressure can be read from the sensor and increases linearly when the fluid pressure increases.

• AC Pressure (Pac) or dynamic pressure is generated from a high-pass filtered version of DC Pressure with additional gain allowing for high resolution of vibrations.

• DC Temperature (Tdc) can be read from the sensor and decreases as the device warms up.

• AC Temperature (Tac) is a high-pass filtered version of DC Temperature that permits for higher resolution of thermal fluxes. This decreases as the device is cooled.

• Hall Sensor (Hall) is available on newer versions of the BioTac and is used by the Shadow Dexterous Hand for joint angle encoding.

Configuration

Details of configuring the BioTacs and data acquisition hardware are covered in the descriptions of the SynTouch stack and the st_biotac package.

Signal Processing and Applications

Many biomimetic features of the sensor such as the inclusion of the fingernail, fingerprints and internal heating give rise to the ability to perform a number of biomimetic signal processing functions including:

• Estimating texture, compliance and thermal properties (Fishel & Loeb, 2012b;Xu et al., 2013)

• Rapidly detecting object contact (Fishel & Loeb, 2012a)

• Determining tri-axial force (Wettels & Loeb, 2011, Lin et al.)

• Determining point of contact (Wettels & Loeb, 2011, Lin et al.)

• Detecting slip (Lin & Fishel, unpublished)

• Estimating the radius of curvature of a contacted object (Wettels & Loeb, 2011)

• Discriminating edges, corners, and flat surfaces (Su et al., unpublished)

When combined with actuators a number of dexterous and perceptual applications can be accomplished including:

• Compliant and fragile grasping (Matulevitch et al., in review)
• Position/Orientation correction when grasping objects (Lin et al., under development)
• Slip detection and grip control (Lin et al., unpublished)
• Contour following (Su et al., under development

• Tactile object discrimination using Bayesian exploration (Fishel & Loeb, 2012b;Xu et al., 2013)

To support these functions we intend to develop ROS packages to be included in the SynTouch stack.

For more literature covering the development of these applications see SynTouch Publications