Tactile Displays Realized Using MEMS Actuator Arrays

Sponsor: Defense Advanced Research Projects Agency

Project Staff: S. James Biggs, Andrew R. Brughera, Lorraine A. Delhorne, Tim T. Diller, Nathaniel I. Durlach, Dennis M. Freeman, Kaigham J. Gabriel (CMU), Charlotte M. Reed, David W. Schloerb, Mandayam A. Srinivasan

Introduction

The purpose of this project is to develop new tactile interfaces that capitalize on advances in MicroElectroMechanical Systems (MEMS) technology to create high-bandwidth displays for stimulating the user's tactile sense. Such devices may also receive manual or other types of pressure/contact input. We envision two major classes of applications: (1) Tactile interfaces mounted on machines could indicate the state of the machine--such as the remaining charge in a battery--or it might respond to the operator's touch in subtle ways not possible with conventional controls, (2) Tactile interfaces attached to the human body--for example, through a glove or wrist band--could be used with wearable computers or communication devices for both input and output. The project is a collaboration between the MIT Touch Lab at RLE and Prof. Kaigham J. Gabriel at the MEMS Laboratory at Carnegie Mellon University (CMU). To date, the work at MIT has consisted of experimental studies of biomechanics and tactile perception. This work is intended to guide the design of tactile interfaces and it will also advance the science of human haptics. The group at CMU is developing the MEMS stimulators and sensors that will be incorporated in future prototype interfaces. Following are summaries of the work accomplished in each task area.

Characterization of Human Skin and Tissue Impedance

The goal of this part of the project is to measure and characterize the mechanical impedance of human skin at various body sites. This involves in vivo measurements of the force response to mechanical displacement of the skin. The measurements are made using the Skin Dynamics Test Apparatus (SDTA) that we have developed. The SDTA uses an Aurora Scientific Inc. Dual-Mode Lever Arm System that is controlled by a PC via an A/D card. The apparatus continuously samples both the position and the resulting force on a 0.5 mm diameter probe as it is first pressed against the skin and then displaced sinusoidally about the mean pre-indentation depth. The resulting biomechanical data will be used to determine parameters in previously developed skin models.

Most of our effort up to this point has been devoted to setting-up the apparatus and establishing the experimental procedures. This includes writing the software to run the experiments and to analyze the data, mechanical construction, tuning, and calibration of the stimulator using a Laser Doppler Vibrometer.

Preliminary data for 3 human subjects has been obtained for the finger pad at frequencies ranging from 1 Hz to 400 Hz, amplitudes ranging from 50 microns to 150 microns, and mean pre-indentation depths of 200 and 300 microns.

Perceptual Resolution Measurements

The goal of this part of the project is to determine the limits of perceptual resolution for various kinds of vibratory tactile stimulation at various body sites. The measurements are made using apparatus that is similar to that used for the skin biomechanics experiments (i.e., the SDTA). The Tactile Perception Test Apparatus (TPTA) will ultimately include two Aurora-Scientific-based tactile stimulators and it will incorporate a 5-axis micro-positioning assembly (x, y, z, theta, and stimulator separation). Under computer control, the motorized micro-positioning assembly will be able to continuously adjust the position of the two stimulators over the course of hundreds of successive experimental trials.

At this point, the apparatus is still under construction although all of the major components have been purchased. The hardware design and the development of the software to run the apparatus have taken longer than anticipated. As an interim measure, we have used the SDTA to make some one-point threshold measurements on the finger pad. This was done in order to help develop our experimental procedures, to verify the acceptability of the Aurora-Scientific-based tactile stimulators for perception experiments, and to obtain some preliminary data.

Detection threshold measurements were obtained on the right index finger of four adult subjects at eight frequencies in the range of 2 to 256 Hz. Thresholds were estimated using a two-interval forced-choice adaptive-level procedure with trial-by-trial correct-answer feedback. Each run began with the stimulus level set well above threshold. Presentation level was changed following two correct responses (resulting in a decrease in stimulus level) or one incorrect response (resulting in an increase in stimulus level). The step size was set initially to 4 dB but was changed to 2 dB after the first reversal. A run was terminated after 8 reversals in level and the threshold for that run was calculated by averaging across the levels of the final 6 reversals. The two observation intervals were 500 msec in duration and were separated by 200 msec. Visual cueing of the observation intervals was provided on a computer terminal. The tactual stimulus was presented in one of the two observation intervals, selected at random on each trial. Data were collected in 8-run blocks with random ordering of the 8 frequencies within each block. Five blocks of data were collected on each of the four subjects, leading to five threshold estimates at each of the 8 frequencies (2, 4, 8, 16, 32, 64, 128, and 256 Hz).

Reliable threshold estimates were obtained in the range of 2-64 Hz; however, threshold measurements at 128 and 256 Hz were limited by both the resolution of the system and the level of the electrical noise. Thus, we will summarize only the measurements in the range of 2-64 Hz. Mean nominal thresholds and standard deviations across the five subjects (in dB re 1 micron peak) were:

Subject means and inter-subject variability at these frequencies are consistent with other measurements reported in the literature. During each observation interval, values of stimulus amplitude as a function of time were sampled and stored. Using these sampled values, a calculation was made of the "actual" presentation level on each trial. When these "actual" values were used to calculate thresholds, they were found to be in excellent agreement with the thresholds from "nominal" stimulus levels (i.e., within 0.1 dB at five of the six frequencies).