JP4471613B2 - Tactile information detection method and tactile information detection system - Google Patents

Description

  The present invention relates to an automatic sensitivity variable tactile information detection method and system capable of automatically adjusting sensitivity according to the output of a strain gauge as a sensor element attached to each measurement point.

  Various tactile receptors exist in human skin, and can be distinguished from a light touch to an impact force. For example, there are about 1500 Meissner bodies and about 750 Merkel tactile discs per square centimeter at the top of the skin, and about 75 Basinian bodies and Rufini organs deep in the skin. A wide measurement range is realized by these four types of tactile receptors with different temporal and spatial responses.

  When considering an environment where humans and robots coexist, it is considered that a tactile sensor attached to the robot always needs an appropriate measurement range. For example, when a detailed work is considered with the fingertip of a robot, a highly sensitive tactile sensor may be required. On the other hand, a sensor capable of measuring a large impact force may be required to avoid danger.

Research on tactile sensors in robots It started around 1960 by Erunst et al. Since then, many tactile sensors have been proposed from various angles. For example, a tactile sensor improved in sensitivity, resolution, linearity, a small wiring mechanism, and mountability has been proposed.

Here, conventionally, a method of measuring tactile information at each measurement point where each tactile sensor is arranged by sequentially arranging tactile sensors in an (M × N) matrix and sequentially switching each tactile sensor. Is widely adopted. The problem with this type of tactile sensor is that a large number of wires are required to read information at each measurement point.

  In order to reduce the number of wires, new wireless tactile sensing has been proposed by Shinoda, for example. This uses a coil to perform power transfer and sensing wirelessly. In addition, this sensor has a configuration in which many resonance circuits are arranged in a flexible material, so that the contact portion can be identified.

  The conventional tactile sensor has a problem that the signal is saturated as soon as a signal with a strong level exceeding a specified value is input, or the sensor resolution is deteriorated when the contact force is extremely small. Therefore, it is necessary to perform automatic gain control (hereinafter referred to as AGC) of the tactile sensor for performing an appropriate gain adjustment in accordance with the input to the tactile sensor.

  When a tactile sensor having a large area is considered, the contact force input to each sensor element is not constant, so that it is expected that the sensor is partially saturated or the resolution is deteriorated. In order to prevent these, it is desirable to be able to adjust the gain of the sensor element for each measurement point.

  Furthermore, in the case of a wired tactile sensor, the number of wires connecting between each sensor element installed at a large number of measurement points and a controller for detecting tactile information at each measurement point from these outputs is reduced. It is desirable to be able to do it.

  An object of the present invention is to propose a tactile information detection method and a tactile sensor system that can automatically adjust the gain of each strain gauge according to the contact force acting on each strain gauge (sensor element) arranged at each measurement point. There is.

Another object of the present invention is to propose a tactile information detection method and a tactile sensor system that can reduce the number of wires between the tactile sensor and the controller in addition to the above-described problems.

The present invention provides a tactile sense that detects a tactile information such as a contact force acting on each measurement point based on an output from each bridge circuit by configuring a bridge circuit for each measurement point by a strain gauge attached to each measurement point. In the information detection method,
Generate a mixed sine wave signal containing sine wave components with different frequencies,
By applying this mixed sine wave signal to each bridge circuit through a bandpass filter, a single frequency sine wave signal determined in advance is applied to each bridge circuit,
Add the outputs obtained from each bridge circuit to generate an added output,
From this added output, by using the orthogonality of the trigonometric function, at least the contact force is obtained from the contact force and the contact direction acting on each measurement point,
The voltage amplitude measured at each measurement point is compared with the target voltage set for each measurement point in advance, and the sine of each frequency applied to the bridge circuit at each measurement point to suppress these errors The gain amplitude of the bridge circuit at each measurement point is controlled by adjusting the voltage amplitude of the wave signal.

Further, the present invention is a tactile information detection system for detecting tactile information such as contact force acting on each measurement point by the above-described tactile information detection method,
A tactile sensor, a controller, a signal output line for supplying the controller with the added output output from the tactile sensor, and a mixed sine wave signal output from the controller for supplying to the tactile sensor A gain control line,
The tactile sensor includes a plurality of sensor units and an addition circuit that adds the outputs of the sensor units and generates the addition output.
Each sensor unit applies the bridge circuit composed of a plurality of strain gauges arranged at measurement points and a sine wave signal having a predetermined single frequency included in the mixed sine wave signal to the bridge circuit. And a bandpass filter for
The controller acts on each measurement point using an orthogonality of a trigonometric function from an AD converter that AD converts the addition output supplied via the signal output line and the addition output after AD conversion. Among the contact force and its contact direction, the analyzer that calculates at least the contact force, the voltage amplitude measured at each measurement point and the predetermined target voltage are compared, and these errors are suppressed, An automatic gain control circuit that adjusts the voltage amplitude of a sine wave signal of each frequency applied to each sensor unit of the tactile sensor, and generates a mixed sine wave including a sine wave of each frequency with the adjusted voltage amplitude Output DA converter.

As described above, according to the present invention, a variable sensitivity type that can individually control the gains of a plurality of sensor elements (sensor units) while the interface between the tactile sensor and the controller is one input and one output. A tactile information detection system can be realized.

  That is, by making the interface between the sensor unit and the controller one input and one output, simultaneous measurement of each measurement point and simultaneous gain adjustment of a detection signal obtained from each measurement point can be performed.

  In addition, the force information acting on each measurement point can be measured in real time by simple signal processing in the analyzer.

Furthermore, since the signal intensity from the tactile sensor can always be kept within a certain range, it is possible to prevent the tactile sensor from being saturated and the resolution of the tactile sensor from being extremely lowered. Therefore, appropriate sensing can always be performed.

  Hereinafter, an automatic sensitivity variable tactile information detection system to which the present invention is applied will be described in detail with reference to the drawings.

  FIG. 1 is an overall configuration diagram showing a tactile information detection system, and FIG. 2 is a block diagram showing a circuit configuration of the tactile sensor. The tactile information detection system 1 includes a strain gauge type tactile sensor 2 and a controller 3. The tactile sensor 2 includes a sensor substrate 4, a plurality of sensor units 5i (i = 1, 2, 3,...) Arranged in a matrix on the surface of the substrate 4, and a polymer gel covering the sensor unit 5i. And a detection surface 6a formed of a flexible material 6 such as the above.

  The sensor unit 5 generates a differential signal from the bridge circuit 53 constituted by a pair of strain gauges 51 and 52 arranged orthogonally at a position corresponding to the measurement point of the detection surface 6 a and the output of the bridge circuit 53. And a band-pass filter (BPF) 55 for applying an input signal having a specific wavelength to the bridge circuit 53. The tactile sensor 2 includes an adder circuit 56 that synthesizes the outputs of the sensor units 5 (outputs of the differential amplifier 54), and an amplifier 57 that amplifies the synthesized signal obtained by the adder circuit 56. The output is supplied to the controller 3 via a single signal output line 58. On the other hand, the controller 3 includes an AD converter 7, an analyzer 8, an automatic gain control circuit (hereinafter referred to as AGC) 9, and a DA converter 10. A gain control signal output from the DA converter 10 is supplied to each sensor unit 5 of the tactile sensor 2 via a single gain control line 11.

  When the number of measurement points increases, for example, a plurality of sets of tactile sensors 2 may be arranged in parallel to the controller 3 as shown in FIG.

  The tactile information detection system 1 can simultaneously measure tactile information (contact force in this example) and gain control of each sensor unit 5 of the tactile sensor 2. For gain control, the DA converter 10 of the controller 3 outputs a sum of sine waves having different frequencies (mixed sine wave signal) as a gain control signal y (t). In each sensor unit 5 placed at each measurement point, a gain control signal including only a sine wave component of a predetermined frequency is applied to the bridge circuit 53 via a band pass filter (BPF) 55. As a result, gain adjustment is performed as described later.

  On the other hand, the amplitude Ai of the signal Vi (t) output from the bridge circuit 53 of the sensor unit 5 i is proportional to the force input to the sensor unit 5 i, and the respective outputs are multiplexed by the adder circuit 56. Therefore, detection outputs at a plurality of measurement points can be collectively AD converted in the AD converter 7 of the controller 3. With this one-input / one-output configuration, the controller 3 can easily measure tactile information such as contact force at a plurality of measurement points and adjust the gain via the DA converter 10.

Next, an outline of processing performed by the analyzer 8 and the AGC 9 of the controller 3 will be described. The analyzer 8 performs processing equivalent to Fourier transform for each i channel in order to calculate force information at each measurement point in real time. In the AGC 9, gain control is continuously performed so that the signal from the sensor unit 5i of the touch sensor 2 becomes a reference value (target value). The role of the AGC 9 is to keep the signal intensity as constant as possible within a certain range and avoid an unstable state in which measurement cannot be performed. For example, if an extremely large contact force is applied to a specific measurement point, the AGC 9 performs control so that the amplitude Ai of the sine wave that controls the gain of the sensor unit 5i arranged at the measurement point becomes small. As a result, the output signal Vi (t) from the sensor unit 5i is maintained at a value within a predetermined range. Further, when a small contact force acts on the measurement point, the AGC 9 increases the gain of the corresponding sine wave by increasing the amplitude Ai so that the signal Vi (t) from the sensor unit 5i falls within a predetermined range. increase.

  The concept of automatic gain control has already been implemented in CCD cameras, MIC amplifiers, etc., and when a signal with a strong level exceeding the specified level is input or when a signal with a weak level less than the specified level is input, The intensity level is kept within a certain range. In the present invention, from the viewpoint of measurement, a new tactile information detection system 1 incorporating an AGC function is realized.

FIG. 4 is a block diagram showing a feedback loop of the i-th sensor unit 5i. In this figure, a signal flow between the tactile sensor 2 and the controller 3 is shown. In order to simultaneously control the gain of the sensor unit 5 i arranged at the measurement point from the controller 3 side, a sum of sine waves having different frequencies is output via the DA converter 10. This mixed sine wave signal y (t) can be expressed as shown in Equation (1).

However, A _i and f _i are the i-th voltage amplitude and frequency. A mixed sine wave signal y (t) as shown in Equation (1) is applied to the bridge circuit 53 at each measurement point, but a single frequency previously determined by the bandpass filter 55 in the bridge circuit 53. Only the sine wave component of fi is added. Therefore, when only A _i sin (2πf _i t) is applied to the i-th bridge circuit 53 and a force is applied to the strain sensors 51 and 52, the output voltage V _i (t) from the i-th sensor unit 5i is Equation (2) is obtained.

Where G _i is the gain of the differential amplifier 54, φ _i is a phase shift from the applied frequency, ΔR _i is the amount of change in resistance of the strain gauges 51 and 52 due to contact force, and R is the balance resistance of the bridge circuit 53. . As can be seen from this equation, since the gain G _i of the differential amplifier 54 is constant, the gain of the sensor unit 5 _i is equivalently changed by changing the voltage amplitude A _i applied to the bridge circuit 53 from the controller 3 side. Will be changing. The output signal V _i (t) from each measurement point is multiplexed by the adder circuit 56 and can be simultaneously measured by the controller 3. Therefore, the input signal V _input (t) taken into the controller 3 can be expressed as in Expression (3).

However, V _input | _max is the maximum input voltage of the AD converter 7 or the like. The force applied to each measurement point is calculated by the analyzer 8, but a force exceeding the maximum input voltage cannot be calculated. Therefore, the gain control is appropriately performed by the AGC 9 so that the intensity level of the signal V _i (t) is kept within a certain range. By forming feedback for gain control in this way, signal saturation can be prevented even if a strong force exceeding a specified value is input to the sensor unit 5i. When a small force is input to the sensor unit 5i, The resolution can be increased by increasing the gain to a level within the specified range. That is, the intensity level of the signal can be kept within a certain range, and an unstable state where measurement cannot be performed can be suppressed. The analyzer 8 and AGC 9 will be described in more detail below.

(analyzer)
As shown in FIGS. 4 and 5A, the output V _i (t) from the sensor unit 5i is amplitude-modulated according to the force acting on the measurement point, as represented by the equations (2) and (3). Frequency components included. Therefore, in order to obtain the force acting on the measurement point, demodulation is performed as follows. Since the frequency (carrier wave) applied to each measurement point is known in advance, it is only necessary to obtain the relationship between the necessary frequency and amplitude. First, as in Expressions (4) and (5), a correlation between a sine wave and a cosine wave is obtained for an output V _sum (t) obtained by AD conversion of the output V _input .

If V _i (t) and Y _y (t) obtained by applying V _x (t) and V _y (t) obtained from the above formulas to X _i (t) and Y _i (t), respectively, the required frequency amplitude, that is, contact force F _i (t) can be written as:

However, d _i is a constant determined by calibration, and phase information Phase _i (t) indicates a contact direction. The phase information indicates, for example, whether the contact is from above or below from the sensor unit 5i. For example, as shown in FIG. 5B, when the phase information Phase _i (t) is positive, it means that the contact direction is downward, and as shown in FIG. 5C, when the phase information Phase _i (t) is negative. Means that the contact direction is upward. Components other than the frequency f _i are cut except for the cut-off frequency F _cut Hz of the LPF due to the orthogonality of the trigonometric function. The cut-off frequency f _cut is determined by how much the input to the tactile sensor 2 is made, and it is necessary to satisfy the following equation.

If it is desired to detect vibration or even a high frequency, the value of the frequency f _i may be increased.

(Automatic gain control)
The purpose of the AGC 9 is to automatically prevent saturation of the AD converter 7 and the like and change the resolution of the contact information. Referring to FIGS. 4 and 6, the AGC 9 performs the following operation to change the voltage amplitude A _i (t) applied to the sensor unit 5i to an appropriate value.

Where A _ri (t) is the voltage amplitude measured at the i-th measurement point, A _io (t) is the target value at the i-th measurement point, and E _i (t) is the error from the i-th target voltage. , ΔW _i (t) indicates the correction amount of the applied voltage. α is a small constant. Equation (10) has an effect of cutting high frequencies by integration, and an error that changes frequently is changed gently.

The voltage amplitude A _i (t) is updated according to the correction amount ΔW _i (t) of the applied voltage by rewriting the memory of the DA converter 10. In general, a combination with a memory is conceivable for generating a sine wave at high speed by the DA converter 10, but the memory rewrite time cannot be ignored. For example, when it is desired to output the ideal output waveform (a) shown in FIG. 7, the actual output waveform from the DA converter 10 is a waveform obtained by multiplying the ideal output waveform (a) and the square wave (c). (A). In the figure, Ti is the memory rewrite time of the DA converter 10, and T is the rewrite update cycle.

(Experimental example)
As shown in FIG. 8, a tactile sensor 2 having a configuration in which a steel plate 21 was cut by 1 mm, divided into a plurality of regions, and a sensor unit 5 was arranged in each region. The sensor unit 5 includes a bridge circuit 53 configured using two strain gauges 51 and 52 attached to measurement points, and performs temperature compensation of the output. From the DA converter 10 of the controller 3, a mixed sine wave signal y (t) was generated by software, and the analog output was updated at 30 kHz. A single sine wave was applied to the bridge circuit 53 at each measurement point via the analog BPF 55. In order to apply only a single sine wave to the bridge circuit 53, a secondary bicut type BPF 55 that can increase the quality factor (a factor that determines the sharpness of the cut-off characteristic) is employed. Since it was difficult for the BPF 55 to accurately adjust the center frequency according to the set value due to variations in elements, the frequency of the sine wave output from the DA converter 10 was adjusted to the center frequency of the BPF 55. Each analog BPF 55 was designed so that the portions that interfere with each other attenuate 100 dB or more. The output from each measurement point was amplified approximately 1000 times by the instrumentation amplifier 54. In order to make one output signal line from the touch sensor 2, the output from each sensor unit 5 was added by the adder circuit 56 to generate an amplitude-modulated frequency multiplexed signal. The output signal from the touch sensor 2 was sampled at 5 kHz by the AD converter 7. In the analyzer 8 for obtaining the correlation between the sine wave and the cosine wave, the digital LPF having a cutoff frequency of 50 Hz used here is designed with a third-order Putterworth characteristic. In the experiment, in order to set τ / T to approximately 1.0, the gain update cycle T was set to 250 ms, and the experiment was performed at τ / T = 0.80. However, in actual measurement, the measurement was performed in a stable state after the memory rewrite update.

FIG. 9 shows experimental results when automatic gain control is not performed on two measurement points. A sine wave of 313 Hz and 604 Hz is applied from the DA converter 10 to each sensor unit (strain gauge). is there. In the measurement, after first touching one of the measurement points to which 604 Hz was applied, two were touched simultaneously. FIG. 9A is a raw waveform of the output from the touch sensor 2, and FIGS. 9B and 9C are data after processing by the analyzer 8. In these figures, the sensor unit to which a 604 Hz sine wave was applied at t = 0.9-1.3 sec was touched, and both sensor units were touched at t = 1.5-1.9 sec. I can read. In FIG. 9B, a small output appears even though there is no input signal during 0.9-1.2 sec. This is considered to be because the attenuation at the frequency of 313 Hz is not sufficient when the applied voltage and the input displacement are large in the BPF 55 having the center frequency of 604 Hz.

  FIGS. 10A and 10B show how the tactile sensor 2 with a constant gain is saturated when automatic gain control is not performed. FIG. 10A shows an output signal from a single measurement point, and FIG. 10B shows an applied voltage from the DA converter 10 supplied to the single measurement point. In FIG. 10A, it can be seen that the AD converter 7 is saturated after t = 3900 ms.

11A to 11D show how the gain of the tactile sensor 2 decreases when the contact force is gradually increased when automatic gain control is performed. 11D shows an output signal from a single measurement point, and FIGS. 11A and 11B are enlarged views of parts of FIGS. 11C and 11D, respectively. In FIG. 11A, T is a gain update period of AGC 9, Td is an effective sensor measurement period, and Ti is a memory rewrite time of the DA converter 10. For the data after Ti, a measurement period Td is provided after waiting for a stable state in consideration of a rise delay due to BPF or LPF. From FIG. 11C and 11D, it can be seen that the gain of the sensor unit 5 decreases according to the contact force.

It is a schematic block diagram which shows the tactile information detection system to which this invention is applied. It is a schematic block diagram which shows the circuit structure of the sensor unit in the tactile information detection system of FIG. It is a block diagram which shows an example of the tactile information detection system with which the tactile sensor of FIG. 1 is equipped with two or more sets. It is a schematic block diagram which shows the signal feedback loop of the i-th sensor unit of FIG. It is a schematic block diagram which shows the processing operation of the analyzer of FIG. It is explanatory drawing which shows the meaning of the phase information obtained with an analyzer. It is explanatory drawing which shows the meaning above the phase obtained with an analyzer. It is a schematic block diagram which shows the processing operation of AGC of FIG. It is explanatory drawing which shows the output signal waveform from the DA converter of FIG. It is explanatory drawing which shows the structure of the tactile sensor of the tactile sensing system used for the experiment for confirming the effect of this invention. It is a graph which shows the output signal waveform of the tactile sensor obtained when not performing automatic gain control. It is a graph which shows the output signal waveform of the tactile sensor obtained when not performing automatic gain control. It is a graph which shows the output signal waveform of the tactile sensor obtained when not performing automatic gain control. It is a graph which shows the output signal waveform of a tactile sensor when not performing automatic gain control. It is a graph which shows the input signal with respect to the tactile sensor in the case of FIG. 10A. It is a graph which shows the output signal waveform from a tactile sensor at the time of performing automatic gain control. It is a graph which shows the input signal with respect to a tactile sensor. It is a graph which shows the output signal waveform from a tactile sensor at the time of performing automatic gain control. It is a graph which shows the input signal with respect to a tactile sensor.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Tactile information detection system 2 Tactile sensor 3 Controller 5i Sensor unit 51, 52 Strain gauge 53 Bridge circuit 54 Differential amplifier 55 Band pass filter 56 Adder 57 Amplifier 58 Signal output line 6a Detection surface 7 AD converter 8 Analyzer 9 AGC
10 DA converter 11 Gain control line

Claims (2)

In a tactile information detection method for detecting tactile information such as contact force acting on each measurement point based on an output from each bridge circuit by configuring a bridge circuit for each measurement point by a strain gauge attached to each measurement point. ,
Generate a mixed sine wave signal containing sine wave components with different frequencies,
By applying this mixed sine wave signal to each bridge circuit through a bandpass filter, a single frequency sine wave signal determined in advance is applied to each bridge circuit,
Add the outputs obtained from each bridge circuit to generate an added output,
From this added output, by using the orthogonality of the trigonometric function, at least the contact force is obtained from the contact force and the contact direction acting on each measurement point,
The voltage amplitude measured at each measurement point is compared with the target voltage set for each measurement point in advance, and the sine of each frequency applied to the bridge circuit at each measurement point to suppress these errors A tactile information detection method for controlling the gain of a bridge circuit at each measurement point by adjusting the voltage amplitude of a wave signal. A tactile information detection system for detecting tactile information such as contact force acting on each measurement point by the tactile information detection method according to claim 1,
A tactile sensor, a controller, a signal output line for supplying the controller with the added output output from the tactile sensor, and a mixed sine wave signal output from the controller for supplying to the tactile sensor A gain control line,
The tactile sensor includes a plurality of sensor units and an addition circuit that adds the outputs of the sensor units and generates the addition output.
Each sensor unit applies the bridge circuit composed of a plurality of strain gauges arranged at measurement points and a sine wave signal having a predetermined single frequency included in the mixed sine wave signal to the bridge circuit. And a bandpass filter for
The controller acts on each measurement point using an orthogonality of a trigonometric function from an AD converter that AD converts the addition output supplied via the signal output line and the addition output after AD conversion. Among the contact force and its contact direction, the analyzer that calculates at least the contact force, the voltage amplitude measured at each measurement point and the predetermined target voltage are compared, and these errors are suppressed, An automatic gain control circuit that adjusts the voltage amplitude of a sine wave signal of each frequency applied to each sensor unit of the tactile sensor, and generates a mixed sine wave including a sine wave of each frequency with the adjusted voltage amplitude A tactile information detection system provided with a DA converter for outputting.

Images