JPH07132472A - Approach method by plural distance sensors - Google Patents

Abstract

PURPOSE:To enable a controlled body to approach the other face such as the ground surface while being controlled appropriately even if the other face is uneven or wavy, or part of distance sensors is in failure. CONSTITUTION:A controlled body 1 is provided with plural distance sensors 2 for measuring the distance to the approaching other face 3. The relative specific measured value (the minimum value, for instance) is selected from the measured value of these distance sensors 2, and this relative specific measured value is used as control information for the approach of the controlled body 1 to the approaching other face 3.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method of approaching an object to be controlled by a plurality of distance sensors which is applied to leg control of a legged robot.

[0002]

2. Description of the Related Art For example, a legged robot 61 which walks trot as shown in FIG. 6 (a gait in which two diagonally paired legs are switched one after another) is provided with four legs 62, two legs forward and backward. It is provided. The conventional method of bringing the legs 62 (controlled body) closer to the ground 63 (opposite surface) is disclosed in Japanese Patent Laid-Open No. 61-193973 for measuring the distance between the legs 62 and the ground 63 on each of the legs 62. One non-contact distance sensor (not shown) such as that disclosed in Japanese Patent No. 6-32242 is provided, and the ground approaching speed of the leg 62 is determined by the distance sensor information to determine the leg 62.
Was controlling his movements.

[0003]

If the ground to be walked is flat or close to a flat surface, one distance sensor for each leg 62 can be provided to control a substantially satisfactory approach speed to the ground. However, the normal ground as seen outdoors has unevenness and undulations, and a single distance sensor causes a distance measurement error. In addition, the output of the distance sensor may fluctuate due to the nature of the ground surface (rock, mud, water), extraneous light, etc., which may cause a distance measurement error. Further, even if the distance sensor fails, the distance measurement becomes impossible, and for example, even if a problem such as always instructing the maximum distance to the control unit occurs, it is impossible to cope with the case where there is only one distance sensor. There was a problem.

The present invention has been made to solve these problems. For example, in controlling the legs of a robot, even if there is unevenness or undulations on the mating surface such as the ground,
It is an object of the present invention to provide an approach method using a plurality of distance sensors that can appropriately control a leg that is a controlled object and can approach a partner surface even if a part of the distance sensor fails.

[0005]

According to the approach method using a plurality of distance sensors of the present invention, a plurality of distance sensors for measuring the distance to an approaching partner surface are provided on a controlled object, and relative distances are selected from the measured values by the distance sensors. It is characterized in that a target specific measurement value (for example, a minimum value) is selected and this relative specific measurement value is used as control information for approaching the approaching partner surface of the controlled object. Further, the measurement value of the distance sensor is a non-continuous value (for example, large,
It is also characterized by being medium and small).

[0006]

With this configuration, even if there is undulation or unevenness on the mating surface, a relatively most appropriate measurement value (for example, a distance closest to the mating surface) is selected from a plurality of measurement values to be controlled. Control the body. Further, for example, even if one of the distance sensors fails, another distance sensor can handle it. Further, even if a distance sensor whose measured value is a discontinuous value is used, the same effect can be obtained, and the distance sensor can be simplified.

[0007]

Embodiments of the present invention will be described below with reference to the drawings. In the first embodiment, the leg of a legged robot is taken as an example of the controlled object, and the leg of this robot is controlled by the approaching method of this embodiment based on the measurement value of the distance sensor. They are arranged as shown in FIGS. 1 (a) and 1 (b).

In FIG. 1 (a), reference numeral 1 is a leg of a legged robot which is a controlled object, and 2 is provided vertically downward on the surface of a disc-shaped sole 1a of the leg 1, It is a plurality of distance sensors whose measured values of the distance to the ground 3 are obtained as continuous analog values. The distance sensor 2 is composed of a transmitter (not shown) that sends a distance measurement signal such as ultrasonic waves or infrared rays to the ground 3 direction, and a receiver (not shown) that receives the measurement signal reflected and returned from the ground. There is. In this embodiment, as shown in FIG. 1 (b), one distance sensor 2 is provided at the center of the back of the leg 1a, and the other distance sensors 2 have the structure of the leg 1 and generate disturbances in the direction of the bottom twist. Assuming a mechanism with a degree of freedom of sliding to suppress the influence of moment, and further, in order to simplify the sensor signal processing without using a slip ring, a total of eight equidistant parts are arranged on the circumference of the sole 1a. Nine distance sensors 2 are arranged.

In the approaching method of this embodiment, the leg 1 having such a structure is controlled by the approaching speed V proportional to the ground distance value Lg obtained by measuring with the nine distance sensors 2. There is.

V = K · Lg K: constant …… 1 Distance from the ground 3 to the ground 3 where the leg 1 is about to touch the ground
There are an average method and a closest data method as a calculation method of. The averaging method arithmetically averages the measured distance values of the total distance sensor 2 as shown in Equation 2 to obtain the ground distance value Lg of the leg 1. The closest data method adopted by the present invention is As shown in Expression 3, the smallest measured value of the constituent distance sensors 2 is used as the distance value Lg of the leg 1 to the ground with the measured value of the distance sensor 2.

Average method Lg = Lave = ΣLi / n 2 Li: Measurement value of i-th distance sensor 2 (i = 1 to 9) n: Number of distance sensors 2 Closest data method Lg = Lmin = Lx. 3 Lx: Measured value of the distance sensor 2 that measured the smallest measured value Therefore, the approaching speed V of the leg 1 in the case of both methods is 1
It is expressed by the equations 4 and 5, respectively.

In the case of the averaging method V = K · Lave ...... 4 In the case of the closest data method V = K · Lmin …… 5 FIG. 2 shows various methods for comparing and evaluating the two methods of the averaging method and the closest data method. Experiments were carried out on the grounds of shapes (a) to (d) using the two methods, and the average and variance of the approach speed V when the leg 1 touches the ground 3 are plotted with the number of distance sensors 2 on the horizontal axis. It is a graph in which the data of the average method is shown by a solid line and the data of the closest data method is shown by a broken line.

The condition (a) is that the ground 3 has a sinusoidal two-dimensional undulation as shown in the same figure, and the undulation surface is expressed by the following equation. Z = h COS (2π · x / D) In this embodiment, D = 480 mm (four times the diameter of the leg 1) and h = 9.4 mm is shown.

The condition (b) is that the ground is a convex surface, and the convex surface is expressed by the following equation. Z = h COS (2π · x / D) COS (2π · y / D) In this embodiment, the case of D = 960 mm and h = 152 mm is shown.

The condition (c) is that the ground surface is a concave surface and the concave surface is expressed by the following equation. Z = −h COS (2π · x / D) COS (2π · y / D) In this embodiment, the case of D = 960 mm and h = 152 mm is shown.

In these equations, Z is the ground height, h is the amplitude of the undulations and irregularities, D is the wavelength of the undulations and irregularities, and x and y are the coordinates of the horizontal plane. The condition of (d) is a case where there are projecting surfaces with a width of 24 mm on the ground, and three projecting surfaces with heights of 6 mm, 12 mm, and 24 mm are randomly distributed.

As can be seen from the graphs of FIG. 2, the average of the approaching speeds V is 3 to 3 in the case of the closest approach data method.
In the case of the requirement that the approach speed V is decelerated (the number of sensors is 3 to 5 or more), the maximum deceleration is 5 or more, and the dispersion of the approach speed V when the leg 1 comes into contact with the ground 3 is the maximum. It can be seen that the approach data method is smaller than or similar to the average method, and the approach characteristics are superior. Further, even if the number of the distance sensors 2 is increased, the approach speed V cannot be reduced to the permissible speed (V / Vmax is 0.1 or less) by the averaging method, but the number of sensors is about 9 by the closest approach data method, and the permissible speed is almost Can be brought close to the neighborhood.

3, the number and arrangement of the distance sensors 2 on the leg 1 and the condition of the ground 3 are the same as those in FIG. 2, but the approach of the leg 1 when the distance sensor 2 has a certain distance measurement error. It is data of the average and variance of the velocity V. As can be seen from the graphs in Fig. 3, the average of the approach velocity V is not much different from the data in Fig. 2, and in the case of dispersion, it is closest to the averaging method under conditions (b) (c) (d). Although both data of the data method are close to each other, in the case of the condition (a), the closest approach data method is required under the condition that the approach speed V can be decelerated as in the case of FIG. 2 (the number of sensors is 3 to 5 or more). Is smaller than the averaging method, indicating that the closest data method has better approach characteristics.

Considering the experimental results shown in FIGS. 2 and 3, the averaging method shows that the approach speed V is obtained when the unevenness of the ground is small.
However, if the irregularities are large, the merit of averaging is lost. When the unevenness is large, the closest approach data method, which can determine the approaching speed V by the measurement value of the closest distance, is more effective for performing soft landing. Further, it is desirable that the number of the distance sensors 2 is large, but it can be understood that practically about 9 is sufficient.

The second embodiment is on / off as a distance sensor.
This is an example of an approach method using a sensor and using information of this on / off sensor. In this embodiment, as shown in FIGS. 4 (a) and 4 (b), nine on / off sensors 4 are mounted on the leg 1 as in the first embodiment.

The on / off sensor 4 is composed of a pair of ON / OFF type proximity sensors. When the reference distance is Lmax, the sub-reference distance is 0.5 Lmax, and one of the two proximity sensors is measured. The threshold is set to Lmax and the other measurement threshold is set to 0.5 Lmax. That is, one proximity sensor outputs ON when the measured value is less than Lmax and OFF when the measured value is more than Lmax, and the other proximity sensor outputs 0.5 measured value.
ON when less than Lmax, OFF when more than 0.5 Lmax
Is set so that the ON / OFF sensor 4 combines the ON / OFF information of the two proximity sensors to output a discrete output to the control unit.

The individual on / off sensors 4 of this embodiment are shown in FIG.
As shown in (c), when both of the two proximity sensors that are configured are both OFF, the first output is output, and when the approach speed V of the leg 1 to the ground is Vmax, the second output is output when only one is ON. broth,
The approach speed V is set to 0.5 Vmax, and when both are ON, the third output is output and the approach speed V is set to 0.1 Vmax.

The above is the case where there is only one on / off sensor 4, but in this embodiment nine on / off sensors 4 are provided, and the ground of the leg 1 is the same as in the first embodiment. The closest approach data method is adopted as the approach method. In other words, 9 on / O
When all the proximity sensors of the FF sensor 4 are OFF, that is, when only the first output is provided, the approach speed V of the leg 1 is set to Vmax.
In the case where ON and OFF are mixed, there is no third output, and when the first and second outputs are mixed, the second output is prioritized and the approach speed V is set to 0.5 Vmax, and when the third output is mixed. Third
Set the approach speed V to 0.1 Vmax, giving priority to the output. Thus, the approach speed V in this embodiment is set discretely. The method of the second embodiment is called the closest data + fuzzy method in order to distinguish it from the closest data method of the first embodiment.

Using the leg 1 on which such an on / off sensor 4 is arranged, with respect to the protruding surface shape among various ground shapes,
FIG. 5 shows the result of an experiment conducted to compare the method of this embodiment and the method using the analog output distance sensor according to the first embodiment. FIG. 5 shows the average and variance data of the approach velocity V when the leg 1 comes into contact with the ground 3, the average method by the analog distance sensor of the first embodiment is shown by a solid line, and the closest approach data method is shown by a broken line. The closest data + fuzzy method of the second embodiment is indicated by a chain double-dashed line. 5
Figure (a) shows the state where all the sensors are operating normally, and Figure (b) shows the case where the central sensor outputs the maximum output due to disconnection. Further, in the experimental ground shape, the pitch between the projections is 60 mm, and the height h of the projections is 0 ≦ h ≧ 60 mm.

As can be seen from the graphs in FIG. 5, the average of the approaching speeds V has a strict projection surface shape. Therefore, even the closest approaching data or the closest approaching data + fuzzy method can meet the requirement that the approaching speed V can be reduced. Requires about nine sensors, but when this requirement is satisfied, the variance of the closest approach data + the approach speed V of the fuzzy method is similar to the variance of the closest approach data method by the analog distance sensor. Was given. This is
As can be seen from Fig. 5 (b), the same tendency can be seen when the central sensor outputs the maximum output due to the disconnection or the like.

Considering the experimental results shown in FIG.
It can be seen that the closest approach data + fuzzy approach method based on the discrete output information of the embodiment of the second embodiment achieves the same level of performance as the closest approach data method of the analog output information of the first embodiment. Further, the precision is inferior to the case where a small number of high-accuracy analog sensors of the first embodiment are used, but the required number of discrete output sensors of the second embodiment provided a smaller dispersion of the approach speed. It was found to be an excellent method for performing soft landing when applied to a legged robot. The present invention is not limited to the above-mentioned embodiments, and can be modified and carried out without changing the gist.

[0027]

According to the present invention, the controlled object is controlled by selecting the most appropriate relative specific measured value for approaching the opponent surface from the plurality of distance measured values, so that the controlled object can be controlled appropriately. It can be brought into contact with the opponent's surface at an approaching speed. Further, according to the approach method of the present invention, the controlled object can be appropriately controlled even if a simple sensor whose measured value is a discontinuous value is used for the plurality of distance sensors.

[Brief description of drawings]

FIG. 1 is an explanatory diagram of an arrangement distribution of distance sensors in a controlled body for explaining a method according to a first embodiment of the present invention.

FIG. 2 is data of average and dispersion of approach speeds when the controlled object shown in FIG. 1 is controlled by the average method for comparison with the closest approach data method of the present invention.

FIG. 3 is data of average and variance of approach speeds when the distance sensor having an error is mounted on the controlled object shown in FIG. 1 and the closest approach data method and the average method performed under the ground condition of FIG. 2 are used.

FIG. 4 is an explanatory diagram in which an on / OFF sensor is used as a distance sensor and is arranged on a controlled object in order to explain the method of the second embodiment.

FIG. 5 is data of average and dispersion of approach speeds when the method of this example and the method of the first example are compared on a projecting surface.

FIG. 6 is an explanatory diagram showing a walking state of a legged robot.

[Explanation of symbols]

 1 ... leg 1a ... sole of foot 2 ... distance sensor 3 ... ground 4 ... on / OFF sensor

Claims (2)

[Claims] 1. A controlled object is provided with a plurality of distance sensors for measuring a distance from an approaching partner surface, a relative specific measured value is selected from the measured values by the distance sensors, and the relative specific measured value is calculated. An approach method using multiple distance sensors, which is characterized by using control information for approaching the approaching partner surface of the controlled object. 2. The approaching method using a plurality of distance sensors according to claim 1, wherein the measured value of the distance sensor is a discontinuous value.