JPH052421A - Controller for unmanned carrier - Google Patents

Abstract

PURPOSE:To enable an unmanned carrier to reach a target point with high accuracy by compensating effectively the track error caused by a dynamic fluctuation error of the effective value of the unmanned carrier body produced during the run of the carrier. CONSTITUTION:The unmanned carrier runs with the right and left driving wheels at a target speed and in a target curvature. A learning storage means III computes the dynamic error variance value between the radius and the tread length of the driving wheel based on the target and actual speeds and the curvature and stores the error variance value. A compensating means IV computes the compensation values of the carrier speed and the carvature from the target speed ant the error variance value. A conversing means V computes the driving speed command values of the right and left driving wheels from the compensation values obtained by the means IV and the prescribed tread length. Then a driving means VI controls the driving wheels based on the driving speed command values.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a control system for an automated guided vehicle, and more particularly, to compensate for a dynamic variation error of the automated guided vehicle during traveling to accurately drive the automated guided vehicle on a target track. The present invention relates to a control device.

[0002]

2. Description of the Related Art Conventionally, a road map is stored in advance,
There is an automatic guided vehicle that travels while checking its own position on the map by detecting a target point installed at a key point on the travel path or calculating a travel distance.

When this automatic guided vehicle passes near or on the target point or is positioned at the target stop position, conventionally, the traveling route between the two is smoothly continuous due to the positional relationship between the target position and the automatic guided vehicle. Generate as a curve trajectory,
(A) When the error from the target trajectory becomes larger than a predetermined threshold, discard the target trajectory up to that point and create a new target trajectory from that position to the target position, or (b) target The driving means of the automatic guided vehicle was controlled so that the error from the track was fed back to the control input of the motor for driving or steering, so that it travels on the generated track and reaches the target point. ..

[0004]

Generally, a cause of the track error is a dynamic fluctuation of the effective value of the vehicle body constant of the automatic guided vehicle which occurs during traveling. That is, since the vehicle body constant set value in the control program and the effective value do not match during running,
That is, the drive means does not travel according to the generation trajectory.

That is, in the case of the above method (a), a new trajectory which is smoothly and continuously connected to the previously generated target trajectory is generated, and the original target trajectory is restored along the new trajectory. By doing so, the error is compensated, but since the dynamic fluctuation error that occurs during traveling with respect to the corrected target trajectory is not compensated, it is difficult for the driving means to follow accurately. However, there was a problem that the target point could not be reached as a result. Further, since a large amount of calculation time is required for trajectory correction, there is a problem that a high speed computer is required.

Further, in the case of the above-mentioned method (a), it is difficult to predetermine an appropriate negative feedback gain with respect to the fluctuation of the vehicle body constant effective value during traveling, so that the target point cannot be reached as a result. There was a problem. Further, since this negative feedback gain is not the same for each vehicle body of the automatic guided vehicle, it has been necessary to design it individually for each vehicle body change.

The present invention has been made in view of the above-mentioned problems of the prior art.
A controller for an automated guided vehicle that can accurately reach a target point by effectively compensating for a trajectory error based on a dynamic variation error of the vehicle body constant effective value of the automated guided vehicle that occurs during traveling. The purpose is to provide.

[0008]

According to a first aspect of the present invention, there is provided an automatic guided vehicle control apparatus which outputs a target trajectory, which is to be traveled by an automated guided vehicle having two left and right driving wheels, as a target vehicle speed and curvature. Based on the target vehicle speed and the target curvature from the generation means and the target trajectory generation means, a dynamic variation error of the automatic guided vehicle that occurs during traveling is compensated and a target vehicle speed compensation value and a target curvature compensation value are output. Compensation means, conversion means for calculating a drive speed command value for the left and right drive wheels from the target vehicle speed compensation value and target curvature compensation value, and the left and right drive wheels based on the calculated left and right drive speed command values. And a driving unit that controls driving.

The automatic guided vehicle of the first aspect of the invention travels straight or turns according to the respective driving speeds of the two drive wheels disposed on the left and right of the vehicle body, that is, the rotation speeds. The target track generation means outputs the track on which the automatic guided vehicle should travel as the target vehicle speed and curvature. The compensating means compensates the dynamic fluctuation error of the traveling automatic guided vehicle on the basis of the target vehicle speed and the target curvature given in advance so that the actual vehicle speed and the curvature become the target vehicle speed and the target curvature, respectively. Compensation values for vehicle speed and target curvature are calculated. The conversion means is based on the calculated target vehicle speed compensation value and target curvature compensation value,
A drive speed command value for the left and right drive wheels is calculated. The driving means is
The left and right drive wheels are drive-controlled based on the calculated left and right drive speed command values.

[0010]

As described above, the present invention compensates a dynamic variation error of an automatic guided vehicle that occurs during traveling with respect to a predetermined target vehicle speed and curvature, and a vehicle speed compensation value as a new target vehicle speed and curvature. And the curvature compensation value are calculated, and the left and right drive wheels are driven and controlled based on the vehicle speed and the curvature compensation value, so that the automated guided vehicle can travel along the target track with high accuracy and in real time with a simple configuration. it can.

[0011]

[Other Inventions] A control system for an automatic guided vehicle according to a second aspect of the present invention includes a target trajectory generating means for outputting a target trajectory that an automatic guided vehicle having one steering drive wheel should travel as a target vehicle speed and curvature. Based on the target vehicle speed and the target curvature from the target trajectory generation means, the dynamic fluctuation error of the unmanned guided vehicle that occurs during traveling is compensated, and a compensation means that outputs a target vehicle speed compensation value and a target curvature compensation value, A conversion unit that calculates a steering command value and a drive speed command value of the steering drive wheel from the target vehicle speed compensation value and the target curvature compensation value, and the steering drive based on the calculated steering command value and drive speed command value. Drive means for steering and driving the wheels.

The unmanned guided vehicle of the second aspect of the invention travels straight or turns according to the steering angle and the driving speed, that is, the rotation speed of one steering driving wheel provided on the left and right center lines of the vehicle body. The target track generation means outputs the track on which the automatic guided vehicle should travel as the target vehicle speed and curvature. The compensating means compensates the dynamic fluctuation error of the traveling automatic guided vehicle on the basis of the target vehicle speed and the target curvature given in advance so that the actual vehicle speed and the curvature become the target vehicle speed and the target curvature, respectively. Compensation values for vehicle speed and target curvature are calculated. The conversion means calculates a steering command value and a drive speed command value of the steering drive wheels based on the calculated target vehicle speed compensation value and target curvature compensation value. The drive means performs steering control and drive control of the steering drive wheels based on the calculated steering command value and drive speed command value.

According to the second aspect of the invention, similarly to the first aspect of the invention, the dynamic fluctuation error of the automatic guided vehicle that occurs during traveling is compensated for the predetermined target vehicle speed and curvature,
The vehicle speed compensation value and curvature compensation value as the new target vehicle speed and curvature are calculated, and the left and right drive wheels are drive-controlled based on the vehicle speed and curvature compensation value. Has the advantage of being able to travel along the target trajectory.

A third aspect of the present invention further embodies the first aspect of the present invention, wherein the compensating means uses the target vehicle speed and the target curvature to dynamically change the radius of the left and right driving wheels and the tread length. It has a vehicle speed compensating means for calculating the vehicle speed compensating value in consideration of it, and a curvature compensating means for calculating the curvature compensating value in consideration of the dynamic fluctuations of the radius of the left and right driving wheels and the tread length from the target curvature. ..

That is, the third aspect of the present invention is such that the compensation values for the target values of the vehicle speed and the curvature are determined in consideration of the dynamic fluctuations of the radii of the left and right driving wheels and the tread length of the automatic guided vehicle during traveling.
Each is calculated independently.

The fourth invention, as shown in FIG.
Target trajectory generating means I for outputting a target trajectory for an unmanned guided vehicle having two left and right driving wheels to travel as a target vehicle speed and curvature, and measuring means II for measuring the actual vehicle speed and curvature of the unmanned guided vehicle. During learning travel, a dynamic error during traveling of the radius of the left and right drive wheels and the tread length based on the target vehicle speed and target curvature from the target trajectory generating means I and the actual vehicle speed and actual curvature from the measuring means II. A vehicle speed compensation value and a curvature compensation value are calculated from a learning storage means III for calculating and storing a variation value, a target vehicle speed and a target curvature from the target trajectory generating means I, and an error variation value from the learning storage means III. Compensation means IV, and conversion means V for calculating the drive speed command value for the left and right drive wheels based on the tread length between the left and right drive wheels of the automatic guided vehicle set in advance from the vehicle speed compensation value and the curvature compensation value. Based on the computed left and right drive speed command value, and has a driving means VI for each drive control of the left and right driving wheels.

That is, in a fourth aspect of the present invention, in addition to the first aspect, a measuring means II for measuring the actual vehicle speed and the curvature of the automatic guided vehicle is further provided, and the measured value of the vehicle speed and the curvature and the target value are more left and right. The dynamic variation value of the driving wheel radius and the tread length during traveling is grasped and stored in advance by learning traveling, and at the time of actual traveling, the compensating means IV, based on the stored dynamic variation value, The vehicle speed and the curvature compensation value are sequentially calculated.

As a result, according to the fourth aspect of the invention, the dynamic fluctuation error can be compensated under a wide range of running conditions.

Here, the calculation contents of the compensating means IV will be described in more detail. Vehicle speed compensation value Vc which is the output value of the compensation means IV
And to the curvature compensation value Rc, the drive command velocity V _L and V _R of the left and right, which is the output value of the converter V, using the tread length TR of the right and left drive wheels, is given by the following equation. _{V L = Vc + TR · Rc} · Vc / 2 ... (1) V R = Vc -TR · Rc · Vc / 2 ... (2)

On the other hand, the actual vehicle speed V and the actual curvature R of the automatic guided vehicle are calculated by the following equations by the dynamic fluctuations ΔV _L and ΔV _R of the diameters of the left and right driving wheels receiving the load of the vehicle body and the dynamic fluctuation ΔTR of the tread length. expressed. ^{_{V = (V L * + V}} R *) / 2 ... (3) R = 2 (V L * -V R *) / {TR * (V L * + V R *)} ... (4) However, V _L ^* _{= ΔV L · V L V R} * = ΔV R · V R TR * = TR + ΔTR ΔV L, ΔV R: dimensionless quantity ΔTR: [m]

Therefore, the expressions (3) and (4) are expressed as follows using the expressions (1) and (2). V = {ΔV _L + ΔV _R + TR · Rc (ΔV _L −ΔV _R ) / 2} Vc / 2 ... (5) R = R1 · R2 / R3 (6) where R1 = 2 / (TR + ΔTR) R2 = ΔV _{L -ΔV R + TR · Rc (} ΔV L + ΔV R) / 2 R3 = ΔV L + ΔV R + TR · Rc (ΔV L -ΔV R) / 2

By the way, in an ideal state, the target vehicle speed Vd and the target curvature Rd output from the target trajectory generating means I are obtained.
Assuming that the actual vehicle speed V and the actual curvature R are the same, the following equations can be obtained by rearranging the equations for Vc and Rc with Vd = V and Rd = R in the above equations (5) and (6). become.

Vc = a * Vd + b * Vd * Rd (7) Rc = (e * Rd + f) / (g * Rd + h) (8) where A, B, E, F, G, and H are It is a coefficient and is expressed by the following equation. a = (ΔV _L + ΔV _R ) / (2ΔV _L ΔV _R ) b = − (TR + ΔTR) (ΔV _L −ΔV _R ) / (4 ΔV _L Δ
_{V R) e = (TR +} ΔTR) (ΔV L + ΔV R) f = -2 (ΔV L -ΔV R) g = -TR (TR + ΔTR) (ΔV L -ΔV R) / 2 h = TR (ΔV L + ΔV R)

Therefore, the compensating means inputs the target vehicle speed Vd and the target curvature Rd, and outputs the vehicle speed compensation value Vc and the curvature compensation value Rc by the calculation of the equations (7) and (8). The dynamic fluctuation amounts ΔV _L , ΔV _R, and ΔTR of the diameters of the left and right driving wheels and the tread length are grasped in advance under various traveling conditions, and are mapped and stored in the learning storage means III.

Alternatively, by using a neural network instead of the above map, accurate vehicle speed and curvature compensation values can be obtained without using a large scale map.

[0026]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS First Embodiment Hereinafter, the present invention will be described with reference to a first embodiment for performing a positioning operation at a stop. FIG. 2 is a schematic configuration diagram of an automated guided vehicle to which the traveling control device according to the first embodiment is applied.

The automated guided vehicle 1 is controlled so as to follow a preset traveling course and is guided to a position just before the stop point. This guiding method is executed as follows.

The amount of rotation of the measuring wheel is detected by the encoder, and the position and direction of the vehicle are calculated by the measuring device.
Furthermore, the running speed, position, and
The unmanned vehicle is guided along the traveling course by controlling the motors that drive the left and right driving wheels so that the respective directional errors e ₀ , e ₁ , e ₂ are minimized.

That is, the evaluation function E _S is set by the expression (9), and the accelerations α _L and α _R of the left and right drive wheels are set by the expressions (10) and (11) so that the evaluation function E _S becomes the minimum. It is calculated by the partial differential equation of. _{E S = k (k 0 e} 0 + k 1 e 1 + k 2 e 2) ... (9) α L = -∂E S / ∂α L ... (10) α R = -∂E S / ∂α R ... ( 11) However, k, k ₀ , k ₁ and k ₂ are preset gain constants.

Since there is a cumulative error in the measured values of the position and direction of the automatic guided vehicle by the encoder, immediately before the stop position, the unmanned side of the mark whose light reflectance is different by the optical rangefinders installed in front of and behind the vehicle. When the vehicle passes, it is recognized as a correction mark and the relative distance to the mark is measured.

The mark, which represents the reference position and direction, is appropriately installed in advance at a predetermined position on the traveling course, and the accurate position and direction of the vehicle are calculated from the relative distance to the mark obtained as described above. Correct the measurement data.

The automated guided vehicle 1 of this embodiment includes a travel control device 2, drive devices 3a and 3b, motors 4a and 4b, and pulse counters 5a and 5b for detecting the rotational speeds of drive wheels.
, Reduction gears 6a, 6b, drive wheels 7a, 7b, measurement wheels 8a, 8b arranged coaxially with the drive wheels, encoders 9a, 9b for detecting the number of rotations of the measurement wheels, and a measurement device 10.
, Caster wheels 11a, 11b, and optical rangefinder 12
a and 12b.

The drive wheels 7a and 7b are independently rotated at positions equidistant from the center of the vehicle body to the left and right by a motor fixed to the vehicle body through output shafts of reduction gears 6a and 6b attached to the motors. The unmanned vehicle can be driven in any direction on a plane by being driven and by the caster wheels 11a and 11b arranged in front of and behind the vehicle body.

The measuring device 10 includes the rotational speeds N _{L of} the left and right measuring wheels 8a and 8b detected by the encoders 9a and 9b,
From N _R , the actual vehicle speed V and the actual curvature R of the movement locus of the automated guided vehicle 1 are calculated by the following equation, and output to the travel control device 2. _{V = 0.5 Lp (N L +} N R) / Tc ... (12) R = ± 2 (N L -N R) / {TR (N L + N R)} ... (13) where, Tc: position measurement period Lp : Amount of movement per pulse of measuring wheel TR: tread length between measuring wheels _NL : number of pulses of left measuring wheel 8a during position measuring cycle Tc _NR : number of pulses of right measuring wheel 8b during position measuring cycle Tc

However, the upper side of the positive and negative signs shows the case where the automatic guided vehicle 1 turns clockwise, and the lower side shows the case where the automatic guided vehicle 1 turns counterclockwise. In the first embodiment, the measurement cycle is Tc = 50 msec. In addition, the measurement wheel is a driven wheel that is separate from the drive wheel that receives the load of the vehicle body,
Furthermore, it has a vertically movable structure with respect to the vehicle body so that it can accurately follow unevenness on the road surface.

As will be described later, the traveling control device 2 generates a traveling track of the unmanned guided vehicle 1 based on a preset target position and direction of the unmanned guided vehicle 1, and travels along the traveling track. The speed command is output to the left and right drive wheels 7a and 7b. Further, the traveling control device 2 controls a vehicle speed and curvature measurement period described later and a control period of the drive device by an internal clock. The drive devices 3a and 3b are driven by the motor 4 based on the speed command from the travel control device 2.
The rotation of a and 4b is controlled. As a result, the automated guided vehicle 1 travels on a predetermined track to reach the target position.

FIG. 3 shows a system block diagram of the traveling control device 2. The travel control device 2 uses the target trajectory generation device 20.
And a compensator 21 and a converter 24.

In the target trajectory generation device 20, the automatic guided vehicle 1 is used.
Using the current position and orientation of the above as the starting point and the preset target position and orientation of the automated guided vehicle 1 as the end point, an interpolation function for smoothly and continuously connecting the two points is calculated. For example, as shown in FIG. 5, in the XY Cartesian coordinate system set on the traveling plane based on the target stop position and azimuth, the target trajectory is the current position (Xs, Ys) and the target stop position (0, 0). Interpolation function (x (t), y (t)) for time t that smoothly and continuously connects
Is calculated by the following formula. The position of the automatic guided vehicle is the position of the center of the vehicle body (the center in the tread direction between the left and right driving wheels in the first embodiment) in the XY orthogonal coordinate system, and the azimuth is the direction of the automatic guided vehicle in the XY orthogonal coordinate system. Is the direction of travel.

(Txs <t <Txe) x (t) = a ₀ + a ₁ t + a ₂ t ² + a ₃ t ³ + a ₄ t ⁴
+ A ₅ t ⁵ (t ≧ Txe) x (t) = 0 (Tys <t <Tye) y (t) = b ₀ + b ₁ t + b ₂ t ² + b ₃ t ³ + b ₄ t ⁴
+ B ₅ t ⁵ (t ≧ Tye) y (t) = 0

Since the automatic guided vehicle is traveling on a straight course at the starting point, the coefficients a _{0 to} a ₅ and b _{0 to} b ₅ are obtained from the following boundary conditions. x (t = 0) = Xs x '(t = 0) = Vxs y (t = 0) = Ys y' (t = 0) = Vys x '' (t = 0) = x '' (t = Txe ) = X '(t = Txe) = x (t = T
xe) = 0 y '' (t = 0) = y '' (t = Tye) = y '(t = Tye) = y (t = T
ye) = 0

Here, x ′ and y ′ are velocities in the X and Y directions, respectively, and x ″ and y ″ are accelerations in the X and Y directions, respectively. Further, (Xs, Ys) is the position at the start of positioning when stopped, and (Vxs, Vys) is the speed at the start of positioning when stopped. Further, the positioning start time at the time of stop is set to 0, and the positioning control end times in the X and Y directions are set to Txe and Tye, respectively.

Further, Txe and Tye are parameters for adjusting the shape of the orbit. That is, in order to orient the vehicle in the X-axis direction when the vehicle is stopped, Txe> Tye, and both max | x ″ (t) | and max | y ″ (t) | Set Txe and Tye in advance so as not to exceed the capacity.

Through the above procedure, the target trajectory is calculated as an interpolation function. The compensator described later inputs the target vehicle speed and the target curvature, so that an interpolation function other than the above,
For example, it is possible to generate a trajectory by a clothoid function, a spline function, or the like.

The target trajectory generating device 20 calculates the target curvature Rd of the automatic guided vehicle 1 from the curvature of the interpolation function thus obtained, and also calculates the desired curvature Rd of the automated guided vehicle 1 from the time differential term of the interpolation function. Control cycle of target vehicle speed Vd (50mse
The calculation is performed in synchronization with c), and each is output to the compensator 21 for each control cycle.

FIG. 4 shows a system block diagram of the compensator 21. The compensator 21 includes a target vehicle speed compensator 22 including a vehicle speed compensator 221, a vehicle speed response delay compensator 222, and an adder 223, and a target curvature including a curvature compensator 231, a curvature response delay compensator 232, and an adder 233. Compensation device 23.

The vehicle speed compensator 221 calculates the sum a.Vd + b.Vd.Rd of the product of the target vehicle speed Vd and the compensation coefficient a and the product of Vd and the target curvature Rd and the compensation coefficient b as the vehicle speed compensation output.

In the vehicle speed response delay compensator 222, the product of the difference between the target vehicle speed Vd and the target vehicle speed Vd ⁻ one time before (control cycle) and the compensation coefficient c, the target vehicle speed Vd and one time later (control cycle). The product of the difference between the target vehicle speed Vd ⁺ and the compensation coefficient d is added, and the vehicle speed response delay compensation output c · (Vd −Vd ⁻ ) + d
-Calculate as (Vd ⁺ -Vd).

The adder 223 adds the vehicle speed compensation output and the vehicle speed response delay compensation output, and calculates the compensated vehicle speed command value Vd 'by the following equation. Vd '= a · Vd + b · Vd · Rd + c · (Vd -Vd -) + d · (Vd + -Vd) ... (14)

The coefficient a is for compensating the magnitude of the variation of the left and right driving wheel diameters and the influence of the variation on the target vehicle speed, and the coefficient b is the difference between the variation of the left and right driving wheel diameters and the variation. The effect of the amount on the target vehicle angular velocity is compensated, and the coefficients c and d compensate for the response delay of the vehicle speed.

In the curvature compensator 231, the term obtained by adding the compensation coefficient f to the product of the target curvature Rd and the compensation coefficient e is divided by the term obtained by adding the compensation coefficient h to the product of Rd and the compensation coefficient g to obtain the curvature compensation output. It is calculated as (e.Rd + f) / (g.Rd + h).

In the curvature response delay compensator 232, the product of the difference between the target curvature Rd and the target curvature Rd ⁻ one time before (control cycle) and the compensation coefficient i, the target curvature Rd and one time after (control cycle). The product of the difference of the target curvature Rd ⁺ and the compensation coefficient j is added, and the curvature response delay compensation output i · (Rd −Rd ⁻ ) + j
-Calculate as (Rd ⁺ -Rd).

The adder 233 adds the curvature compensation output and the curvature response delay compensation output, and calculates the compensated curvature command value Rd 'by the following equation. Rd '= (e · Rd + f) / (g · Rd + h) + i · (Rd -Rd -) + j · (Rd + -Rd) ... (15)

It should be noted that the coefficients e and h are for compensating for the deviation of the variation of the tread length to the target curvature, and the coefficients f, g and h are for compensating the influence of the variation of the left and right driving wheel diameters on the target curvature. The coefficients i and j are for compensating the curvature response delay.

The converter 24 uses the vehicle speed command value Vd 'and the curvature command value Rd' compensated by the compensator 21 for the speed command values V _L , V _{L for} the left and right drive wheels 7a, 7b.
_R is calculated and output to the driving devices 3a and 3b. Since the automatic guided vehicle 1 is a two-wheel independent drive system, the command value of the vehicle speed and the curvature Vd 'and Rd' and the drive wheel tread length TR, by the following equation, the left and right driving wheel speed command value V _L, V _R Can be calculated. _{V R = -Vd '· (-1} ± 0.5 · TR · Rd') ... (16) V L = Vd '· (1 ± 0.5 · TR · Rd') ... (17) However, (16), (17 ), The upper side of the positive / negative sign indicates the case where the automated guided vehicle 1 turns clockwise, and the lower side is
The case where the automatic guided vehicle 1 turns counterclockwise is shown.

As described above, in order to realize the calculated speed command values V _R and V _L as the traveling speed of the automatic guided vehicle 1, the motors 4a and 4b are feedback-controlled by the drive devices 3a and 3b. As this feedback signal,
The drive devices 3 and 4 configured by servo circuits using pulse counters 5a and 5b coaxially connected to the motors 4a and 4b control the speeds of the motors 4a and 4b by the output pulses of the pulse counters 5a and 5b. Then, the left and right driving wheel speeds are made to match V _R and V _L. This motor 4a, 4
The output torque from b is transmitted to the drive wheels 7a and 7b via the speed reducers 6a and 6b and the output shafts thereof. As described above speed command value V _R, V _L is outputted every control cycle.

According to the first embodiment, it is possible to effectively compensate for the track error due to the vehicle body constant effective value of the automatic guided vehicle, the dead time of the unmanned vehicle system, and the dynamic fluctuation during the dynamics, which have been difficult in the past. This made it possible to reach the target point with high accuracy. Furthermore, as an additional effect,
It becomes possible to compensate for an error in assembling the drive system at the time of assembling the automatic guided vehicle and a trajectory error factor such as a change with time of the drive system, so that the initial characteristics of the drive performance of the automatic guided vehicle can be maintained. As described above, it is possible to accurately position the automatic guided vehicle at the time of stop.

Second Embodiment In the second embodiment, a target vehicle speed compensating device 22 and a target curvature compensating device 23 constituting a compensating device 21 are respectively hierarchical neural networks (abbreviated as NN hereinafter).
And the coefficients a to j in the equations (14) and (15) are
It is estimated by. The rest of the configuration is the same as that of the first embodiment, so the explanation is omitted.

NN-1 constituting the target vehicle speed compensator 22
Has a two-layer structure as shown in FIG. The input layer is the target vehicle speed Vd output from the target trajectory generation device 20, the product Vd · Rd of the target vehicle speed Vd and the target curvature Rd, and the target vehicle speeds Vd ⁻ , Vd ⁺ before and after one time (control cycle). And four input units 101 to 10 for respectively inputting
It is composed of 4. All four input units output a linear function value for each input value.

The output layer also includes four input units 10
The output unit 105 outputs the linear function value represented by the following equation, with the sum of the values output from 1 to 104 multiplied by the respective weighting factors W ^V _ij as the input value In. F (In) = C · In + β (18) where C: coefficient, β: bias

NN-2 constituting the target curvature compensator 23
Has a three-layer structure as shown in FIG. The input layer includes the target curvature Rd output from the target trajectory generation device 20 and the target curvatures Rd ⁻ and Rd ⁺ before and after one time (control cycle).
And three input units 111 to 1 for respectively inputting
It is composed of 13. All three input units output a linear function value for each input value.

The hidden layer is composed of a plurality of intermediate unit groups 114, and each intermediate unit takes as an input value a value obtained by multiplying the output value of the input unit 111 for inputting the target curvature Rd by each weighting coefficient. Output the sigmoid function value for the input value. The number of intermediate units is
Depending on the accuracy of the compensation value of the target curvature finally obtained, the higher the required accuracy is, the more the number of them is provided.

The sigmoid function G (x) is defined by the following equation for the input value In and the bias β. G (In) = 2 / {1 + exp (-In + β)}-1 (19)

Further, the output layer uses the sum of values obtained by multiplying the output values of the plurality of intermediate unit groups 114 and the three input units 111 to 113 by the respective weighting factors as the input value and the equation (18) The output unit 115 outputs the same linear function value.

Next, a method of learning these NNs will be described. The automated guided vehicle 1 is actually run in advance based on a target trajectory composed of a plurality of running patterns. And the position and direction on the target trajectory in each traveling pattern,
By using each deviation amount from the actual position and azimuth based on the vehicle speed and curvature obtained from the rotation speeds of the measurement wheels 8a and 8b as teacher data, a learning method (back propagation method) for energy minimization as described below, Learning is performed by correcting each weighting coefficient of NN.

That is, first, with respect to the vehicle speed and the curvature, the error functions ε _V and ε _R and the energy functions E _V and E _R are defined as follows. _{ε V = Vd -V ε R =} Rd -R E V = ε V 2/2 E R = ε R 2/2

Next, with respect to the weighting coefficient W ^V _ij between the input and output units of NN-1 and the weighting coefficient W ^R _ij between the input, intermediate and output units of NN-2, respectively, Find the partial differential value of the energy function as shown in. ∂E _V / ∂W ^V _ij ∂E _R / ∂W ^R _ij

Using the obtained partial differential values, the respective weighting coefficient update amounts ΔW ^V _ij and ΔW ^R _ij are obtained as shown below. ΔW ^V _ij = −η (∂E _V / ∂W ^V _ij ) ΔW ^R _ij = −η (∂E _R / ∂W ^R _ij )

The weight coefficient update amounts ΔW ^V _ij and ΔW ^R
_The weighting coefficient is updated by _ij according to the following equation. W ^V _ij = W ^V _ij + ΔW ^V _ij W ^R _ij = W ^R _ij + ΔW ^R _ij

Under the updated weighting factors, NN-1 and NN-2 calculate output values V and R, respectively, based on the new input values, and for these calculation results, their target vehicle speed and curvature are calculated. Each weighting coefficient is updated again using Vd and Rd as new teacher data.

The above procedure is repeated for each error function ε _V and ε _R.
When the magnitude of is smaller than a preset threshold value, learning is terminated.

As a result, the energy function is minimized, and the target vehicle speed and the target curvature substantially match the actual vehicle speed and the actual curvature, respectively, so that the deviation from the target trajectory can be made extremely small. It is possible to further reduce the deviation amount and increase the degree of coincidence between the actual value and the target value by always learning during traveling.

In the second embodiment, since the neural network performs the compensation calculation for the dynamic fluctuation of the vehicle body constant effective value of the automatic guided vehicle, that is, the fluctuation during traveling, the target is realized in real time with an extremely simple structure. It becomes possible to travel on the track accurately.

In the second embodiment, during learning traveling,
The NN weighting coefficient was corrected using the actual vehicle speed and the actual curvature obtained from the measuring device in real time. However, without using the measuring device, for example, the deviation of the position and direction from the target trajectory measured from the outside of the automated guided vehicle Amounts may be used.

Third Embodiment Although the automatic guided vehicles in the first and second embodiments are straight or turnable according to the rotational speeds of the two left and right drive wheels, the third embodiment is The present invention relates to an automatic guided vehicle that travels straight or turns according to the steering angle and the rotation speed of one steering drive wheel.

In the third embodiment, the command values of the steering angle and the driving speed of the steered driving wheels are calculated based on the target vehicle speed and the curvature compensation value calculated in the same manner as the first and second embodiments. Only the difference from the first and second embodiments.
Only the differences will be described below.

FIG. 8 shows the turning state of the automatic guided vehicle in the third embodiment. The automated guided vehicle 1 has one steering drive wheel 30 on the left and right centerlines of the vehicle body at the rear of the vehicle body.
Is provided, and two left and right driven wheels 31a, 3 are provided at the front of the vehicle body.
1b and measurement wheels 32a and 32b are provided. The center of the vehicle body of the automatic guided vehicle is located at the center of the left and right driven wheels and the measurement wheel, and the center of the vehicle body and the center of the steering drive wheel are separated by a distance L ₁ .

In the automatic guided vehicle 1, depending on the steering angle θ and the rotation speed ω of the drive steered wheels 30, the center of the vehicle body turns at a speed of Vd ′ at a distance of 1 / Rd ′ from the rotation center O on the traveling plane. I shall. In addition, Rd 'and Vd' are
These are the command values of curvature and vehicle speed output from the compensator 21, respectively. The distance between the center of the steering wheel and the rotation center O at this time is L ₂ .

From the geometrical relationship of each part in FIG. 8, the angle between the center of the vehicle body at the center of rotation O and the center of the steered drive wheel is equal to the steering angle θ of the steered drive wheel 30. Holds. θ = tan ^-1 (Rd '· L ₁ ) (20)

On the other hand, since the angular velocity at the vehicle body center with respect to the rotation center O and the angular velocity at the center of the steering drive wheels are equal, the following relationship holds. ω = Vd '(1 / Rd ') / L 2 = Vd 'cos (θ) ... (21)

Therefore, as shown in the equations (20) and (21), the drive command value and the drive speed command value of the steering drive wheels can be determined by the curvature and vehicle speed command values obtained from the compensator.

[Brief description of drawings]

FIG. 1 is a block diagram showing a basic configuration of a fourth invention.

FIG. 2 is a schematic configuration diagram of an automated guided vehicle according to an embodiment.

FIG. 3 is a system block diagram of a travel control device.

FIG. 4 is a system block diagram of compensating means.

FIG. 5 is a diagram showing a target trajectory represented by an interpolation function.

FIG. 6 is a configuration diagram showing a configuration of a neural network as a target vehicle speed compensation device.

FIG. 7 is a configuration diagram showing a configuration of a neural network as a target curvature compensation device.

FIG. 8 is a diagram showing a turning situation of an automatic guided vehicle in the third embodiment.

[Explanation of symbols]

1 ... Automated guided vehicle 2 ... Travel control device 3a, 3b ...
Drive device 4a, 4b ... Motor 5a, 5b ... Pulse counter 6a, 6b ... Reduction gear 7a, 7b ... Drive wheel 8a, 8
b ... Measuring wheel 9a, 9b ... Encoder 10 ... Measuring device 11a, 11b ... Caster wheel 12a, 12b ... Optical distance meter 20 ... Target trajectory generation device 21 ... Compensation device 22 ...
Target vehicle speed compensation device 23 ... Target curvature compensation device 24 ... Conversion device 30 ...
Steering drive wheels 31a, 31b ... Driven wheels 32a, 32b ... Measuring wheels

Claims (1)

Claim: What is claimed is: 1. A target track generation means for outputting a target track to be traveled by an automatic guided vehicle having two left and right driving wheels as a target vehicle speed and a curvature, and the target track generation means. Compensation means for compensating the dynamic fluctuation error of the unmanned guided vehicle generated during traveling based on the target vehicle speed and the target curvature and outputting the target vehicle speed compensation value and the target curvature compensation value, and the target vehicle speed compensation value and the target A conversion unit that calculates a drive speed command value for the left and right drive wheels based on the curvature compensation value; and a drive unit that controls the drive of the left and right drive wheels based on the calculated left and right drive speed command values. A control device for an unmanned guided vehicle.