JP2003198233A - Control method for array antenna - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To lean to improve performance by each repetition of search, without needing a long learning sequence signal, as compared with the conventional ones. <P>SOLUTION: A method includes steps where an adaptive control type controller 20 randomly perturbs a bias voltage vector V (n), using a bias voltage value V<SB>m</SB>as one component from a specified initial value by means of a random vector R (n) generated by a random number generator, an objective function value J (n) which is an cross correlation coefficient for the bias voltage vector V (n) before perturbation is compared with an objective function value J (n+1), being a cross-correlation coefficient for the bias voltage vector V (n+1) after perturbation; a bias voltage value V<SB>m</SB>corresponding to the increase of the cross- correction coefficient before and after perturbation is selected and set and then the vector V (n) is randomly perturbed and set from the selected bias voltage of each of variable capacitance diodes 12-1 to 12-6; and this setting is repeated. <P>COPYRIGHT: (C)2003,JPO

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an array antenna control method capable of changing the directional characteristics of an array antenna device including a plurality of antenna elements, and more particularly,
Electronically controlled waveguide array antenna device (Electronically S
teerable Passive Array Radiator (ESPAR) Antenna;
Hereinafter referred to as ESPAR antenna. ), A method of controlling an array antenna capable of adaptively changing the directional characteristic.

[0002]

2. Description of the Related Art A conventional ESPAR antenna is, for example,
Prior Art Document 1 "T. Ohira et al.," Electronically s
teerable passive array radiator antennas for low-c
ost analog adaptive beamforming, "2000 IEEE Intern
ational Conference on PhasedArray System & Technol
ogy pp. 101-104, Dana point, California, May 21-2
5, 2000 "and Japanese Patent Laid-Open No. 2001-24431. The ESPAR antenna is provided with an exciting element to which a radio signal is fed, and a distance from the exciting element by a predetermined distance, and at least 1 to which the radio signal is not fed.
A non-excitation element and an array antenna composed of a variable reactance element connected to the non-excitation element are provided, and the directional characteristic of the array antenna can be changed by changing the reactance value of the variable reactance element. it can.

The beamforming method based on spatial power combination such as the ESPAR antenna can achieve variable directivity and obtain a high gain with a simple hardware configuration and low power consumption, so that a practical terminal ( Especially, it can be expected as an adaptive antenna mounted on a mobile user terminal).

However, in the case of the ESPAR antenna, the signal on the passive element cannot be observed. Therefore, it is necessary to observe only the output of a single port and process it as feedback for adjusting the reactance value. In other words, most of the methods made for conventional adaptive arrays cannot be applied directly to ESPAR antennas.

To solve this problem, for example, in the patent application of Japanese Patent Application No. 2000-307548, the so-called "steepest gradient method" is used to sequentially change the reactance value of each variable reactance element by a predetermined shift amount. Perturb,
A tilt vector of a predetermined evaluation function value for each reactance value is calculated, and the main beam of the array antenna is directed in the direction of the desired wave so that the evaluation function value becomes maximum or minimum based on the calculated tilt vector. A control method (hereinafter referred to as a first conventional example) for calculating and setting the reactance value of each variable reactance element for directing the null in the direction of the interference wave has been proposed.

[0006]

However, this first prior art example requires successive perturbations to determine each component of the gradient vector, which results in (M + 1) times in each iteration of the perturbation. Requires calculation of the objective function value of
In the case of the ESPAR antenna, there is a problem that a learning sequence having a length at least (M + 1) times as long as that of the conventional adaptive array is required and the amount of calculation becomes large.

On the other hand, the prior art document 2 "Yukihiro Kamiya et al.,
"Basic study of ESPAR antenna-SIN based on adaptive control
Statistical evaluation of R characteristics- ", IEICE technical report,
A.P2000-175, SANE2000-156,
pp. 17-24, The Institute of Electronics, Information and Communication Engineers, 200
"Issued January, 1st", the following "random search method"
The procedure (hereinafter referred to as the second conventional example) is used. (1) Consider a column vector x having the reactance value of each variable reactance element as an element, and use this as a reactance matrix. Such a matrix is generated by uniform random numbers within a predetermined range to generate a reactance matrix population. (2) Each reactance matrix included in the generated population is loaded into the ESPAR antenna one by one, and a sample of the received signal is observed in each case, and a predetermined mutual phase between the received signal and the learning sequence signal is observed. Calculate the number of relationships. (3) Of the obtained plurality of cross-correlation coefficients, the reactance matrix that gives the maximum cross-correlation coefficient is adopted as the weighting coefficient.

This second conventional example requires only one cross-correlation coefficient calculation for each iteration. However, this has the disadvantage that the next trial is independent of the previous trial and nothing is learned when the trial is complete.

An object of the present invention is to solve the above problems, and in a method for controlling an ESPAR antenna, when a main beam is directed to a desired wave and a null is directed to an interference wave, it is compared with a conventional example. An object of the present invention is to provide a control method of an array antenna that does not require a long learning sequence signal and can perform learning so that the performance improves at each search iteration.

[0010]

An array antenna control method according to the present invention comprises an excitation element for receiving a radio signal, and a plurality of non-excitation elements provided at a predetermined distance from the excitation element. , A plurality of variable reactance elements respectively connected to the plurality of non-excitation elements, by changing each reactance value set in each of the variable reactance elements, each of the non-excitation elements are respectively a waveguide or a reflection. In the method of controlling the array antenna, which operates as a container and changes the directional characteristics of the array antenna,
When the reactance value of each variable reactance element is randomly perturbed from a predetermined initial value and set, when the learning sequence signal included in the radio signal transmitted from the transmitter of the other party is received by the array antenna, A predetermined cross-correlation coefficient before and after perturbation between the received signal and the learning sequence signal generated with the same signal pattern as the learning sequence signal is calculated, and the cross-correlation coefficient before and after the perturbation is increased. It is characterized by including a step of repeating the step of selecting and setting a reactance value corresponding to the occasion and then randomly perturbing and setting the reactance value of each of the selected variable reactance elements.

In the array antenna control method, the initial value preferably has a maximum cross-correlation coefficient among the reactance values of the variable reactance elements corresponding to a plurality of predetermined radiation patterns. It is characterized in that it is a reactance value of each variable reactance element corresponding to one radiation pattern.

Further, in the array antenna control method, the plurality of radiation patterns are preferably
(A) a plurality of sector beam patterns each having a maximum gain in the direction from the excitation element to each non-excitation element; and (b) in the direction from the excitation element to an intermediate position between adjacent non-excitation elements. At least one set of a plurality of sector beam patterns having a maximum gain is included.

[0013]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram showing the configuration of an array antenna control apparatus according to an embodiment of the present invention. As shown in FIG. 1, the array antenna control device of this embodiment includes one excitation element A0 and a variable capacitance diode 1.
An array antenna device 100, which is an ESPAR antenna, is provided with six parasitic elements A1 to A6 loaded with 2-1 to 12-6, respectively, and a ground conductor 11, an adaptive control type controller 20, and learning. A sequence signal generator 21 and a bias voltage table memory 22 connected to the adaptive controller 20 are provided.

Here, the adaptive control type controller 20 is
For example, the learning sequence signal included in the radio signal transmitted from the transmitter of the other party is constituted by a digital computer such as a computer, and before the start of the radio communication by the demodulator 4, the exciting element A0 of the array antenna apparatus 100 is supplied.
And the learning sequence signal d generated by the learning sequence signal generator 21 having the same signal pattern as the learning sequence signal and the received signal y (n).
Based on (n), each variable capacitance diode 12 for directing the main beam of the array antenna device 100 in the direction of the desired wave and the null in the direction of the interference wave by executing the adaptive control process described later. It is characterized in that the bias voltage value V _m (m = 1, 2, ..., 6) applied to −1 to 12-6 is searched and set. In particular,
The adaptive control type controller 20 includes a random number generator, and a random vector R generated by the random number generator.
(B) The bias voltage vector V (n) having the bias voltage value V _m as a component is randomly perturbed from a predetermined initial value by (n), and is a cross-correlation coefficient with respect to the bias voltage vector V (n) before perturbation. The function value J (n) is compared with the objective function value J (n + 1) which is a cross-correlation coefficient for the bias voltage vector V (n + 1) after perturbation,
After the bias voltage V _m corresponding to the increase in the cross-correlation coefficient before and after the perturbation is selected and set, the bias voltage of each of the variable capacitance diodes 12-1 to 12-6 is randomly perturbed. And repeat setting. Therefore, starting from the initial value of the bias voltage, a random vector R (n) is generated and perturbed, and the corresponding bias voltage V when the cross-correlation coefficient before and after the perturbation increases.
_{After m} is selected and set, a random vector R (n) is further generated from the selected bias voltage, perturbed, and the above-described processing is repeated to select the random vector R (n) while sequentially generating it. The bias voltage thus generated is updated, whereby the main beam of the array antenna device 100 is directed toward the desired wave and the null is directed toward the interference wave so that the objective function value J (n) is maximized. To search the bias voltage vector V (n) of each of the variable capacitance diodes 12-1 to 12-6, and to have each bias voltage value V _m (m = 1, 2, ..., 6) found as a result of the search. Bias voltage value signal is applied to each variable capacitance diode 12
Output to -1 to 12-6 and set.

In FIG. 1, an array antenna device 100 is shown.
Is composed of an excitation element A0 and non-excitation elements A1 to A6 provided on the ground conductor 11, and the excitation element A0 includes six non-excitation elements A1 to A provided on the circumference of a radius r.
It is arranged so as to be surrounded by 6. Preferably, the parasitic elements A1 to A6 are provided on the circumference of the radius r at equal intervals. Each of the excitation elements A0 and the non-excitation elements A1 to A6 is configured to be, for example, a monopole element having a length of about 1/4 with respect to the wavelength λ of the desired wave, and the radius r is λ / 4. Is configured to be. The feeding point of the excitation element A0 is connected to the low noise amplifier (LNA) 1 via the coaxial cable 5, and the non-excitation elements A1 to A6 are the variable capacitance diodes 12-1.
To 12-6, and these variable capacitance diodes 1
2-1 to 12-6 change the reactance value by setting the bias voltage value signal from the adaptive control type controller 20.

FIG. 2 is a vertical sectional view of the array antenna device 100. The excitation element A0 is electrically insulated from the ground conductor 11, and the non-excitation elements A1 to A6 are grounded to the ground conductor 11 via the variable capacitance diodes 12-1 to 12-6 in a high frequency manner. Variable capacitance diode 12-
Explaining the operations of 1 to 12-6, for example, the excitation element A
0 and the parasitic elements A1 to A6 have substantially the same length in the longitudinal direction, for example, the variable capacitance diode 12-
When 1 has an inductance property (L property), the variable capacitance diode 12-1 serves as an extension coil, the electric lengths of the non-exciting elements A1 to A6 become longer than that of the exciting element A0, and they function as a reflector. On the other hand, for example, when the variable capacitance diode 12-1 has a capacitance property (C property), the variable capacitance diode 12-1 becomes a shortening capacitor, and the electrical length of the non-excitation element A1 becomes shorter than that of the excitation element A0. , Works as a director. In addition, the parasitic element A connected to the other variable capacitance diodes 12-2 to 12-6.
The same applies to 2 to A6.

Therefore, the array antenna device 100 of FIG.
In the array antenna, the bias voltage values applied to the variable capacitance diodes 12-1 to 12-6 connected to the parasitic elements A1 to A6 are changed to change the reactance value which is the junction capacitance value thereof. Device 1
00 plane directional characteristics can be changed.

In the array antenna controller of FIG.
The array antenna device 100 receives the radio signal and
The received signal is low noise amplified through the coaxial cable 5.
(LNA) 1 to be amplified and then down
The converter (D / C) 2 converts the amplified signal to a predetermined intermediate value.
Low-frequency conversion is performed into a frequency signal (IF signal). Furthermore, A
The / D converter 3 digitizes the analog signal converted into the low frequency band.
A / D conversion to digital signal and adaptive control of the digital signal
It is output to the type controller 20 and the demodulator 4. Then
The adaptive control type controller 20, as will be described later in detail,
Random vector R generated by random number generator
Bias voltage value V depending on (n)_mBahia as an ingredient
The random voltage vector V (n) from a predetermined initial value
Perturb the bias voltage vector V (n) before perturbation.
The objective function value J (n), which is the cross-correlation coefficient, and the perturbation
For the latter bias voltage vector V (n + 1),
By comparing with the objective function value J (n + 1) which is a correlation coefficient,
When the cross-correlation coefficient before and after perturbation increases, the corresponding
As voltage V_mAfter selecting and setting, select each of the above
Bias voltage of the variable capacitance diodes 12-1 to 12-6
Repeatedly setting by perturbing the above randomly from pressure
You Therefore, starting from the initial value of the bias voltage,
Generate and perturb the dam vector R (n) before and after the perturbation.
Bias voltage V corresponding when the cross-correlation coefficient increases
_mAfter selecting and setting the selected bias voltage
Is further perturbed by generating a random vector R (n)
By repeating the above process, the random vector
The selected bias voltage while sequentially generating R (n).
The objective function value J (n) is updated to the maximum value.
As described above, the main beam of the array antenna device 100 is
The null toward the desired wave and toward the interference wave
Of the variable capacitance diodes 12-1 to 12-6 for
Searching for the Ias voltage vector V (n) and finding the result of the search
Each bias voltage value V _m(M = 1, 2, ..., 6)
Each of the variable capacitance diodes 12 has a bias voltage value signal
Output to -1 to 12-6 and set. On the other hand, demodulator 4
Performs demodulation processing on the input received signal y (n).
Thus, the demodulated signal which is a data signal is output.

The transmitting station for transmitting the radio signal received by the array antenna 100 is a learning sequence signal generator 21.
In accordance with a digital data signal of a predetermined symbol rate including the same learning sequence signal as that generated in step 1, a carrier signal of radio frequency is modulated using a digital modulation method such as BPSK or QPSK, The modulated signal is power-amplified and transmitted to the array antenna apparatus 100 of the receiving station. In the embodiment according to the present invention, the wireless signal including the learning sequence signal is transmitted from the transmitting station to the receiving station before the data communication, and the adaptive control process by the adaptive controller 20 is executed in the receiving station. To be done.

The prior art phased array antenna directly controls the weight vector (amplitude and phase) of each element. On the other hand, in the array antenna device 100 which is the ESPAR antenna, the weight circuit does not exist, and instead, the variable reactance elements (in the present embodiment, the variable capacitance diode 12-1) loaded on the non-excitation elements A1 to A6 are used.
12-6) to control the reactance value. Therefore,
The concept of an “equivalent weight vector” corresponding to the weight vector of the prior art is introduced, and this is related to the reactance value. The array antenna device 100 is essentially different from the phased array antenna of the related art in that (1) the output of the received signal is one system, (2)
Positive use of inter-element coupling, and (3) integration of non-excitation element and variable reactance element
It is a point. These are the essence of the operation for the array antenna device 100, and must be considered in the antenna design stage and the control theory construction stage.

Here, the received signal y (t) output from the array antenna device 100 composed of the ESPAR antenna
Is formulated as a function of each reactance value (x ₁ , ..., X ₆ ) of the parasitic elements A1 to A6. In the description of the formulation, a time variable t is used. In the control process, the iterative function parameter n corresponding to the time is used to execute the digital process using the recurrence formula. The variable beam forming in the array antenna apparatus 100 is performed by the bias voltage value V on each variable capacitance diode 12-1 to 12-6.
_{This is performed by controlling m} (m = 1, 2, ..., 6), and as a result, controlling these reactance values.

The parasitic element A1 is positioned relative to the exciting element A0.
Angle as the reference axis _q(Q = 0, 1,
, U) signal u with angle of arrival (DOA)_q(T)
Suppose there are a total of Q + 1 sources
It s_m(T) (m = 0, 1, ..., 6) is m of the antenna
In the second element Am (ie, excitation element or non-excitation element)
When the incident RF signal is represented, the signal vector s (t) is
RF signal s in the m-th component_mA column vector with (t)
Suppose. At this time, the signal vector s (t) is
Can be expressed as

[0024]

[Equation 1]

Where a (θ _q ) is

[Equation 2] Is the steering vector defined by. Where r
Is the radius of the array antenna device 100, and φ _m
Represents the angle at which each parasitic element Am is positioned with respect to the excitation element A0, and φ _m = 2π (m−1) / 6 (m = 1, ...,
6). A received signal y (t) (a high-frequency signal (RF signal) in the preceding stage of the LNA 1 is referred to in the following explanation of the principle for convenience of explanation.) Is an RF output signal of a single port of the antenna, which is given by the following equation. .

[0026]

## EQU00003 ## y (t) = i ^T s (t) + n (t)

Here, i = [i ₀ , i ₁ , i ₂ , ..., i
_M] ^T is, RF current _{i m (m = 0, ...} , 6) appearing respectively on the excitation element A0 and the parasitic elements A1 through A6 is an RF current vector having a component, n (t) is an array antenna device Represents additional white Gaussian noise at 100.

According to the analysis of the ESPAR antenna disclosed in the first conventional example, the RF current vector i is formulated as follows.

[0029]

I = v _s (I + YX) ⁻¹ y ₀

Where I is the 6 + 1 order identity matrix,
v _s is a constant gain factor. Further, in the equation 4, the reactance value x on the variable capacitance diodes 12-1 to 12-6 is x
Diagonal matrix X = diag containing _m (m = 1, ..., 6) in the component
[50, jx ₁ , jx ₂ , ..., jx ₆ ] is called a reactance matrix. The reactance value x _m is a function of the bias voltage V _m on the varactor diodes 12-1 to 12-6. Furthermore, in Expression 4, Y = [y _kl ]
_{(6 + 1) × (6 + 1)} is called the admittance matrix,
y _kl is the antenna element Ak and the antenna element Al (0 ≦
represents the mutual admittance between k and l ≦ 6). The vector y ₀ is the first column of the admittance matrix Y. For the value of mutual admittance y _kl , according to the known reciprocity theorem,
Similar to the normal array antenna device, y _kl = y _lk holds. The value of the mutual admittance y _kl is also constant depending on the physical structure of the antenna, such as the radius, the space interval, and the element length, and further, from the rotational symmetry of the array antenna apparatus 100, Fulfill.

[0031]

Y ₁₁ = y ₂₂ = y ₃₃ = y ₄₄ = y ₅₅ = y ₆₆

Y ₀₁ = y ₀₂ = y ₀₃ = y ₀₄ = y ₀₅ = y ₀₆

Y ₁₂ = y ₂₃ = y ₃₄ = y ₄₅ = y ₅₆ = y ₆₁

Y ₁₃ = y ₂₄ = y ₃₅ = y ₄₆ = y ₅₁ = y ₆₂

Y ₁₄ = y ₂₅ = y ₃₆

Therefore, the admittance matrix Y has six components of mutual admittance, y ₀₀ , y ₁₀ , y ₁₁ , y ₂₁ and y.
Determined by only ₃₁ and y ₄₁ , vector y
It can be seen that ₀ is determined only by the two components of mutual admittance, y ₀₀ , y ₁₀ .

From the equations 3 and 4, the current vector i and the received signal y (t) are the reactance values (x ₁ , ..., X ₆ ).
Therefore, it can be seen that the received signal y (t) is a function of the bias voltage value applied to each of the variable capacitance diodes 12-1 to 12-6. Therefore, in the array antenna control method according to the embodiment of the present invention, the directivity pattern of the array antenna device 100 is formed by changing each of the bias voltage values.

In the experiments conducted by the present inventors, the bias voltage of the variable capacitance diodes 12-1 to 12-6 applied is set in the range of -0.5V to 20V. In practice, digital to analog (D / A) converters are used to set the bias voltage. This D / A converter is encoded with 12 bits and is -2.
Digital values from 048 to 2047 can be accepted. For ease of notation, we refer to the encoded value of the bias voltage as the digital voltage.
As described above, the reactance value x _m is the bias voltage V
Note that it is a function of _m . In our experiments,

It is assumed that x _m = −0.0217V _m −49.21. Here, V _m is the value of the digital voltage.

Next, a control method for controlling the array antenna device formulated as described above will be considered. From the above discussion, it can be seen that it is difficult to apply a conventional control method such as the LMS algorithm to the ESPAR antenna. The main reason for this is that the structure of the simple ESPAR, ie the antenna has only a single output signal y (t). The signal y (t) received at the single port is observed, but the surrounding parasitic element A1
It is not possible to observe the signal at A6. Therefore, there is a need to develop a special adaptive control method for ESPAR antennas.

In the second conventional example, a random search method for the directivity pattern of the ESPAR antenna is studied.
Let V = [V ₁ , V ₂ , ..., V _M ] be an M-dimensional bias voltage vector whose components are bias voltages on the reactance value x _m (m = 1, 2, ..., M), respectively. To do. Note that the reactance value x _m in Equation 4 is a function of the bias voltage value V _m . This function depends on the implementation circuit for the reactance value. A series of bias voltage vectors V (n) =, where n is the number of search iterations.
[V ₁ (n), ..., V _M (n)] is generated according to the following equation.

[0037]

(11) V (n) = R (n) (N = 1, 2, ..., N)

Here, R (n) = [R ₁ (n), ..., R
_M (n)] is a random vector of voltage values selected by the random number generator to have a uniform distribution over the range of bias voltages on each varactor diode. The index n indicates the number of search iterations. The value of the bias voltage vector V (n) is fed to the loaded terminal and the output of the receiver, the received signal y (n) (sample for the nth iteration on the received signal y (t)) is measured, The objective function value J (n) = J (V (n)) was then calculated. At the end of the random search phase, we find that the bias voltage vector V with the largest objective function value J (n).
The value of (n) was discovered.

This method, called the "(pure) random search method", has the disadvantage that nothing is learned when the trial is completed in step n. The next trial in step n + 1 is independent of the previous trial. this is,
For example, like the "steepest gradient method" according to the first conventional example,
No consideration is given to the local continuity property of the curved surface of the objective function. To this end, the embodiments of the present invention employ a more efficient "sequential" random search method.

Also in the sequential random search method proposed in the embodiment according to the present invention, the bias voltage vector V (n) is randomly changed. Objective function value J before and after change
(N) (for example, the cross-correlation coefficient between the received signal y (n) and the learning sequence signal d (n)) is calculated, and the two calculated values are compared. If the change increases the objective function value J (n), the change is accepted. If not augmented, the change is rejected and a new random change is attempted. This procedure can be described algebraically as follows.

[0041]

## EQU12 ## V (n + 1) = V (n) + (1/2) × {1
+ Sgn [J (V (n) + R (n))-J (V
(N))]} R (n); n = 1, 2, ..., N-1

Here, R (n) is a random M-dimensional vector (M = 6 in this embodiment), and J (V
(N) is an objective function value based on P samples of y (t) in Equation 3 when the bias voltage vector is set to V (n) (that is, sample y of the received signal y (t)).
(N) is the evaluation value of the cross-correlation coefficient between the learning sequence signal d (n) and J (V (n) + R (n)) is the bias voltage vector set to V (n) + R (n). It is an evaluation value of the objective function value based on P samples of y (t) when set. Further, the sign operator sgn [z] is z ≧ 0.
Is +1 when, and -1 when z <0.

Each component in the random vector R (n) of equation (12) is (i) -a random variable uniformly distributed over the range from b to (b), and (ii) zero mean and variance σ.
, And a Gaussian sequence with Here, b and σ are positive. The values of b and σ may be constant. However, it appears more reasonable that the range of the uniform distribution and the variance of the Gaussian distribution are reduced during the iterative procedure of Eq. Therefore, as an alternative, the following equation is used as the range parameter b (n) and the variance σ (n), which change according to the iteration number parameter n.

[0044]

[Equation 13]

[Equation 14]

Here, the coefficient b ₀ of the range parameter, the coefficient σ _{0 of} the variance, the step parameter τ, and the iteration number parameter n are positive constants. Equation 13 and Equation 1
4, the range parameter b (n) and the variance σ are used.
The value of (n) decreases as the number of iterations increases, as illustrated in FIG. Here, the value set as the coefficient b ₀ of the range parameter and the coefficient σ ₀ of the variance 15
00 is represented by a digital voltage.

Referring to FIG. 4, a graph illustrating the perturbation of the bias voltage vector V (n) by the random vector R (n) generated by the adaptive controller 20 is illustrated. In FIG. 4 and its description, the bias voltage on the horizontal axis in FIG. 4 is represented by a first-order component element, not a vector. 4A is a graph showing the probability density of the random vector R (n) that perturbs the bias voltage vector V (n), and FIG. 4B is a graph showing the change of the objective function value J due to the perturbation. It is a graph which shows. Bias voltage value V _m which is a component of the bias voltage vector V (n)
When (n) is applied to the variable-capacitance diode 12-m, the adaptive controller 20 calculates the average V _m (n).
And the bias voltage value V _m (n + 1) is randomly selected from the bias voltage values (FIG. 4A) that are Gaussian distributed with the variance σ (n). In other words, mean 0 and variance σ
The bias voltage value V _m (n + 1) is obtained by perturbing the bias voltage value V _m (n) by the component R _m (n) of the random vector that is randomly selected from the Gaussian-distributed bias voltage in (n). . Perturbed bias voltage value V
_The candidate bias voltage values selected as _m (n + 1) are concentrated with a variance σ (n) around the perturbed bias voltage value V _m (n).

As shown in FIG. 4B, the objective function value J (n) = J (V based on the bias voltage value V _m (n).
_m (n)) (that is, the bias voltage vector V (n) including the bias voltage value V _m (n) is output to the variable capacitance diodes 12-1 to 12-6 and set, and the objective function value J (n) is set. )) Rather than the bias voltage value R _m (n) + V
_m objective function value _J based on the (n) (R m (n ) + V
_{When m} (n) is larger, V _m (n + 1) = R
_m (n) + _Vm (n) is accepted as the new bias voltage value. Unlike the case of FIG. 4B, the objective function value J
When (R _m (n) + V _m (n)) is less than or equal to the objective function value J (n), the perturbation due to the component R _m (n) of the random vector is rejected, and the average V _m (n) and the variance σ. At (n), the next bias voltage value is again randomly selected from the Gaussian-distributed bias voltage values.

In the iteration of equation 12, in this embodiment,
The cross-correlation coefficient between the received signal y (n) and the learning sequence signal d (n) is adopted as the objective function J (n). Hereinafter, d (n) represents the P-dimensional column vector of the learning sequence signal, and y (n) is the received signal y (t) in Equation 3.
Let us denote the P-dimensional column vector that is the discrete-time sample of The cross-correlation coefficient J (n) = ρ (n) between the received signal y (n) and the learning sequence signal d (n) at the time (that is, the number of repetitions) n is defined by the following equation.

[0049]

[Equation 15]

Here, the superscript ^H indicates a complex conjugate transpose. It should be noted that the received signal y (n) output from the excitation element A0, which is a single port of the array antenna apparatus 100, is a high-order non-linear function of the adjustable reactance value x _m .

Next, the application control processing of the ESPAR antenna by the above-described sequential random search method executed by the adaptive controller 20 will be described with reference to FIGS. 5 and 6.

In step S1 of FIG. 5, the search iteration parameter n is initialized to zero. Then in step S2,
The initial value of the bias voltage vector to be applied to the variable capacitance diodes 12-1 to 12-6 is selected. In step S3, the selected bias voltage vector V (n)
The initial value V (0) of is output to the variable capacitance diodes 12-1 to 12-6 and set. With the bias voltage vector V (0) set, in step S4, the received signal y (n) output from the array antenna apparatus 100 is measured and generated by the learning sequence signal generator 21. Based on the learning sequence signal d (n), the objective function value J (n), which is a cross-correlation coefficient, is calculated using Equation 15.

In step S5, the iteration number parameter n is incremented by 1, and the bias voltage vector V (n) is updated with the value of V (n-1). In step S6, the random vector R (n) is generated using the random number generator provided in the adaptive controller 20.
Here, as described above, the random vector R (n)
The occurrence of may be limited to a uniform or Gaussian distributed range using Eq. Next, the bias voltage vector V (n) + R (n) is set to the variable capacitance diode 1
Output to 2-1 to 12-6 and set. With this bias voltage vector V (n) + R (n) set,
In step S8, the received signal y (n) is measured, and based on the received signal y (n) and the learning sequence signal d (n),
Is used to calculate the objective function value J (n), which is a cross-correlation coefficient.

Next, in step S9, when the objective function value J (n) calculated in step S8 is larger than the previously calculated objective function value J (n-1), in step S10, the bias voltage is calculated. The vector V (n) is updated with the value of the bias voltage vector V (n) + R (n) perturbed by the random vector R (n), and step S12
Proceed to. Step S9 is NO (that is, J (n) ≦ J
(N-1)), the objective function value J (n) is updated in step S11 without updating the bias voltage vector V (n).
Is updated with the value of the objective function value J (n-1), and step S
Proceed to 12. Therefore, when the bias voltage vector V (n) is not updated in the nth search, in the n + 1th search, based on the bias voltage vector V (n-1) and its objective function value J (n-1). , N + 1-th search result (that is, the bias voltage vector V (n + 1)
And its objective function value J (n + 1)) can be evaluated.

In step S12, the iteration number parameter n
Is less than a predetermined threshold value (upper limit value of the number of iterations) N, the process returns to step S5.
Then, the bias voltage vector V (n) is output and set to the variable capacitance diodes 12-1 to 12-6, and the adaptive control process is ended.

As described above, according to the array antenna control method by the sequential random search method of the present invention, the property of local continuity of the curved surface of the objective function J (n) is used, and each iteration step is repeated. In addition, the objective function value J (n) can be controlled so that the objective function value J (n) increases by referring to (learning) the previous result. At least, unlike the “pure” random search method, the objective function value J (n) Can be controlled so as not to decrease.

FIG. 6 shows a subroutine of the bias voltage initial value selection processing in step S2 of FIG. In FIG. 6, first, in step S21, the number I of candidate bias voltage vectors to be selected is set to 12, the initial value J ₀ (0) of the objective function value is set to −1, and the iteration number parameter i ( 1 ≦ i ≦ I) is initialized to 1. In this embodiment,
As the initial value candidates to be selected, 12 bias voltage vectors S (i) (i = 1, 2, ..., 12) shown in the following table and stored in advance in the bias voltage table memory 22 are used. Here, these bias voltage values are represented by the aforementioned digital voltage.

[0058]

[Table 1] Initial value voltage vector S (i) ―――――――――――――――――――――――――――――――――――― S (1) = {1800, -1800, -1800, -1800, -1800, -1800} S (2) = {1800, 1800, -1800, -1800, -1800, -1800} S (3) = {-1800, 1800, -1800, -1800, -1800, -1800} S (4) = {-1800, 1800, 1800, -1800, -1800, -1800} S (5) = {-1800, -1800, 1800, -1800, -1800, -1800} S (6) = {-1800, -1800, 1800, 1800, -1800, -1800} S (7) = {-1800, -1800, -1800, 1800, -1800, -1800} S (8) = {-1800, -1800, -1800, 1800, 1800, -1800} S (9) = {-1800, -1800, -1800, -1800, 1800, -1800} S (10) = {-1800, -1800, -1800, -1800, 1800, 1800} S (11) = {-1800, -1800, -1800, -1800, -1800, 1800} S (12) = {1800, -1800, -1800, -1800, -1800, 1800} ――――――――――――――――――――――――――――――――――――

Here, when the bias voltage vector S (1) is set to the variable capacitance diodes 12-1 to 12-6 corresponding to the parasitic elements A1 to A6, the main beam of the array antenna apparatus 100 has an azimuth angle of 0. Orientation (Excitation element A
The direction is set from 0 to the non-excitation element A1). Similarly, when the bias voltage vectors S (2) to S (12) are set in the variable capacitance diodes 12-1 to 12-6 corresponding to the parasitic elements A1 to A6, respectively, the main elements of the array antenna device 100 are Beams have azimuth angles of 30 °, 60 °, 90 °, 120 °, 150 °, 18
0 °, 210 °, 240 °, 270 °, 300 ° and 3
It is set to face the direction of 30 °. That is, the bias voltage vectors S (1), S (3), S (5), S
The application of (7), S (9), and S (11) forms six sector beam patterns each having a maximum gain in the direction from the excitation element A0 to the non-excitation elements A1 to A6. On the other hand, the bias voltage vectors S (2), S
(4), S (6), S (8), S (10), S (12)
Is applied from the exciting element A0 to the non-exciting elements (A1 and A2, A2 and A3, A3 and A4, A4) which are adjacent to each other.
4 and A5, A5 and A6, A6 and A1) are formed to form six sector beam patterns each having a maximum gain in a direction toward an intermediate position.

In step S22, the bias voltage vector S (i) is set to the variable capacitance diodes 12-1 to 12-6.
Output to and set. This bias voltage vector S
With (i) set, in step S23,
The received signal y (n) is measured, and based on this and the learning sequence signal d (n), the objective function value J _i (0) that is a cross-correlation coefficient is calculated using Equation 15. Step S24
, The objective function value J _i calculated in step S23
(0) is the previously calculated objective function value J _i−1 (0)
If it is larger than the above, in step S25, the bias voltage vector V (0) is updated with the value of the bias voltage vector S (i), and the process proceeds to step S27. Step S24
Is NO (that is, J _i (0) ≦ J _i−1 (0)), in step S26, the bias voltage vector V
The objective function value J _i (0) is updated with the value of the objective function value J _i−1 (0) without updating (0), and the process proceeds to step S27. Therefore, in the i-th selection, the bias voltage vector V
When (0) is not updated, in the next i + 1-th selection, the i + 1-th time is selected based on the bias voltage vector V (0) at the i-1-th time point and its objective function value J _i (i-1). The selection result of can be evaluated.
In step S27, when the selection of the initial value is executed for all the candidates S (i) of the bias voltage vector (that is, when the iteration number i reaches 12), the final bias voltage vector V ( 0) is selected as an initial value and the process returns to step S3 of FIG. 5, otherwise,
Increment the number of iterations i by 1 and step S22
Return to.

The initial value selection process of the bias voltage vector includes, in addition to selecting from the plurality of bias voltage vectors stored in advance as described above, an omnidirectional vector (for example, V (0) = {0 , 0,0,0,0,0})
Is used, or a random vector is used. In the experiments conducted by the present inventors, a random vector was used as an initial value. However, when the bias voltage vector initial value selection processing described with reference to FIG. 6 is used, the beam directivity can be set in accordance with the approximate arrival direction of the desired wave. It is expected that performing a sequential random search will give better results than using an omnidirectional vector or a random vector as the initial value.

[0062]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The present inventors have performed a simulation of the control device for the array antenna of FIG. 1, and the results will be described below. The SINR gain (dBo) of the ESPAR antenna controlled based on the above-described sequential random search method is evaluated with reference to the experimental results. In this case SI
NR (Signal to Interference to Noise Power Ratio) gain is expressed as the difference between the output SINR and the input SIR (Signal to Interference Power Ratio), d
Bo means the SINR gain obtained by the adaptive antenna compared to the omnidirectional antenna. Our experiments were performed on the signal model of Eq.
Here, the gain coefficient v _s of the equation 4 is selected as v _s = 100. The (Q + 1) source signals are generated in BPSK (binary phase shift keying) mode. Number 1
The size of the data block for calculating each of the objective functions J that are the cross-correlation coefficients defined in
= 100 is adopted. The number of block iterations is N
= 100.

In most of the experiments conducted by the present inventors, the applied variable capacitance diodes 12-1 to 12- are applied.
The bias voltage of 6 is set over the range of −0.5V to 20V, and the bias voltage of 12 is set as described above.
D / A converted by bits to obtain a digital value (digital voltage) V _m (m = 1, ..., From −2048 to 2047)
6). In the experiments conducted by the present inventors, the reactance value x _m ranges from −93.6Ω to −4.8Ω corresponding to the equation ₁₀ .

The statistical analysis performed by the present inventors is SIN.
The complement of the cumulative distribution function (CDF) value of the R gain is adopted. The complement of the CDF value indicates the probability that the SINR gain Z exceeds the given real number z.

[0065]

## EQU16 ## Pr (Z ≧ z) = 1−Pr (Z <z)

In the calculation of the complement of these CDF values, the desired wave signal is fixed so as to arrive at an angle of 15 °, and the DOA (arrival azimuth angle) of Q = 3 interference signals is from 0 ° to 359 °. It is set to be uniformly random in the range of. The input SIR is assumed to be -4.77 dB (ie, the power of each signal is 1). A total of 1000 sets of DOAs are used for these statistics. A random vector was used as the initial value of the bias voltage vector applied to the variable reactance elements 12-1 to 12-6.

As described above, the random vector R (n) in the equation 12 belongs to the “uniform” random distribution or the “Gaussian” distribution. FIG. 7 shows a random vector R
Distribution range [-b based on coefficient b ₀ with different (n)
7 is a graph showing the statistical performance of SINR gain of the ESPAR antenna in the case of a uniform random vector having (n), b (n)]. Hereinafter, except for the experiment of FIG. 11, the value of the step parameter τ of the equation 12 is always set to 5. The setting value of the coefficient b ₀ of the range parameter is 10
When increased from 0 to 1500, the statistical performance improves and the trend for improvement tends towards saturation. Referring to the graph of FIG. 8, the statistical performance of the SINR gain of the ESPAR antenna is shown in FIG. 7 even with a “Gaussian” random vector R (n) having a variance σ (n) based on different coefficients σ ₀ . We can observe the same phenomenon as in. The curve when the dispersion coefficient σ ₀ = 1500 shows that the ESPAR antenna has a probability of at least 5 dBo SIN with a probability of 70%.
It means that R gain can be provided.

In FIG. 9, we have shown that the random vector
Coefficient b of range parameter of R (n) ₀Or coefficient of variance σ
₀Two curves for the average value of SINR gain for
Have lots. Average based on 1000 pairs of DOAs
Is being done. σ₀= 1500 and b₀= 1500
Then, using a Gaussian-distributed random vector R (n)
The average value of SINR gain when
0.8 dB larger than when using the vector R (n)
Yes. In FIG. 10, the present inventors have found that a uniformly distributed random vector.
Cuttle R (n) and Gaussian random vector R
Two curves relating to the complement of the CDF value in the case of (n)
I'm comparing. Here, b in FIG.₀When = 1500
Graph and σ in FIG.₀With the graph when = 1500
Compared. Clearly the Gaussian distribution is more uniform
It's working better than when.

Next, the present inventors consider the effect of the step parameter τ in Equation 14 when using a Gaussian-distributed random vector R (n). Figure 11
In addition, the present inventors show a curve of the average value of the SINR gain with respect to the step parameter τ when σ ₀ = 1500. When the step parameter τ is larger than 5, it seems that the step parameter τ has almost no effect on the SINR gain.

In FIG. 12, we compare the complements of the CDF curves for different input SNRs when using a Gaussian distributed random vector R (n). We have observed that the curve shifts slightly to the left when the SNR is changed from 30 dB to 20 dB. However, the input SNR is 10d
When reduced to B and 0 dB, performance drops significantly.

Finally, we have an input SNR of 30.
In the case of dB, the sequential random search method (when the Gaussian-distributed random vector R (n) is used) proposed in the present invention is compared with the pure random search method according to the second conventional example. FIG. 13 shows C according to two different search methods.
The complement of the DF curve is shown. The proposed sequential random search method works obviously better than the pure random search method. SINR based on 1000 DOAs
By averaging, we find that the average SINR gain of the sequential random search method is 1.7 dB higher than that of the pure random search method. This is mainly because the present inventors consider the local continuity property of the curved surface of the objective function in the sequential random search method.

<Modification> In the above embodiment, 6
Although the non-exciting elements A1 to A6 of the book are used, if the number of the non-exciting elements A1 to A6 is at least plural, the directional characteristics of the array antenna device can be electronically controlled. Alternatively, more than six parasitic elements may be provided. Further, the arrangement shape of the non-excitation elements A1 to A6 is not limited to the above-described embodiment, and may be a predetermined distance from the excitation element A0. That is, the intervals between the parasitic elements A1 to A6 may not be constant.

In the above embodiment, the non-excitation element A is used.
1 to A6 are variable capacitance diodes 12-1 to 12-6
However, the present invention is not limited to this, and any variable reactance element capable of controlling the reactance value may be used.
Since variable capacitance diodes are generally capacitive circuit elements,
The reactance value is always negative. The reactance value of the variable reactance element ranges from a positive value to a negative value, and for this purpose, for example, the variable capacitance diode 12 is used.
The reactance value can be changed from a positive value to a negative value by inserting a fixed inductor in series with -1 to 12-6 or by making the lengths of the parasitic elements A1 to A6 longer. .

In this embodiment, the cross-correlation coefficient between the received signal y (n) and the learning sequence signal d (n) is used as the objective function J (n), but any other objective function may be used. . For example, using the square of the cross-correlation coefficient J (n),
Since it is not a function including the square root like Equation 15,
The calculation can be simplified.

Further, as the distribution of the range for perturbing the bias voltage vector V (n), not only the uniform distribution and the Gaussian distribution but also other distributions (eg, gamma distribution)
May be used.

In the experiments described in this specification, a random vector was used as the initial value of the bias voltage vector, but as described with reference to FIG. A preferred initial value may be selected. The present inventors compared the case where the random vector or the omnidirectional vector is selected as the initial value of the bias voltage vector with the initial value selection processing as shown in FIG. It is confirmed that the directivity pattern of the array antenna converges fastest when the length learning sequence signal is used. It is not intended to limit the predetermined set of bias voltage vectors to those illustrated in Table 1.

In the above embodiment, the adaptive control processing using the learning sequence signal is executed before the actual communication is started. However, the present invention is not limited to this, and may be executed at the beginning of the communication. It may be performed every time period.

As described above, according to the array antenna control method of the embodiment of the present invention, it is possible to provide a more efficient “sequential” random search method for the ESPAR antenna. In this method, a plurality of loaded reactance values are randomly changed at the same time. Objective function value before and after the change (eg cross-correlation coefficient)
Are calculated and then the calculated values are compared. If the change leads to an increase in the objective function value, this is accepted. If not, it is rejected and another new random change is attempted. Experiments have shown that the sequential random search method improves the performance of the adaptive ESPAR antenna over the case of the pure random search method according to the second conventional example.

The present inventors have proposed a sequential random search method for adaptive control of the ESPAR antenna. The proposed method considers the property of local continuity of the curved surface of the objective function. The experimental results show that the proposed sequential random search method is 1.
It is shown to provide a 7 dB improved average SINR gain. Furthermore, in the proposed sequential random search method,
The Gaussian random vector R (n) works better than the uniform distribution. Mean S for Gaussian distribution
The INR gain is about 0.8 dB larger than in the case of uniform distribution.

The control device for the array antenna can be easily mounted on an electronic device such as a notebook computer or PDA as an antenna for a mobile communication terminal.
Regardless of which direction the main beam is scanned in the horizontal plane, all the non-exciting elements effectively function as a director or a reflector, and it is also very suitable to control the directional characteristics with respect to incoming waves and a plurality of interference waves.

[0081]

As described above in detail, according to the array antenna control method of the present invention, in the control method of the ESPAR antenna, the reactance value of each variable reactance element is randomly perturbed from a predetermined initial value. When set, it is generated with the same signal pattern as the received signal when the learning sequence signal included in the radio signal transmitted from the transmitter of the other party is received by the array antenna, and the learning sequence signal. A predetermined cross-correlation coefficient before and after perturbation with the learning sequence signal is calculated, and when the cross-correlation coefficient before and after perturbation increases, the corresponding reactance value is selected and set. The step of repeating the random perturbation and setting based on the reactance value of the variable reactance element is included. Therefore,
It is possible to learn so that the performance improves with each iteration of the search, and it is possible to significantly reduce the convergence time to the optimal solution. This reduces the amount of calculation and does not require a long learning sequence signal.

In the array antenna control method, the initial value preferably has the maximum cross-correlation coefficient of the reactance values of the variable reactance elements corresponding to a plurality of predetermined radiation patterns. It is a reactance value of each variable reactance element corresponding to one radiation pattern. Therefore, by starting the search from the optimum initial value, the convergence time to the optimum solution can be greatly shortened, and the amount of calculation can be reduced.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of an array antenna control device according to an embodiment of the present invention.

2 is a cross-sectional view showing a detailed configuration of the array antenna apparatus 100 of FIG.

3 is a graph showing a range parameter b (n) and a variance σ (n) of a random vector R (n) generated by the adaptive controller 20 of FIG.

4A is a graph showing a probability density of a random vector R (n) that perturbs a bias voltage vector V (n), and FIG. 4B is an objective function value J due to the perturbation.
It is a graph which shows the change of.

5 is a flowchart showing adaptive control processing of an ESPAR antenna by a sequential random search method executed by the adaptive control type controller 20 of FIG.

FIG. 6 is a flowchart showing a bias voltage vector initial value selection process (step S2) that is a subroutine of FIG. 5;

7 is a simulation result of the array antenna apparatus 100 of FIG. 1, showing a complement 1-of a cumulative distribution function value Pr (Z <z) with respect to SINR gain when a uniformly distributed random vector R (n) is used. It is a graph which shows Pr (Z <z).

8 is a simulation result of the array antenna apparatus 100 of FIG. 1, showing a complement 1-Pr of a cumulative distribution function value Pr (Z <z) with respect to SINR gain when a Gaussian-distributed random vector R (n) is used. It is a graph which shows (Z <z).

9 is a simulation result of the array antenna apparatus 100 of FIG. 1, showing SINR gains when a uniformly distributed random vector R (n) is used and when a Gaussian distributed random vector R (n) is used. It is a graph which shows the comparison of the average value of.

10 is a simulation result of the array antenna apparatus 100 of FIG. 1, showing cumulative distributions when a uniformly distributed random vector R (n) is used and when a Gaussian distributed random vector R (n) is used. Function value Pr (Z
6 is a graph showing a comparison of complement 1-Pr (Z <z) of <z).

11 is a simulation result of the array antenna apparatus 100 of FIG. 1, showing S for a step parameter τ of variance of a Gaussian-distributed random vector R (n).
It is a graph which shows the comparison of the average value of INR gain.

FIG. 12 is a simulation result of the array antenna apparatus 100 of FIG. 1, showing a complement 1-of a cumulative distribution function value Pr (Z <z) for different input SNRs when a Gaussian-distributed random vector R (n) is used. Pr (Z <
It is a graph which shows z).

FIG. 13 is a graph showing a comparison of the complement 1-Pr (Z <z) of the cumulative distribution function value Pr (Z <z) between the conventional random search method and the sequential random search method of the present invention.

[Explanation of symbols]

A0 ... Excitation element, A1 to A6 ... Non-excitation element, 1 ... Low noise amplifier (LNA), 2 ... Down converter, 3 ... A / D converter, 4 ... demodulator, 5 ... coaxial cable for feeding, 11 ... Ground conductor, 12-1 to 12-6 ... Variable capacitance diode, 20 ... Adaptive control type controller, 21 ... Learning sequence signal generator, 22 ... Bias voltage table memory, 100 ... Array antenna device.

   ─────────────────────────────────────────────────── ─── Continued front page    (72) Inventor Takashi Ohira             2-2 Kodai, Seika-cho, Soraku-gun, Kyoto             International Telecommunications Basic Technology Research Institute Co., Ltd. F-term (reference) 5J020 AA03 BA02 BA04 BC08 DA03                 5J021 AA01 AB02 BA01 DB00 EA04                       FA02 FA30 GA01 HA05

Claims (3)

[Claims] 1. An excitation element for receiving a radio signal,
A plurality of non-excitation elements provided at a predetermined distance from the excitation element, and a plurality of variable reactance elements respectively connected to the plurality of non-excitation elements, each reactance set in each variable reactance element By changing the value, each of the non-excitation elements is operated as a director or a reflector, and in the array antenna control method of changing the directional characteristics of the array antenna, the reactance value of each variable reactance element is set to a predetermined value. When perturbed and set randomly from the initial value, the received signal when the learning sequence signal included in the radio signal transmitted from the transmitter of the other party is received by the array antenna, and the same as the above learning sequence signal. Predetermined cross-correlation before and after perturbation with a learning sequence signal generated with a signal pattern After calculating the coefficient and selecting and setting the corresponding reactance value when the cross-correlation coefficient before and after perturbation increases, set by randomly perturbing from the reactance value of each selected variable reactance element An array antenna control method comprising the step of repeating. 2. The variable reactance element corresponding to one radiation pattern having the maximum cross-correlation coefficient among the reactance values of the variable reactance elements corresponding to a plurality of predetermined radiation patterns as the initial value. The array antenna control method according to claim 1, wherein the reactance value is 3. The plurality of radiation patterns include: (a) a plurality of sector beam patterns each having a maximum gain in a direction from the excitation element to each non-excitation element; and (b) each of the excitation beam elements adjacent to each other. The array antenna control method according to claim 2, further comprising at least one set of a plurality of sector beam patterns each having a maximum gain in a direction toward an intermediate position between the non-excitation elements.