SU1520478A1 - Adaptive device for identifying linear objects - Google Patents

Abstract

The invention relates to the management of stationary objects and can be widely used in the design and testing of various systems and objects. The purpose of the invention is to improve the accuracy of identification. This is achieved by introducing the fourth and fifth adders, which makes it possible at each identification step to refine the values of the diagonal elements of the dispersion matrix of the vector of identifiable parameters. 3 il.

Description

The invention relates to the stationing of stationary objects and can be widely used in the design and testing of various systems and objects. .

The purpose of the invention is to improve the accuracy of identification.

FIG. 1 shows a functional diagram of the proposed device; in fig. 2 - functional diagram of the unit for calculating dispersions; in fig. 3 is a diagram of the first RAM block.

The proposed device contains (FIG. 1) an identification object 1, a first RAM unit 2, a dispersion calculation unit 3, a first multiplication unit 4, a third adder 5, a third multiplication unit 6, a fourth adder 7, a first memory unit 8, a fifth adder 9, fourth multiply unit 10, second RAM unit 11, first adder 12, first

dividing unit 13, second multiplying unit 14, second adder 15.

The dispersion calculation unit 3 comprises (FIG. 2) a second block 16 pgms, a fifth multiplication unit 17, a sixth multiplication unit 18, an eighth adder 19, a seventh adder 20, a seventh multiplication unit 21, a third operational block 22, a sixth adder 23 , second division unit 24, fourth operative memory unit 25.

The operational memory unit 2 contains (FIG. 3) a control signal register 26, a matrix B element memory 27, a first matrix multiplier 28, an adder 29, a vector of identifiable parameters vector memory 30, a matrix element memory 31, a second matrix multiplier 32, register 33 output signals, block 34

sd

differentiation, memory device 35.

 The device works as follows.

The evaluation of the identifiable parameters is carried out according to the formula

X (K + 1) X (K) -) f (K + 1) R (K + 1) X (K) + + J (K + 1) F (K + 1) (K + 1) Y (K +1) -F (K + 1) X (K),

 If | TR (K + T) ii

one

(K) + f (K + T) Q- (K + T) F (K + T) 11

 (K + 1) R (K). + F (K + 1) Q (K + 4-1) F (K + 1) R; + Z: F (i) Q (i) F (i);

X (K) is the vector estimate of the identifiable parameters of the object at the Kth step;

R is the inverse a priori dispersion matrix of the vector of identifiable object parameters;

R (K + 1) is the inverse dispersion matrix of the vector of the identified parameters of the object at the K + 1 st assessment step;

Q (K + 1) is the inverse dispersion matrix of the measurement noise vector at the K + 1-M estimation step;

F (K + 1) is the size matrix (Р-п), formed from the values of the input and output signals of the identification object according to the formula F (K + 1) Y (K-H), Y (K), ... , Y (K-n),

U (K), U (K-1), ..., U (K-nj-l),

where U (K) is the value of the input signal; Y (K) is the value of the output signal;

{(K + l) is some numeric sequence.

In the second memory block 16, the coefficients aj are calculated according to the magnitude of the smoothing interval Q and the order of the final difference, g, calculated previously

)

r-j

(g)

(2m-f1) (2r) j. (T-j) l

, 1, ..., r.

In the third memory block 22, a priori dispersion is recorded 5

0

Q

five

0

on the matrix of the interference vector Q measurements. The signals from the input and output of the object 1 are recorded in the registers of the first RAM block 2. The input signals and (K) are recorded sequentially in the control signal register 26, and the output Y (K) in the output signal register 33. In the memory devices 27, 31 and 30, the matrices B, A and the initial value vector are identified in advance; i 1x of the parameters X (0)

The elements of the matrix F (K) are derivatives of the output signals Y (K) by identifiable parameters X (K),

Therefore, to obtain its elements, it is necessary to obtain a vector of predictable parameters for the time K, to form the matrix F (K) by differentiating the values of Y (K) over X (K).

From the outputs of the first 28 and second 32 matrix multipliers, signals BU (K-l) and AX (K-1) cc are run in adder 29, recorded in memory 30 and fed to block 34, to the other input of which signal Y (K) is received.

A matrix is written to memory 35.

0

0

five

five

As soon as in the first block 2 of the RAM, 1 + hl-m values of the input and output signals will be accumulated, the device begins to estimate the identifiable parameters of the object, i.e. estimation is performed with a delay of 1 + r + m cycles. The matrix F (K + 1) formed in the first block 2 of the working memory enters the first 4, third 6, and fourth 10 multiplications blocks. In addition, the values of the output signals Y (K + ro + j + 1) and Y (K-m + j) from the corresponding registers of the first RAM block 2 go to the dispersion calculation unit 3, the fifth fifth and sixth 18 multiplications blocks . The values of the coefficients a come from the second block 16 of the memory into the fifth 17 and sixth 18 blocks of multiplication. The fifth 17th and sixth 18th block t forms the products (K + m + j + 1) and a: .y (). Values aj-Y (K + m + j + 1) of the n

multiplication unit 17 is fed to the summing inputs of adders 19 and 20. The values a; .Y (K-m + j) from the sixth multiplication unit 18 are fed to the second summing input of adder 19, where the values

X and j Y (K + m + j +1) + Y (K-m + j)

 -on

and subtracting the input of the adder 20, where

values are formed

Ta; Y (K + m + j + 1) -Y (K-m + j) 1. ) Values obtained from adders

19 and 20 arrive in the seventh multiplication unit 21, where their products are formed. The obtained products from the seventh postz multiplication unit 21 are placed in the sixth adder 23. Here, from the third operating memory unit 22, the value of the prior dispersion matrix of the measurement interference vector Q is supplied. The sixth adder 23 generates the values of the dispersion matrix of the measurement vector Q (K + 1) which is fed to the third RAM block 22 is recorded instead of a priori data for use in the next steps of calculating the dispersion matrix. The value of the matrix Q (K + 1) is also supplied to the second dividing unit 24. Since the dispersion matrix is diagonal, to obtain an inverse matrix, it is necessary and sufficient to replace its diagonal elements with their inverse values. The inverse dispersion matrix of the measurement interference vector formed in the second division unit 24 is recorded in the fourth RAM memory unit 25. The value of the inverse matrix Q (K + 1 of the fourth RAM block 25 is supplied to the first 4 and fourth 10 blocks of i. The first multiplication unit 4 forms the product F (K + 1) Q- (K + 1) F (K + O which is fed to the adder 7, in which the accumulation of values

k + n

F (i)) F (i).

This matrix serves to correct the elements of the matrix U previously written in the first memory block 8. From the first block of 8 memories, the values of the a priori dispersion matrix of the vector of the identified parameters of the object K are received at the eighth adder 9, where c3 are mixed with the elements of the matrix

(i) (i) F (i),

;

coming there from the adder 7. From the output of the first adder 9 matrix elements

 .K-t-l

R (K-H) R; +2: F (i) Q-4i) F (i)

arrive at the first adder 12 and 0 the second multiplication unit 1A. The first adder 12 forms the norm of the matrix

II R

five

0

0

K + i

 - ZlF CDQ-cDFCDlf,

I

equal to the sum of the modules of all its elements. The obtained matrix norm from the first matrix 12 enters the first division block 13, where the values yCK + l) are formed, equal to

Ok-il -I

| k; + z: F (i) (i) F (i) iij,

The values of Y (K + 1) go into blocks 10 and 14 of multiplication. The second multiplication unit 14 from the second operative memory block 11 receives the value of the identification object parameter evaluation,

calculated on the previous clock. I

So in the second block 14

multiplications are calculated

0 1

jp (K + 1) R (K + 1) X (K), which are fed to the subtracting input of the second adder 15, From the second memory block 11, the evaluation values X (K) are fed to the third multiplication unit 6,

5 where the value of F (K + 1) X (K) is formed. This value is fed to the subtracting input of the third adder 5, to the summing input of which the values of the output signal Y (K + 1) are supplied from the first RAM block 2, and the third adder 5 generates the values

Y (K + I) -F (K + I) X (K), which are supplied to the fourth multiplication unit 10. The fourth and fifth multiplication unit forms the values

y (K + i) F (K-n) (K + I) Y (K + I) X

F (K + 1) X (K),

which are fed to the summing input of the second adder 15. To the second summing input of the second adder 15, the evaluation value X (K) from the second memory block 11 is applied. In this way, the adder 15 generates the values of the vector for estimating the identifiable parameters of the object, equal to the K + 1-M step

X (K + 1) X (R) - | (K-H) R (K-f) X (K) +

+ | C (K + i) F (K + i) (K + I) Y (K + I) -F (K + 1) X (K).

The synchronization of the device is as follows.

In a pre-zeroed device, pre-calculated prior data is recorded. So, for example, the coefficients a calculated from the data on the size of the smoothing interval and the order of the finite difference r are recorded in the second memory block 16, and the priori Dispersive matrix of the measurement noise vector Q is recorded in the third memory block 22. In the device, synchronous prin

accumulation and calculation zip

No information and used conveyor mode. In this case, the pumping mode of the input information is determined by the execution time of the calculation process. The results of calculations are given at the same rate. In the above structure, signals are processed in each block as they accumulate in the input registers. Thus, various elements of the device simultaneously process information using their own algorithms with different completion times and then transfer the results of the solutions to the following blocks, the operation of which begins after receiving the previous values of the calculated information.

The overall operation of the blocks is synchronized with a clock pulse generator (not shown).

In comparison with the known device, the proposed device has a higher identification accuracy, due to the possibility of identifying at each step of identification the values of the diagonal elements of the dispersion matrix of the vector of identifiable parameters of the object and thereby exclude

methodological estimation errors, St - I

0

0

five

- 35

40

45 50

It is unknown with the inaccuracy of setting an a priori dispersion matrix of a vector of identifiable parameters of an object.

Claims (1)

Invention Formula An adaptive device for identifying linear objects comprising a dispersion calculation unit, a first memory block, a first RAM unit and a first multiplication unit connected in series, a first adder sequentially connected, a first division unit, a second multiplication unit, a second adder, and a second unit the memory unit, the third multiplication unit, the third adder and the fourth multiplication unit, the second input of which is connected to the second input of the third multiplication unit and the first input of the first multiplication unit. The third input is from the output ohm of the first dividing unit, and the output with the second input is the second of the adder, the third input of which is connected to the output of the second RAM memory unit and the second input of the second multiplication unit, the second output of the first RAM memory unit is connected to the second input of the third adder, the first and the second information input - respectively to the input and output of the object, the third and fourth outputs - respectively to the first and second inputs of the dispersion calculation unit connected by the output to the second INPUT5 The first multiplication unit and the fourth input of the fourth multiplication unit, differing in that, in order to improve the identification accuracy, the fourth and fifth adders are introduced into the device, the fourth adder is connected to the first input of the fifth adder, the second input of which is connected to the output of the first memory block, and the output is connected to the input of the first adder and to the third input of the second multiplication unit, the input of the seventh adder is connected to the output of the first multiplication unit. 26 21 thirty 31 uH 2nd P R Je 55 J / f f 4V Phie. five