RU2279125C1 - Device for parametric estimation of message streams distribution law - Google Patents

Abstract

FIELD: radio engineering, computer science.

SUBSTANCE: device for parametric estimation of message streams distribution law contains: input amplifier, parameters computation block, average arithmetic values computation block, block for determining distribution type, distribution computing device, control block, intensiveness analysis block, block for analyzing trustworthiness of messages. Thus, it is possible to estimate parameters of distribution of homogeneous and heterogeneous streams of multi-packet messages with high trustworthiness, specific signs of pauses for which may be identified both quantitatively and qualitatively - untrustworthily, due to comparative analysis in block of bit series received in binary code from position of their match to signs of pauses and mathematically correct transformation, on basis of neuron networks theory, of untrustworthily identified signs of pauses to form, useable for performing procedure of parametric estimation of message streams distribution law.

EFFECT: increased trustworthiness if estimating under conditions, appropriate for real functioning process of multi-channel radio-communication networks and local area networks under conditions of untrustworthy identification of signs of pauses between multi-packet messages, circulating in given networks.

4 cl, 8 dwg

Description

The invention relates to radio engineering and computer engineering and is intended for parametric estimation of the law of distribution of multi-packet message flows in multichannel radio communication networks (MLCS) and in local area networks (LAN).

A device is known for parametric estimation of the distribution law for ed. St. USSR No. 1024935, G 06 F 15/36, G 06 G 7/52, 1983, bull. 23, comprising an input amplifier, a memory unit, a consent check unit, a control unit, and a distribution function synthesizer.

The disadvantage of this device is the relatively low rate of flow analysis when evaluating the distribution of multi-packet message streams (MPS) in SMKRS and LAN, due to the need to iterate over a set of M combinations of distribution parameters of a message stream consisting of many packets.

A device is known for parametric estimation of the law of distribution of message flows, comprising an input amplifier, a parameter calculation unit, an arithmetic mean calculation unit, a distribution type determination unit, a distribution calculator, and a control unit (see RF patent No. 2094844, G 06 F 17/18, 1997 , bull. 30).

However, this device has a narrow scope, since it cannot be used to evaluate the distribution law of heterogeneous (mixed) streams of MPS occurring in the SMKRS and LAN and due to the presence of information sources in them, issuing as clearly expressed independent single user information messages (ISP) , and packs of ICPs whose lengths are distributed according to a geometric law. In this case, when approximating real statistics of incoming ICP flows, it is impossible to reduce them to three classical modifications of message flows.

The closest in technical essence to the claimed device (prototype) is a device for parametric evaluation of the law of distribution of message flows (see RF patent No. 2165100, 7 G 06 F 17/18, 2001, bull. 10), containing an input amplifier, the input of which a binary pulse sequence, a parameter calculation unit, an arithmetic mean calculation unit, a distribution type determination unit, a distribution calculator, a control unit and an intensity analysis unit are supplied, the clock output of the control unit being connected the clock input of the distribution calculator, the zeroing and control outputs of the control unit are connected respectively to the zeroing and control inputs of the parameter calculation unit, the information input of which is connected to the output of the input amplifier, the first and second information outputs of the parameter calculation unit are connected respectively to the first and second information inputs of the average calculation unit arithmetic values, the first information output of which is connected to the first information input of the distributor computer and with the input of the distribution type determining unit, the first, second and third modification outputs of which are connected respectively to the first, second and third modification inputs of the distribution calculator, the second information input of which is connected to the second information output of the arithmetic mean value calculation unit, the output of the distribution calculator is the output devices, the input of the intensity analysis unit is connected to the second information output of the parameter calculation unit, the first and second parameters of sul intensity analysis unit outputs respectively connected to first and second inputs of the calculator parametric distribution control output intensity analysis unit connected to the control input of the arithmetic unit calculating values.

In the prototype, it is possible to evaluate the distribution parameters (message intensity - λ _c and the ratio of information lengths of packets and messages - ξ) of both homogeneous and heterogeneous (mixed) streams of MPS - streams that periodically change their intensity during the operation of the SIRC and LAN, thereby expanding the scope of the device.

However, the prototype has a drawback - the relatively low reliability of the assessment under conditions of inaccuracy (insufficiency, incompleteness and inconsistency) of the signs (parameters) of the circulating flows of multi-packet messages, i.e. the relatively low reliability of assessing the state of the flows of MPS SMKRS and LAN with unreliable (insufficient, incomplete) identifiable signs. This device allows with high reliability to estimate the distribution parameters of only those homogeneous and heterogeneous (mixed) message flows, the initial signs (parameters) of the pauses of which are quantitatively set, consistently and fully identified, while the signs of the start and end of a pause for a large number of high-speed message flows actually circulating in computing and other telecommunication networks can only be identified qualitatively (at a qualitative level - unreliably, disadvantages chno, incomplete), based on the views of a large number of experts.

The "signs" of pauses in a message flow are defined as the established time intervals between messages - time intervals between the last bit of the flag for the end of the information part of the previous packet (the last in the previous message) and the first bit of the flag for the start of the information part of the next packet (the first in the next message), as well as the time of identifying the flag bit of the end of the information part of the previous packet (the last in the previous message) and the identification time of the flag bit of the beginning of the information part of the next chum (the first in the next post).

The aim of the invention is the creation of a device for parametric estimation of the law of distribution of message flows, providing increased reliability of the assessment in the conditions inherent in the real process of functioning of the SMKRS and LAN, i.e. in case of unreliability (insufficiency, incompleteness and inconsistency) of identifying pause signs of circulating streams of multi-packet messages, a device capable of highly reliably estimating the distribution parameters of homogeneous and heterogeneous (mixed) streams of MPS, the identifying signs of pauses of which (signs of the start and end of a pause between messages) can be identified both quantitatively and qualitatively - unreliably, incompletely.

This goal is achieved by the fact that in the known device for parametric estimation of the law of distribution of message flows, containing an input amplifier, the input of which is the input of the device, a parameter calculation unit, an arithmetic mean calculation unit, a distribution type determination unit, a distribution calculator, an intensity analysis unit, and a unit control, the clock output of which is connected to the clock input of the distribution computer, the control output of the control unit is connected to the control input of the block calculation of parameters, the information output of which is connected to the first information input of the arithmetic mean value calculation unit, the first information output of which is connected to the first information input of the distribution calculator and to the input of the distribution type determination unit, the first, second and third modification outputs of which are connected to the first, second and the third modification inputs of the distribution calculator, the second information input of which is connected to the second information output arithmetic mean value calculation unit house, the output of the distribution calculator is the device output, the first and second parametric outputs of the intensity analysis unit are connected respectively to the first and second parametric inputs of the distribution calculator, the control output of the intensity analysis unit is connected to the control input of the arithmetic mean value calculation unit, it is additionally introduced message reliability analysis unit, the output of the input amplifier being connected to the information input of the unit and the calculation of the parameters and the information input of the message reliability analysis unit, the information output of which is connected to the second information input of the arithmetic mean value calculation unit and the input of the intensity analysis unit, the nulling output of the control unit is connected to the nulling input of the message reliability analysis unit, the verification output of which is connected to the verification input block computing parameters.

The message reliability analysis block consists of a pause controller, a message controller and a message counter, the output of which is the information output of the block, the pause controller input is the information input of the block, the pause controller output is connected to the message controller input, the output of which is the verification output of the block and connected to the direct input a message counter, the nulling input of which is the nulling input of the message reliability analysis unit.

The parameter calculation block consists of a packet header decoder, a multi-packet message length information counter and a packet information message length ratio calculator, the output of which is an information output of the block, the packet header decoder input is a block information input and connected to the first input of a multi-packet message length information counter packet header decoder is connected to the second input of the counter of information length of multi-packet messages, verification the first input of which is a verification input unit, an output information multipacket message length counter is connected to the direct input information calculating ratios of the lengths of packets and messages, which control input is a control input parameter calculating unit.

Due to the new set of essential features due to the introduction of a message reliability analysis unit that provides a comparative analysis of bit sequences arriving in binary code from the point of view of their correspondence to pause features and the conversion of unreliably (incompletely) identified pause features, the claimed device allows preliminary analysis and transformation of pause features identified unreliably (incompletely) to a species suitable for the implementation of the parametric procedure Who assessment message flow distribution law, which leads to higher reliability evaluation under the conditions inherent in the functioning of the real process SMKRS and LAN - in conditions of uncertainty (lack of, incomplete and often contradictory) signs pauses circulating multipacket message flows.

The analysis of the prior art made it possible to establish that analogues that are characterized by a combination of features identical to all the features of the claimed technical solution are absent, which indicates the compliance of the claimed device with the patentability condition of "novelty".

Search results for known solutions in this and related fields of technology in order to identify features that match the distinctive features of the claimed object from the prototype showed that they do not follow explicitly from the prior art. The prior art also did not reveal the popularity of the impact provided by the essential features of the claimed invention, the transformations on the achievement of the specified technical result. Therefore, the claimed invention meets the condition of patentability "inventive step".

The claimed device is illustrated by drawings, on which are presented:

figure 1 - structural diagram of a device for parametric evaluation of the law of distribution of message flows;

figure 2 is a structural diagram of a block analysis of the reliability of messages;

figure 3 is a structural diagram of a message controller;

figure 4 is a structural diagram of a neural network transmitter;

figure 5 is a structural diagram of a unit for calculating parameters;

figure 6 is a structural diagram of a block for calculating arithmetic mean values;

Fig.7 is a structural diagram of a control unit;

on Fig is a structural diagram of a block analysis of intensity.

The device for parametric estimation of the law of distribution of message flows, shown in figure 1, consists of an input amplifier 1, the input of which is the input of the device, a unit for calculating parameters 2, a unit for calculating arithmetic mean values 3, a unit for determining the type of distribution 4, a distribution calculator 5, block control 6, intensity analysis unit 7 and message validation analysis unit 8. Clock output 61 of control unit 6 is connected to clock input 56 of distribution calculator 5, control output 63 of control unit 6 is connected to the control input 23 of the parameter calculation unit 2, the information output 24 of which is connected to the first information input 31 of the arithmetic mean value calculation unit 3, the first information output 33 of which is connected to the first information input 51 of the distribution calculator 5 and to the input of the distribution type determination unit 4, the first 41, the second 42 and the third 43 modification outputs of which are connected respectively to the first 53, second 54 and third 55 modification inputs of the distribution calculator 5, second and information input 52 of which is connected to the second information output 34 of the arithmetic mean value calculation unit 3. The output 59 of the distribution calculator 5 is the output of the device, the first 73 and second 74 parametric outputs of the intensity analysis unit 7 are connected respectively to the first 57 and second 58 parametric inputs of the distribution calculator 5 , the control output 72 of the intensity analysis unit 7 is connected to the control input 35 of the arithmetic mean value calculation unit 3. The output of the input amplifier 1 is connected to inf by the input of input 21 of the unit for calculating parameters 2 and the information input 81 of the unit for analyzing the reliability of messages 8, the information output 83 of which is connected to the second information input 32 of the unit for calculating arithmetic mean values 3 and the input 71 of the unit for analyzing the intensity 7, resetting the output 62 of the control unit 6 is connected to the resetting the input 84 of the block analysis of the reliability of messages 8, the verification output 82 of which is connected to the verification input 22 of the block computing parameters 2.

The message reliability analysis unit 8 (FIG. 2) is intended for fixing (recording) pauses, comparative analysis of pause signs in terms of their full correspondence to the space of pause signs, conversion of unreliably (incompletely) identified pause signs and determining current reliable values of a parameter characterizing the flow intensity messages - λ.

The message reliability analysis unit 8 consists of a pause controller 8.01, a message controller 8.02, and a message counter 8.03. The output 833 of the message counter 8.03 is the information output 83 of the block 8, the input 811 of the pause controller 8.01 is the information input 81 of the block 8, the output 812 of the pause controller 8.01 is connected to the input 821 of the message controller 8.02, the output 822 of which is the verification output 82 of the block 8 and is connected to a direct the input 831 of the message counter 8.03, the nulling input 832 of which is the nulling input 84 of the message analysis analysis unit 8.

The pause controller 8.01 of the message reliability analysis unit 8 is intended for fixing (recording) pauses characterizing the end of the message. The pause controller 8.01 can be technically implemented as a typical storage device based on a shift register with sequential input and output of information described in [Bystrov Yu.A., Velikson Ya.M., Vogman V.D. et al. Electronics: Reference Book / Ed. Bystrova Yu.A. - St. Petersburg: Energoatomizdat, 1996. S.291-292, Fig.6.7].

Message controller 8.02 of message reliability analysis unit 8 (Fig. 3) is intended for comparative analysis of pause features in terms of their full correspondence to the space of pause features and the conversion of unreliably (incompletely) identified pause features to a form suitable for carrying out a procedure for reliable parametric estimation of the law of flow distribution messages. The message controller 8.02 consists of a shifting element 8.02-1 and a neural network transmitter 8.02-2. The input 8021 of the biasing element 8.02-1 is the input 821 of the message controller 8.02, the reliable output 8022 of the biasing element 8.02-1 is connected to the inhibit input 8024 of the neural network transmitter 8.02-2, is combined with the output 8027 of the neural network computer 8.02-2 and is the output 822 of the message controller 8.02, the unreliable output 8023 of the biasing element 8.02-1 is connected to the enable input 8025 and the information input 8026 of the neural network calculator 8.02-2.

The shifting element 8.02-1 of the controller of messages 8.02 is intended for the implementation of the procedure of sequential comparison (in terms of the number of bits) of the signs of pauses and deciding on their mathematical nature - the signs of pauses are identified quantitatively (parametrically) or qualitatively (unreliably, incompletely) and need to be verified using analytically formulated subjective opinions (associations) of experts on specific values of time intervals (actually pauses) between messages. The shift element 8.02-1 can be technically implemented as a commercially available shift register for left shift, as shown in the literature [Sidorov AM, Gusev VV, Lebedev ON Fundamentals of pulsed and digital technology. - St. Petersburg: SPVVIUS, 1995. S.158-160, Fig. 5.28 (b)].

The neural network transmitter 8.02-2 of the message controller 8.02 (Fig. 4) is intended to carry out a procedure for converting pause signs that are identified unreliably (incompletely) to a form suitable for performing a parametric estimation procedure. The neural network transmitter 8.02-2 is a programmable parallel four-bit arithmetic logic element (ALE) with one enable and one inhibit input, n information inputs and one combined logic output, where n can take values from 2 to 23. The enable input M ALE is enable the input 8025 of the neural network calculator 8.02-2, the inhibitory input C ALE is the inhibitory input 8024 of the neural network calculator 8.02-2, n inputs of the ALE are combined and are the information input 8026 neural network calculates To 8.02-2 combined logic output STE F 8027 is an output of neural network calculator 8.02-2. The message controller 8.02 can be technically implemented as a parallel four-bit arithmetic-logic device based on the commercially available programmable CMOS microcircuit (microcircuit based on complementary field-effect transistors with a metal-oxide-semiconductor structure) of the 564 series (for example, K564IPZ), as shown in [ Shilo V.L. Popular digital circuits. Directory. - M.: Radio and Communications, 1987. S. 273-275, Fig. 2.70].

The message counter 8.03 of the message analysis analysis unit 8 is intended to determine reliable values of the intensity λ by pulses from the output of the message controller for a time interval determined by the control unit. The message counter 8.03 can be technically implemented on the basis of a commercially available counter, as shown in [Sobotka Z., Stari Ya. Microprocessor systems. - M .: Energoizdat, 1981. S.96-100].

The parameter calculation unit 2 shown in FIG. 5 is intended for analyzing the packet flow, determining the information length of the message and calculating the current values of the parameter ξ, which characterizes the ratio of the information lengths of the packets and messages. The parameter computing unit 2 consists of a packet header decoder 2.01, an information length counter for multi-packet messages 2.02, and a calculator for the ratio of information packet lengths and messages 2.03. The output of the calculator of the ratio of the information lengths of packets and messages 2.03 is the information output 24 of block 2, the input of the packet header decoder 2.01 is the information input 21 of block 2 and connected to the first input 221 of the information length counter of multi-packet messages 2.02, the output of the packet header decoder 2.01 is connected to the second input 222 the counter of information length of multi-packet messages 2.02, the verification input 223 of which is the verification input 22 of block 2, the output of the counter of information length of multi-packet messages with Communication 2.02 is connected to the direct input 231 ratio calculator lengths information packets and messages 2.03, a control input 232 which is the control input parameters 23 calculating unit 2.

The packet header decoder 2.01 of the parameter calculation block 2 is designed to analyze the binary pulse sequence arriving at its input and detect identified signs of the beginning and end of the information part of the packet in it. The packet header decoder 2.01 can be technically implemented on the basis of a commercially available decoder described in the book [Bogdanovich M.I., Grel I.N., Prokhorenko V.A. et al. Digital Integrated Circuits: A Guide. - Minsk: Belarus, 1991. S.432-436, Fig.4.46].

The information length counter of multi-packet messages 2.02 of the parameter calculation unit 2 is used to determine the current values of the parameter m _s - the information length of the message. The information length counter of multi-packet messages 2.02 can be technically implemented as a binary summing counter with sequential transfer on T-triggers, as described in the literature [Ugryumov EP Design of computer components and assemblies. - M.: Higher School, 1987. S.138-141, Fig. 9.1.].

The calculator of the ratio of information lengths of packets and messages 2.03 of the block computing parameters 2 is intended to determine the current values of the parameter ξ, which characterizes the ratio of information lengths of packets and messages. The calculator of the ratio of information lengths of packets and messages 2.03 can be technically implemented on the basis of a commercially available counter-divider, as shown in [Gusev VV, Lebedev ON, Sidorov AM Fundamentals of pulsed and digital technology. - St. Petersburg: SPVVIUS, 1996. S.175-182, Fig.5.46 (b)].

The input amplifier 1, which is part of the general block diagram, is designed to carry out the amplification procedure, normalizing the amplitude and duration of the input stream — a binary pulse sequence. Input amplifier 1 can be technically implemented as a Schmitt trigger, as shown in [Gusev VV, Lebedev ON, Sidorov AM Fundamentals of pulsed and digital technology. - SPb .: SPVVIUS, 1996. S.117-119, Fig. 4.12 (a)].

и

). Его структурная схема известна, описана, например, в прототипе (см. патент РФ №2165100, 7 G 06 F 17/18, 2001, бюл. 10) и представлена на фиг.6 данного описания. Блок вычисления средних арифметических значений 3 состоит из счетчика-делителя значений длительности 3.01, счетчика-делителя значений интенсивности 3.02 и элемента ЗАПРЕТ 3.03. Вход счетчика-делителя значений длительности 3.01 является первым информационным входом 31 блока вычисления средних арифметических значений 3, выход счетчика-делителя значений длительности 3.01 является первым информационным выходом 33 блока 3. Вход счетчика-делителя значений интенсивности 3.02 является вторым информационным входом 32 блока вычисления средних арифметических значений 3, выход счетчика-делителя значений интенсивности 3.02 подключен к разрешающему входу 331 элемента ЗАПРЕТ 3.03. Запрещающий вход 332 элемента ЗАПРЕТ 3.03 является управляющим входом 35 блока вычисления средних арифметических значений 3, выход элемента ЗАПРЕТ 3.03 является вторым информационным выходом 34 блока 3.The block for calculating arithmetic mean values 3, which is part of the general block diagram, is designed to obtain numerical values of flow parameters averaged over the observation interval (

and

) Its structural diagram is known, described, for example, in the prototype (see RF patent No. 2165100, 7 G 06 F 17/18, 2001, bull. 10) and is presented in Fig.6 of this description. The block for calculating arithmetic mean values 3 consists of a counter-divider of values of duration 3.01, a counter-divider of intensity values 3.02 and the element BAN 3.03. The input of the counter-divider of values of duration 3.01 is the first information input 31 of the block for calculating arithmetic mean values 3, the output of the counter-divider of values of duration 3.01 is the first information output of block 33 of 3. The input of the counter-divider of intensity 3.02 is the second information input 32 of the block for calculating arithmetic mean 3, the output of the counter-divider of intensity values 3.02 is connected to the enable input 331 of the element BAN 3.03. The inhibitory input 332 of the element FORBID 3.03 is the control input 35 of the unit for calculating the arithmetic mean values 3, the output of the element FORBID 3.03 is the second information output 34 of the block 3.

с единичным порогом и принятия решения о принадлежности однородного входного потока к одной из трех известных модификаций потоков многопакетных сообщений. Схема блока определения типа распределения 4 и принцип его действия известны и описаны в прототипе (см. патент РФ №2165100, 7 G 06 F 17/18, 2001, бюл. 10, фиг.6).The block for determining the type of distribution 4, which is part of the general block diagram, is designed to perform the procedure of comparing the values of the parameter of the ratio of duration

with a single threshold and deciding whether the homogeneous input stream belongs to one of the three known modifications of multi-packet message streams. The block diagram for determining the type of distribution 4 and the principle of its action are known and described in the prototype (see RF patent No. 2165100, 7 G 06 F 17/18, 2001, bull. 10, Fig.6).

Distribution computer 5, which is part of the general block diagram, is designed to obtain estimated values of the probability density of the flow. Its structural scheme and principle of operation are known and described in the prototype (RF patent No. 2165100, 7 G 06 F 17/18, 2001, bull. 10, figure 3).

The control unit 6, which is part of the general block diagram, is designed to adjust the observation interval (t _nab ), packet length (m _p ) and the number of packets (k). The structural diagram of the control unit 6 is known, described in detail in the prototype (RF patent No. 2165100, 7 G 06 F 17/18, 2001, bull. 10) and is presented in Fig.7 of this description. The control unit 6 contains a clock generator 6.01 and three dividers with a variable division ratio (DPKD) 6.02-6.04. The output of the clock pulse generator 6.01 is connected in parallel to the inputs of all three dividers with a variable division ratio, the output of the first of which (6.02) is the clock output 61 of block 6 and connected to the clock input 56 of the distribution calculator 5. The output of DPKD 6.04 and the output of DPKD 6.03 are respectively zeroing 62 and control 63 outputs of block 6 and are connected to zeroing 22 and control 23 inputs of block calculation parameters 2.

The intensity analysis unit 7, which is part of the general block diagram, is intended for the current estimation and comparison of the previous (λ ₁ ) and subsequent (λ ₂ ) values of the samples of the intensity parameter λ of the input stream. Its structural diagram is known, described in detail, for example, in the prototype (RF patent No. 2165100, 7 G 06 F 17/18, 2001, bull. 10) and is presented in Fig. 8 of this description. The intensity analysis unit 7 consists of a comparison element 7.01, a random access memory 7.02, a primary element And 7.03 and a secondary element And 7.04. The information input 701 of the comparison element 7.01, the input of random access memory 7.02 and the information input 706 of the secondary element And 7.04 are connected in parallel with each other, and their common input is input 71 of the intensity analysis unit 7, the auxiliary output 707 of the random access memory 7.02 is connected to the auxiliary input 702 comparison element 7.01, the output of which is the control output 72 of the intensity analysis unit 7 and is connected in parallel with the signal inputs 703 and 705 of the primary 7.03 and secondary 7.04 elements And accordingly Actually, the outputs of which are respectively the first 73 and second 74 parametric outputs of the intensity analysis unit 7. The information output 708 of the operative storage element 7.02 is connected to the information input 704 of the primary element And 7.03. The comparison element 7.01 is a digital comparison node and can be technically implemented as a comparison node (digital comparator), as shown in [Gusev VV, Lebedev ON, Sidorov AM Fundamentals of pulsed and digital technology. - St. Petersburg: SPVVIUS, 1996. S.149-152, Fig.5.19]. Operational memory element (OZE) 7.02 can be technically implemented on the basis of a commercially available programmable memory device in accordance with the description presented in [Gusev VV, Lebedev ON, Sidorov AM Fundamentals of pulsed and digital technology. - St. Petersburg: SPVVIUS, 1996. S.197-199, Fig.6.10].

A device for parametric estimation of the law of distribution of message flows works as follows. It is known that from the point of view of verification of the observed signs of pauses within the framework of the estimated message flow, it is possible to represent the space of signs of pauses in the form of a set (transposed vector) [1, 3, 5] of the form:

- транспонированный вектор, характеризующий (параметрически или неполно) пространство признаков пауз. Элементами данного вектора являются: t_бо - время идентификации бита флага окончания информационной части предыдущего пакета (последнего в предыдущем сообщении); t_бн - время идентификации бита флага начала информационной части следующего пакета (первого в следующем сообщении);

- временной интервал между сообщениями (от t_бо до t_бн). Существует возможность обнаружения (опознавания, детектирования) признаков пауз, идентифицируемых как количественно, так и качественно (недостоверно, неполно, противоречиво). Эта возможность реализуется путем последовательных преобразований с использованием нейросетевых вычислительных методов и алгоритмов, позволяющих осуществить переход от недостоверно (недостаточно, неполно) поставленной задачи оценивания к параметрической.Where

- a transposed vector characterizing (parametrically or incompletely) the space of signs of pauses. The elements of this vector are: t _bo is the identification time of the flag bit of the end of the information part of the previous packet (the last in the previous message); t _bn is the identification time of the flag bit of the beginning of the information part of the next packet (the first in the next message);

- the time interval between messages (from t _bo to t _bn ). There is the possibility of detecting (recognizing, detecting) signs of pauses, identifiable both quantitatively and qualitatively (unreliable, incomplete, contradictory). This possibility is realized by sequential transformations using neural network computational methods and algorithms that allow the transition from an unreliable (insufficient, incomplete) stated estimation task to a parametric one.

In this case, expression (1), which characterizes false information (incomplete knowledge) about the belonging of the analyzed pause feature to the pause feature space, has the form:

- транспонированный вектор пространства недостоверно (неполно) идентифицируемых признаков пауз, элементы которого получены с помощью экспертов в рамках простейших, аппаратно реализуемых нейросетевых вычислительных методов и алгоритмов. Данная трактовка модели идентификации признаков пауз позволяет ввести алгоритм последовательного сведения недостоверно (неполно) идентифицируемых признаков пауз к виду, обуславливающему возможность параметрической и однозначной идентификации этих признаков, повысить достоверность параметрического оценивания закона распределения потоков сообщений.Where

- the transposed space vector of unreliably (incompletely) identifiable signs of pauses, the elements of which were obtained with the help of experts in the framework of simple, hardware-based neural network computational methods and algorithms. This interpretation of the pause sign identification model allows us to introduce an algorithm for consistently reducing unreliably (incompletely) identifiable pause signs to a form that makes it possible to parametrically and unambiguously identify these signs, to increase the reliability of parametric estimation of the law of message flow distribution.

In the framework of the traditional expert system, to solve the problem of combining the opinions of experts whose knowledge is used in the form of pre-formed data on the possible belonging of a particular pause attribute to the attribute space, one of the typical computational algorithms of the theory of neural networks is used - a neural network extrapolating computational algorithm, or the so-called extrapolating neural network (ENS), which is a type of well-known computational models of associative memory [3].

A computational neural network algorithm (extrapolating neural network) of this class consists of two layers of calculators (neurons) - the input layer S _a and the output layer S _b . The input layer S _a is composed of n _Rin neurons possessing a set of forward and reverse links with n _O output layer neurons S _b, wherein the number of input and output images n equals (n = n _Rin = n _O), depends on the number of experts and appropriate amount device inputs of hardware implementation of a neural network algorithm. In our case, n can take values from 2 (two) to 23 (twenty three), by the number of inputs of the K564IP3 chip. In the ENS, the so-called cognitive map is used, which is completely defined by a matrix of relations of the form:

A cognitive map characterizes the causal relationship of individual signs of pauses and is formulated by experts, the principle of forming cognitive maps is described in detail in [2, 3].

In this case, n branches of the algorithm are responsible for the conversion of unreliable (incomplete) expert opinions on the belonging of a particular pause attribute to the attribute space. Each element w _{ij of the} matrix (3) defines a connection from the i-th element (a separate attribute) of the space of unreliably identifiable signs of pauses to the j-th element, with positive connections being encoded 1, negative ones -1, and the absence of connections between elements being encoded 0.

характеризующий множество известных как параметрически (достоверно), так и недостоверно заданных элементов (отдельных признаков пауз). Определяется, какие q элементов (отдельных признаков пауз), составляющих подмножество Ω_q из множества исходных идентифицируемых признаков пауз

в данный момент времени, заданы количественно (параметрически), а какие элементы (отдельные признаки пауз) из множества идентифицируемых признаков пауз, необходимые (предпочтительные) для осуществления задачи достоверного оценивания, - идентифицированы неопределенно (недостоверно, неточно). В целях получения достоверных результатов оценивания параметров потока необходимо реконструировать недостоверно (неполно, неточно) заданные признаки пауз, характеризующие интервалы между сообщениями.An input image is received at the input of a computational neural network algorithm (ENS)

characterizing many known both parametrically (reliably) and unreliably specified elements (individual signs of pauses). It is determined which q elements (individual signs of pauses) constituting a subset of Ω _q from the set of initial identifiable signs of pauses

at a given moment of time, quantitatively (parametrically) are specified, and which elements (individual signs of pauses) from the set of identifiable signs of pauses, necessary (preferred) for the task of reliable assessment, are identified indefinitely (unreliably, inaccurately). In order to obtain reliable results of the estimation of flow parameters, it is necessary to reconstruct unreliably (incompletely, inaccurately) specified signs of pauses characterizing the intervals between messages.

. Иными словами, вычислители (нейроны) входного слоя приводятся в начальные состоянияThe functioning of the computational neural network algorithm is as follows. The input layer S _{a of the} network is activated in an input manner characterizing the identifiable signs of pauses

. In other words, the calculators (neurons) of the input layer are brought into the initial state

The initialization of the calculators (neurons) of the output layer is carried out in accordance with the expression:

The input layer neurons are reduced to the state of the output layer neurons:

The new states of calculators (neurons) of the output layer are calculated for all i∈ Ω _q using the formula:

The calculation of expressions (6) and (7) is repeated until the ENS reaches a stable state:

принимающий ряд значений состояний, которые определяются на основе выражения (7) и полностью (достоверно) характеризуют интегрированное мнение экспертов о принадлежности конкретного признака паузы между сообщениями к пространству признаков.At the output of a computational neural network algorithm (ENS), we have an output vector

receiving a number of state values, which are determined on the basis of expression (7) and completely (reliably) characterize the integrated opinion of experts about the belonging of a particular pause sign between messages to the attribute space.

The considered computational neural network algorithm allows eliminating uncertainty (inaccuracy, incompleteness) when identifying pause signs, thereby unambiguously identifying messages in the MPS stream circulating in really functioning SMKRS and LANs. Clarification, reconstruction of unreliably (incompletely) identifiable signs of pauses can increase the objectivity of the estimated parameters of message flows, and ultimately increase the reliability of estimating their distribution law in the presence of unreliable (incomplete, inaccurate, and often contradictory) information about the state of the medium of propagation of radio waves, interference environment, your own network schedule, the behavior of the network management system and the influence of other destabilizing factors.

With this in mind, the law of distribution of message flows in the claimed device is evaluated. The studied input stream of the MPS of users of SMKRS and LAN in the form of a binary pulse sequence, amplified and normalized by the amplitude and duration of the input amplifier 1, is fed to input 21 of the unit for computing parameters 2, in which the current values of the parameter ξ are determined, which characterizes the ratio of the information lengths of packets and messages , and to the information input 81 of the message analysis analysis unit 8, in which the current reliable values of the parameter λ characterizing the flow rate are determined messages.

The parameter calculation unit 2 may be implemented in accordance with the circuit proposed in FIG. 5. The calculation in block 2 of the current values of the parameter ξ is as follows. The packet header decoder 2.01 analyzes the pulse sequence received at the input 21 of the parameter calculation unit 2 and, when it detects a sign of the beginning of the packet information part through the second input 222, starts the information length counter of multi-packet messages 2.02, the counting completion times are determined by the moments when the decoder 2.01 detects the end of the packet information part. The moments of arrival of the values of the information length of the message m _c to the calculator of the ratio of the information lengths of packets and messages 2.03 and the zeroing of the counter 2.02 are determined by the moments of occurrence of pulses at the verification input 223 of the information length counter of multi-packet messages 2.02 received through verification input 22 of the parameter calculation block 2 from the verification output 82 of the analysis the reliability of messages 8 and fixing the end of the message. In the calculator of the ratio of information lengths of packets and messages 2.03, the values ξ = m _p / m _{c are} determined for a fixed length of packets received in the SMKRS and LAN and installed by the control unit 6, by supplying, through the control input 23 of block 2, to the control input 232 of the calculator 2.03 pulses, the number which corresponds to the value of m _c . The obtained current values of the parameter ξ through the information output 24 of the parameter calculation unit are received in binary code to the first information input 31 of the arithmetic mean value calculation unit 3.

либо все вместе

идентифицированы недостоверно (неполно). В этом случае сдвигающий элемент 8.02-1 не записывает эту информацию, а выполняет функции транзитного узла, отправляя информацию через свой недостоверный выход 8023 на разрешающий 8025 и информационный 8026 входы нейросетевого вычислителя 8.02-2. Если на вход 8021 сдвигающего элемента 8.02-1 поступает в двоичном коде информация с признаками пауз в количестве пяти разрядов, значит все признаки пауз идентифицированы достоверно (однозначно, полно),

, сдвигающий элемент 8.02-1 записывает эту информацию и со своего достоверного выхода 8022 направляет ее на запрещающий вход 8024 нейросетевого вычислителя 8.02-2, а также через выход 822 контроллера сообщений 8.02 на прямой вход 831 счетчика сообщений 8.03 и через верификационный выход 82 блока анализа достоверности сообщений 8 на верификационный вход 22 блока вычисления параметров 2.The current reliable values of the parameter λ characterizing the intensity of the message flow are determined based on the detection and identification of pause signs between messages and are calculated in the message reliability analysis unit 8, which can be implemented in accordance with the scheme proposed in FIG. 2. Moreover, the elimination of uncertainty (unreliability, incompleteness) in the identification of signs of pauses is implemented as follows. From the information input 81 of block 8, the studied input stream of the MPS of the users of the SMKRS and LAN is fed to the input 811 of the pause controller 8.01, which captures (identifies, registers) pauses characterizing the end of the message. From the output 812 of the pause controller 8.01, the quantitatively or qualitatively (unreliable, incomplete) signs of pauses identified in the form of a binary pulse sequence are received at the input 821 of the message controller 8.02. The message controller 8.02 of the message reliability analysis unit 8 can be constructed in accordance with the circuit shown in FIG. In the message controller 8.02, a comparative analysis of the signs of pauses in terms of their full correspondence to the space of signs of pauses and the conversion of unreliably (incompletely) identified signs of pauses to a form suitable for the procedure of reliable parametric estimation is carried out as follows. Identified quantitatively or qualitatively (unreliably, incompletely) the signs of pauses in the form of a binary pulse sequence are fed to the input 8021 of the shifting element 8.02-1, which carries out a comparative analysis of the signs of pauses from the point of view of their full correspondence to the space of signs of pauses as follows. The shifting element 8.02-1 is designed to store five bits of incoming information, if the number of bits exceeds this figure, that means, from the point of view of mathematics, this information contains redundancy characterizing the cause-effect relationship of the signs of pauses, which means that any of the signs of pauses

either all together

identified unreliably (incomplete). In this case, the shifting element 8.02-1 does not record this information, but acts as a transit node, sending information through its unreliable output 8023 to the resolving 8025 and information 8026 inputs of the neural network transmitter 8.02-2. If the input 8021 of the shifting element 8.02-1 receives information in binary code with signs of pauses in the amount of five digits, then all the signs of pauses are identified reliably (unambiguously, completely),

, the shifting element 8.02-1 records this information and sends it from its reliable output 8022 to the inhibit input 8024 of the neural network calculator 8.02-2, as well as through the output 822 of the message controller 8.02 to the direct input 831 of the message counter 8.03 and through the verification output 82 of the reliability analysis block messages 8 to the verification input 22 of the parameter calculation unit 2.

The neural network calculator 8.02-2 of the message controller 8.02 is technically implemented on the basis of a programmable (from the point of view of the weight matrix (3) - causal cognitive relations formulated by experts) parallel four-digit ALE that implements the computational neural network algorithm (4) - (8) in accordance with the circuit depicted in figure 4. The conversion of the signs of pauses, identifiable unreliably (incompletely), to a form suitable for the implementation of the parametric estimation procedure, is carried out as follows. Reliable (complete) information with signs of pauses, arriving in binary code through the inhibit input 8024 of the neural network calculator 8.02-2 to the inhibit input C ALE, blocks the microcircuit. Unreliable (incomplete) information with signs of pauses is received in binary code through the enable input 8025 of the neural network transmitter 8.02-2 to the enable input M of the ALE, initiating the neural network computing process, and enters through the information input 8026 of the neural network transmitter 8.02-2 to n inputs of the programmable ALE.

, имеющие физический смысл информации с недостоверно (неполно) идентифицированными признаками пауз. Набор прямых и обратных связей n_вх с n_вых ЭНС, программно реализованный в рамках программируемого АЛЭ, позволяет учитывать при вычислениях значения элементов (w_ij) матрицы весов W (3) и получать на n выходах (n_вых) в результате арифметических вычислений в соответствии с вычислительным нейросетевым алгоритмом (4)-(8) экстраполированные значения элементов вектора

характеризующие предварительные решения о степени принадлежности конкретного признака паузы пространству признаков. При этом подача (и запись на хранение) в двоичном коде на любой i-ый вход АЛЭ (в любую i-ую ячейку памяти АЛЭ) значения недостоверно (неполно) идентифицируемого признака паузы инициирует выдачу в двоичном коде с i-го выхода АЛЭ (выхода i-го нейрона выходного слоя S_b) запрограммированного в соответствии с вычислительным нейросетевым алгоритмом (4)-(8) значения математически корректно преобразованного, относительно достоверно идентифицированного признака паузы. В результате, на выходе F логического элемента АЛЭ, осуществляющего операцию итоговой верификации (формирование обобщенного мнения), и на выходе 8027 нейросетевого вычислителя 8.02-2 получаем в двоичном коде информацию с достоверно (полно) идентифицированными признаками пауз, получаем количественные значения признаков пауз, полностью (достоверно) характеризующие интегрированное мнение экспертов о принадлежности конкретного признака паузы к пространству признаков.Programmable ALE, relying on the programmed values of the elements (w _ij ) of the weight matrix W (3) - analytically described cause-effect cognitive relationships formulated by experts, performs the calculation (extrapolation) procedure in accordance with the computational neural network algorithm (4) - (8). At the same time, the n inputs (1, ..., n) of the programmed ALE are equal to the n inputs (n _in ) of the calculators (neurons) of the input layer S _a ENS, to which the vector elements are supplied in binary code

having physical meaning of information with unreliably (incompletely) identified signs of pauses. Set of direct and reverse links n _Rin with n _O ENS, software implemented within the programmable STE, allows to take into account when calculating the values of elements (w _ij) weight matrix W (3) and to receive n outputs (n _O) as a result of arithmetic calculations in accordance with computational neural network algorithm (4) - (8) extrapolated values of vector elements

characterizing preliminary decisions on the degree of belonging of a particular sign to a pause to the space of signs. In this case, the supply (and recording for storage) in binary code to any i-th input of the ALE (to any i-th memory cell of the ALE) of the value of an unreliable (incomplete) identifiable sign of a pause initiates the output in binary code from the i-th output of the ALE (output i-th neuron of the output layer S _b ) programmed in accordance with the computational neural network algorithm (4) - (8) the values of mathematically correctly transformed, relatively reliably identified sign of a pause. As a result, at the output F of the logic element of the ALE performing the operation of final verification (formation of a generalized opinion), and at the output of 8027 neural network calculator 8.02-2, we obtain in binary code information with reliably (fully) identified signs of pauses, we obtain quantitative values of the signs of pauses, fully (reliably) characterizing the integrated opinion of experts about the belonging of a particular pause attribute to the attribute space.

Initially, reliable information from the reliable output 8022 of the shifting element 8.02-1 and verified information containing quantitative (relatively reliable) values of the identified signs of pauses obtained in the neural network calculator 8.02-2 are transmitted in binary code via the output 822 of the message controller 8.02 to the direct input 831 of the message counter 8.03 and through the verification output 82 of the message reliability analysis unit 8 to the verification input 22 of the parameter computing unit 2. In the parameter computing unit 2, this information is used It is used as a command that fixes the end of the message and is used to determine the moments when the information length of the message m _s arrives at the computer for calculating the ratio of information lengths of packets and messages 2.03 and resetting the counter 2.02 of parameter calculation block 2.

Current reliable values of the parameter λ characterizing the intensity of the message flow is determined by the message counter 8.03 by pulses from the output 822 of the message controller 8.02 for the time interval determined by the control unit 6 by applying pulses (frequency f _nb ) from the resetting output 62 of the control unit 6 through the resetting input 84 block analysis of the reliability of messages 8 to the resetting input 832 of the counter 8.03.

Thus, relying on the mathematical apparatus of neural networks, using identification criteria (causal cognitive relationships) formulated by experts, using mathematically correct calculations in the message reliability analysis unit 8, it was possible to achieve that output 833 of the message counter 8.03 via information output 83 the reliability analysis unit of messages 8 to the input 71 of the intensity analysis unit 7 and to the second information input 32 of the arithmetic mean value calculation unit 3, despite the hour th present in real communications channels (in the background noise and errors in the channels and SMKRS LAN) unreliability (incomplete) indicia pauses are supplied to a binary pulse sequence verified accurate current value of the parameter λ.

). Блок вычисления средних арифметических значений 3 может быть реализован по схеме, представленной на фиг.6. В случае, когда имеет место однородный (несмешанный) поток МПС пользователей СМКРС и ЛВС, когда источники выдают сообщения регулярно, расчет численных значений усредненных за интервал наблюдения параметров потока (

или

) происходит следующим образом. Значения параметра ξ в виде бинарной последовательности поступают на первый информационный вход 31 блока 3, а затем на вход счетчика-делителя значений длительности 3.01, производящего арифметическую операцию суммирования последовательно поступающих значений ξ и деление полученной суммы на количество отсчетов. На выходе счетчика-делителя значений длительности 3.01 и на первом информационном выходе 33 блока вычисления средних арифметических значений 3 в двоичном параллельном коде получаем значение параметра

, усредненного за интервал наблюдения. Достоверные значения параметра λ в виде бинарной последовательности поступают на второй информационный вход 32 блока 3, а затем на вход счетчика-делителя значений интенсивности 3.02, производящего арифметическую операцию суммирования последовательно поступающих значений λ и деление полученной суммы на количество отсчетов. На выходе счетчика-делителя значений интенсивности 3.02 и на разрешающем входе элемента ЗАПРЕТ 3.03 в двоичном параллельном коде имеем значение параметра

, усредненного за интервал наблюдения. На второй информационный выход 34 блока 3 с выхода элемента ЗАПРЕТ 3.03 значение

попадает лишь в том случае, если в данный момент на управляющий вход 35 блока 3 и на запрещающий вход элемента ЗАПРЕТ 3.03 не подан в двоичном параллельном коде сигнал с управляющего выхода 72 блока анализа интенсивности 7. В противном случае на второй информационный выход 34 блока 3 с выхода элемента ЗАПРЕТ 3.03 значение параметра

не подается. Интервал наблюдения для расчета параметров

и

устанавливается блоком управления 6 в зависимости от требуемой точности.The further stages of the parametric estimation of the law of distribution of message flows are implemented as follows. In the block for calculating arithmetic mean values 3, the first 31 and second 32 information inputs of which in the parallel code respectively receive the current values of the parameter ξ and the current reliable values of the parameter λ, the numerical values of the flow parameters averaged over the observation interval are determined (

) The unit for calculating arithmetic mean values 3 can be implemented according to the scheme shown in Fig.6. In the case when there is a homogeneous (unmixed) MPS stream of users of SMKRS and LANs, when the sources give out messages regularly, calculation of numerical values averaged over the observation interval of the flow parameters (

or

) occurs as follows. The values of the parameter ξ in the form of a binary sequence are sent to the first information input 31 of block 3, and then to the input of the counter-divider of the values of duration 3.01, which performs the arithmetic operation of summing the successively arriving values of ξ and dividing the resulting sum by the number of samples. At the output of the counter-divider of values of duration 3.01 and at the first information output 33 of the block for calculating arithmetic mean values 3 in binary parallel code, we obtain the parameter value

averaged over the observation interval. Reliable values of the parameter λ in the form of a binary sequence are sent to the second information input 32 of block 3, and then to the input of the counter-divider of the intensity values 3.02, which performs the arithmetic operation of summing the successively arriving values of λ and dividing the resulting sum by the number of samples. At the output of the counter-divider of intensity values 3.02 and at the enable input of the element BAN 3.03 in binary parallel code, we have the parameter value

averaged over the observation interval. To the second information output 34 of block 3 from the output of the element BAN 3.03 value

gets only if at the moment the signal from the control output 72 of the intensity analysis block 7 is not sent in binary parallel code to the control input 35 of block 3 and to the inhibitory input of the element BAN 3.03; otherwise, the second information output 34 of block 3 s output element BAN 3.03 parameter value

not served. Observation interval for calculating parameters

and

set by the control unit 6 depending on the required accuracy.

The control unit 6 can be implemented according to the scheme proposed in Fig.7. The formation of control pulse sequences is carried out by installing on the elements of a group of dividers with a variable division coefficient (DLC) the corresponding division factors of the pulse sequence coming from the output of the generator 6.01. Setting the coefficient of element 6.03 determines the value of t _nab = 1 / f _nab , setting the coefficient of element 6.04 determines the packet length accepted for this protocol at the moment of operation of the SMKRS and LAN (m _p ), setting the coefficient of element 6.02 determines the number of packets necessary for calculation (k )

и

соответственно, значение

с первого информационного выхода 33 блока 3 подается на вход блока определения типа распределения 4, в котором происходит его сравнение с единичным порогом. Структурная схема блока определения типа распределения 4 известна и описана в прототипе (см. патент РФ №2165100, 7 G 06 F 17/18, 2001, бюл. 10, фиг.6). Определение типа распределения потока МПС осуществляется следующим образом. Значения

, поступающие с выхода блока 3 в двоичном параллельном коде в цифроаналоговый преобразователь (ЦАП) 4.01 блока определения типа распределения 4, преобразуются в аналоговое напряжение, которое на компараторах 4.02, 4.03 блока определения типа распределения 4 сравнивается с пороговым напряжением, вырабатываемым генератором порогового напряжения 4.04 блока определения типа распределения 4. Значение порогового напряжения соответствует единичному значению

. На выходе 42 компаратора 4.02 появляется сигнал логической единицы в случае превышения значения

единичного значения и поступает на второй модификационный вход 54 вычислителя распределения 5, в обратном случае появление логической единицы происходит на выходе 41 компаратора 4.03 и поступает на первый модификационный вход 53 вычислителя распределения 5. При наличии логических нулей на выходах компараторов 4.02 и 4.03 принимается решение о том, что

=1, об этом свидетельствует наличие логической единицы на выходе 43 элемента ИЛИ-НЕ 4.05 блока определения типа распределения 4 и поступление этой логической единицы на третий модификационный вход 55 блока 5. Таким образом, по результатам сравнения путем подачи управляющего импульса на один из трех модификационных входов 53-55 вычислителя распределения 5 принимается решение о вычислении одного из трех известных и описанных в прототипе (см. патент РФ №2165100, 7 G 06 F 17/18, 2001, бюл. 10, выражения (1-3)) аналитических выражений для плотностей распределения вероятности по значениям параметров

и

поступающих с первого 33 и второго 34 информационных выходов блока 3 соответственно на первый 51 и второй 52 информационные входы вычислителя распределения 5.When at the first 33 and at the second 34 information outputs of block 3 we have simultaneously the values of the parameters

and

accordingly, the value

from the first information output 33 of block 3 is fed to the input of the block determining the type of distribution 4, in which it is compared with a unit threshold. The block diagram of the distribution type determination unit 4 is known and described in the prototype (see RF patent No. 2165100, 7 G 06 F 17/18, 2001, bull. 10, Fig.6). Determining the type of distribution of the MPS stream is as follows. Values

coming from the output of block 3 in binary parallel code to a digital-to-analog converter (DAC) 4.01 of the distribution type determination unit 4, is converted to an analog voltage, which is compared with the threshold voltage generated by the threshold voltage generator 4.04 on the comparators 4.02, 4.03 of the distribution type determination unit 4 determine the type of distribution 4. The value of the threshold voltage corresponds to a unit value

. At the output 42 of comparator 4.02, a signal of a logical unit appears in case of exceeding the value

unit value and goes to the second modification input 54 of the distribution calculator 5, otherwise the appearance of a logical unit occurs at the output 41 of the comparator 4.03 and goes to the first modification input 53 of the distribution calculator 5. If there are logical zeros at the outputs of the comparators 4.02 and 4.03, the decision is made , what

= 1, this is evidenced by the presence of a logical unit at the output 43 of the OR-NOT 4.05 element of the distribution type determination block 4 and the arrival of this logical unit to the third modification input 55 of block 5. Thus, according to the comparison results, by supplying a control pulse to one of the three modification the inputs 53-55 of the distribution calculator 5, a decision is made to calculate one of the three known and described in the prototype (see RF patent No. 2165100, 7 G 06 F 17/18, 2001, bull. 10, expressions (1-3)) analytical expressions for distribution densities yatnosti of parameter values

and

coming from the first 33 and second 34 information outputs of block 3, respectively, to the first 51 and second 52 information inputs of the distribution calculator 5.

The block diagram of the distribution calculator 5 is known and described in the prototype (see RF patent No. 2165100, 7 G 06 F 17/18, 2001, bull. 10, figure 3). Calculations in the distribution calculator 5 can be performed for various values of the number of packets (k), the choice of values is carried out by the control unit 6 by applying pulses to the clock input 56 of block 5 and to the first input 510 of element 5.02, the number of which corresponds to a given value of the number of packets. The arrival at the second input 511 of the element And 5.02 pulses from the first information input 51 of block 5 ensures that the output signal of the element And 5.02 calculator distribution 5 control signal, the number of pulses of which corresponds to a given value of the number of packets. This signal, arriving at the clock input 506 of the arithmetic logic element 5.01 of the distribution calculator 5, initiates the start of the calculation procedure. The calculated distribution values of the uniform MPS stream from the output 59 of the distribution calculator 5 are supplied in binary parallel code to the registration device.

) происходит так, как описано выше, а расчет численных значений параметра λ, характеризующего интенсивность потока, происходит следующим образом. Достоверные (верифицированные) значения параметра λ в виде бинарной последовательности поступают с информационного выхода блока анализа достоверности сообщений 8 на второй информационный вход 32 блока 3 для вычисления усредненного за интервал наблюдения параметра потока (

) и на вход 71 блока анализа интенсивности 7. Блок анализа интенсивности 7 может быть реализован в соответствии со схемой, изображенной на фиг.8. Значения параметра λ поступают параллельно на информационный вход 701 элемента сравнения 7.01 блока 7, на вход оперативного запоминающего элемента 7.02 блока 7 и на информационный вход 706 вторичного элемента И 7.04 блока 7. В оперативном запоминающем элементе 7.02 хранится значение параметра λ на предыдущем шаге - значение λ₁. С вспомогательного выхода 707 оперативного запоминающего элемента 7.02 значение λ₁ поступает на вспомогательный вход 702 элемента сравнения 7.01, который сравнивает значение предыдущего отсчета параметра (λ₁) со значением последующего отсчета (λ₂) - значением, поступившим на информационный вход 701 элемента сравнения 7.01. Сравнение производится по двум критериям: λ₁≠λ₂ и λ₁=λ₂. Если значения отсчетов равны (λ₁=λ₂), это подтверждает, что на входе устройства мы имеем однородный (несмешанный) поток МПС, а значит, элемент сравнения 7.01 не выдает сигнал на управляющий выход 72 блока 7 и на сигнальные входы элементов И 7.03 и 7.04. Это значит, что на управляющий вход 35 блока 3 и на запрещающий вход элемента ЗАПРЕТ 3.03 блока 3 с управляющего выхода 72 блока 7 сигнал не поступает, следовательно, на выходе элемента ЗАПРЕТ 3.03 и на втором информационном выходе 34 блока 3 в двоичном параллельном коде имеем значение параметра

и дальнейший расчет параметров закона распределения потоков сообщений происходит так, как описано выше.In the case when there is an inhomogeneous (mixed) MPS flow of users of SMKRS and LANs, when mixed-type sources issue messages irregularly - by single ICP or packets of ICP with lengths distributed according to a geometric law, calculation of numerical values of the flow parameter averaged over the observation interval (

) occurs as described above, and the calculation of the numerical values of the parameter λ characterizing the flow intensity occurs as follows. Reliable (verified) values of the parameter λ in the form of a binary sequence are received from the information output of the message analysis analysis block 8 to the second information input 32 of block 3 to calculate the stream parameter averaged over the observation interval (

) and to the input 71 of the intensity analysis unit 7. The intensity analysis unit 7 can be implemented in accordance with the circuit shown in Fig. 8. The values of the parameter λ are sent in parallel to the information input 701 of the comparison element 7.01 of block 7, to the input of the operative storage element 7.02 of block 7 and to the information input 706 of the secondary element And 7.04 of block 7. The value of parameter λ is stored in the operative storage element 7.02 in the previous step — the value λ ₁ . From the auxiliary output 707 of the random access memory element 7.02, the value λ _{1 is} supplied to the auxiliary input 702 of the comparison element 7.01, which compares the value of the previous reference parameter (λ ₁ ) with the value of the subsequent reference (λ ₂ ) - the value received at the information input 701 of the comparison element 7.01. The comparison is made according to two criteria: λ ₁ ≠ λ ₂ and λ ₁ = λ ₂ . If the values of the samples are equal (λ ₁ = λ ₂ ), this confirms that at the input of the device we have a homogeneous (unmixed) MPS stream, which means that the comparison element 7.01 does not give a signal to the control output 72 of block 7 and to the signal inputs of elements And 7.03 and 7.04. This means that the signal does not arrive at the control input 35 of block 3 and at the inhibitory input of the element BAN 3.03 of block 3 from the control output 72 of block 7, therefore, at the output of the element BAN 3.03 and at the second information output 34 of block 3, we have the value in binary parallel code parameter

and further calculation of the parameters of the law of distribution of message flows occurs as described above.

на второй информационный вход 52 блока 5. С информационного выхода 708 оперативного запоминающего элемента 7.02 блока 7 значение параметра λ₁ поступает на информационный вход 704 первичного элемента И 7.03, а в оперативном запоминающем элементе 7.02 записывается значение нового параметра. На информационном входе 706 вторичного элемента И 7.04 блока 7 имеем значение λ₂. Сигнал с выхода элемента сравнения 7.01, поступая на сигнальные входы 703 и 705 первичного и вторичного элементов И (7.03 и 7.04) соответственно, открывает эти устройства, позволяя получать на их выходах, а значит, и на первом 73 и втором 74 параметрических выходах блока 7 значения параметра λ₁ и λ₂ соответственно. Значения параметра λ₁ и λ₂ поступают на первый 57 и второй 58 параметрические входы вычислителя распределения 5, а затем на первый 507 и второй 508 параметрические входы арифметико-логического элемента 5.01, который вычисляет значения распределения в соответствии с аналитическим выражением для определения свертки соответствующих вероятностей (см. патент РФ №2165100, 7 G 06 F 17/18, 2001, бюл. 10, выражение (4)), причем наличие управляющего импульса на любом из трех модификационных входов 53-55 вычислителя распределения 5 в этом случае роли не играет. Вычисленные значения распределения неоднородного (смешанного) потока МПС с выхода 59 вычислителя распределений 5 подаются в двоичном параллельном коде на устройство регистрации.If the values of the samples are not equal (λ ₁ ≠ λ ₂ ), this confirms that at the input of the device we have a heterogeneous (mixed) MPS stream, which means that the comparison element 7.01 gives a signal to the control output 72 of block 7 and to the signal inputs of elements And 7.03 and 7.04. The signal from the control output 72 of block 7 is fed to the control input 35 of block 3 and to the inhibitory input of the element BAN 3.03, blocking its output and not allowing to give out values

to the second information input 52 of block 5. From the information output 708 of the operative storage element 7.02 of block 7, the parameter value λ _{1 is} supplied to the information input 704 of the primary element And 7.03, and the value of the new parameter is recorded in the operative storage element 7.02. At the information input 706 of the secondary element AND 7.04 of block 7, we have the value λ ₂ . The signal from the output of the comparison element 7.01, arriving at the signal inputs 703 and 705 of the primary and secondary elements And (7.03 and 7.04), respectively, opens these devices, allowing them to receive at their outputs, and therefore, at the first 73 and second 74 parametric outputs of block 7 the values of the parameter λ ₁ and λ _2, respectively. The values of the parameter λ ₁ and λ ₂ are fed to the first 57 and second 58 parametric inputs of the distribution calculator 5, and then to the first 507 and second 508 parametric inputs of the arithmetic logic element 5.01, which calculates the distribution values in accordance with the analytical expression to determine the convolution of the corresponding probabilities (see RF patent No. 2165100, 7 G 06 F 17/18, 2001, bull. 10, expression (4)), moreover, the presence of a control pulse at any of the three modification inputs 53-55 of the distribution calculator 5 does not play a role in this case . The calculated distribution values of the heterogeneous (mixed) MPS stream from the output 59 of the distribution calculator 5 are supplied in binary parallel code to the recording device.

As a result, at the output of block 5, we have reliable, recorded in binary code, verified using the mathematical apparatus of the theory of neural networks, distribution values of either a homogeneous or heterogeneous (or mixed, depending on the characteristics of the SMKRS and LAN) flow of uniquely identified multi-packet messages.

Thus, the analysis of the principle of operation of the inventive device for parametric estimation of the law of distribution of message flows shows the evidence of the fact that, along with the saved capabilities of estimating the distribution parameters of heterogeneous (mixed) streams of MPS, the device can increase the reliability of estimating the distribution parameters of message flows, the identification signs of pauses of which (signs beginning and end of the pause between messages) can be identified both quantitatively and qualitatively - not reliable, incomplete.

This device provides an increase in the reliability of estimation under conditions inherent in the real process of functioning of the SMKRS and LANs - in the conditions of unreliability (insufficiency, incompleteness and inconsistency) of identifying signs of pauses in the circulating flows of multi-packet messages due to the instability (variability, stochasticity) of the operating conditions of networks of this class, which expands the scope of the device, expands the functionality of the equipment, where the claimed device for A parametric estimate of the law of distribution of message flows will be used.

INFORMATION SOURCES

1. Wassermen F. Neurocomputer technology: Theory and practice. - M.: Mir, 1992 .-- 240 p.

2. Kosko, V. Fuzzy cognitive maps // International Journal of Man-Machine Studies. V.24. N.Y., 1986. P.16-22.

3. Shcherbakov M.A. Artificial neural networks. - Penza: PSTU, 1996 .-- 44 p.

4. Trachtengerts E.A. Computer decision support. - M .: SINTEG, 1998 .-- 342 p.

5. Gorban A.N., Rossiev D.A. Neural networks on a personal computer. - Novosibirsk: Science. Siberian Publishing Company RAS, 1996. - 146 p.

Claims (3)

1. A device for parametric estimation of the law of distribution of message flows, comprising an input amplifier, the input of which is an input to the device, a parameter calculation unit, an arithmetic mean calculation unit, a distribution type determination unit, a distribution calculator, an intensity analysis unit, and a control unit whose clock output is connected to the clock input of the distribution calculator, the control output of the control unit is connected to the control input of the parameter calculation unit, the information output to which is connected to the first information input of the arithmetic mean value calculation unit, the first information output of which is connected to the first information input of the distribution calculator and to the input of the distribution type determination unit, the first, second and third modification outputs of which are connected respectively to the first, second and third modification inputs of the calculator distribution, the second information input of which is connected to the second information output of the arithmetic mean calculation unit values, the output of the distribution calculator is the output of the device, the first and second parametric outputs of the intensity analysis block are connected respectively to the first and second parametric inputs of the distribution calculator, the control output of the intensity analysis block is connected to the control input of the arithmetic mean value calculation unit, characterized in that the block is additionally introduced analysis of the reliability of messages, and the output of the input amplifier is connected to the information input of the calculation unit parameter in and with the information input of the message reliability analysis unit, the information output of which is connected to the second information input of the arithmetic mean value calculation unit and the input of the intensity analysis unit, the nulling output of the control unit is connected to the nulling input of the message reliability analysis unit, the verification output of which is connected to the verification input of the calculation unit parameters. 2. The device for parametric estimation of the law of distribution of message flows according to claim 1, characterized in that the message reliability analysis unit consists of a pause controller, a message controller and a message counter, the output of which is the information output of the block, the input of the pause controller is the information input of the block, output the pause controller is connected to the input of the message controller, the output of which is the verification output of the block and connected to the direct input of the message counter, the zeroing input of which is reset yuschim login Message validity of the analysis unit. 3. A device for parametric estimation of the law of distribution of message flows according to claim 1, characterized in that the parameter calculation unit consists of a packet header decoder, an information length counter for multi-packet messages, and a calculator for the ratio of information lengths of packets and messages, the output of which is the information output of the block, input the packet header decoder is the information input of the block and is connected to the first input of the information length counter of multi-packet messages, the decoder output is Cove packets coupled to the second input information length multipacket message counter, the input of which the verification is a verification input unit, an output of the counter multipacket message length information is connected to the direct input information calculating ratios of the lengths of packets and messages, which control input is a control input parameter calculating unit.

Images