CN106992937A - Jamming control method based on GARCH time series algorithms - Google Patents

Abstract

The present invention relates to wireless communication field, to solve wireless communication congestion problems.The technical solution adopted by the present invention is, jamming control method based on GARCH time series algorithms, based on generilized auto regressive conditional heteroskedastic GARCH algorithms, the time interval of the correct order arrival received using transmitting terminal sets up GARCH (p, q) ARCH forecast model as a time series, the autoregressive exponent numbers of wherein p, q is the exponent number of ARCH, and the time delay that transmitting terminal sends packet is predicted, predicted value is obtained；Then with the size of packet than this upper predicted value, if actual bandwidth is more than this value, illustrate that transmission now is not in congestion phenomenon, otherwise, now congestion phenomenon is occurred in that in network, then be accomplished by being adjusted correspondingly the size of the data transmission window of transmitting terminal congestion.Present invention is mainly applied to wireless communication.

Description

Jamming control method based on GARCH time series algorithms

Technical field

The present invention relates to wireless communication field, the congestion problems in wireless network transmissions are particularly solved, specifically, It is related to the congestion control mechanism based on GARCH (generilized auto regressive conditional heteroskedastic) time series algorithm.

Background technology

Prevailing with wireless network, the emergence of internet, mobile terminal is even more innumerable, PC, cell phone, The terminal devices such as Mac are seen everywhere to the dependence more and more higher of network, therefore network congestion phenomenon.Network congestion how is solved to show As turning into the problem of puzzlement network provider always.Existing bandwidth how is made full use of on the premise of ensuring not increase bandwidth, It is the topic that each network provider is paid special attention to reduce network congestion, is also to bring more efficient beneficial development to company or enterprise Target.If increase bandwidth can bring more inputs in technology, research and development and equipment so that increase cost simply, and net Network utilization rate is also not necessarily because the increase of bandwidth increases.So, how to go to solve this network congestion phenomenon Traditional TCP (transmission control protocol) has less data loss rate in terms of cable network and higher transmission is stable Property, but be average in performance on a wireless network.Traditional TCP (transmission control protocol) is unable to the classification of data discrimination packet loss, Easily cause the judgement of mistake.Most of congestion avoidance algorithms realize that the control of such as PI (proportional plus integral control) rate is calculated in transmitting terminal Method, but the algorithm needs transmitting terminal and all intermediate nodes to make corresponding change, has broken conventional TCP (transmission control protocol) Semanteme, therefore be not widely used.

The content of the invention

To overcome the deficiencies in the prior art, the present invention is directed to propose a kind of be based on ARCH (ARCH) algorithm Congestion control mechanism, solve wireless communication congestion problems.The technical solution adopted by the present invention is, based on the GARCH times The jamming control method of sequence algorithm, based on generilized auto regressive conditional heteroskedastic GARCH algorithms, is received with transmitting terminal The time interval that correct order is reached is a time series, sets up GARCH (p, q) ARCH forecast model, its The middle autoregressive exponent numbers of p, q is the exponent number of ARCH, and the time delay that transmitting terminal sends packet is predicted, predicted Value；Then with the size of packet than this upper predicted value, this obtained ratio can be for the band in measurement now network Width, if actual bandwidth is more than this value, illustrates that transmission now is not in congestion phenomenon, otherwise, now occurs in network Congestion phenomenon, then be accomplished by being adjusted correspondingly the size of the data transmission window of transmitting terminal congestion, if a certain Moment, the time that the order that transmitting terminal is received correctly is responded has exceeded a certain preset value, i.e. response timeout, is now considered as packet Lose, it is necessary to lost data packets be resend, and reduce the size of congestion window, so that packet can be in existing network smoothly Transmit on ground.

Specific steps are further refined as：

The first step：It is determined that the time that correct command to answer correctly ACK instructions are reached, then obtains response according to this time Time interval sequence；

Second step：This time series produced according to the first step sets up ARCH (p, q), the autoregressive exponent numbers of p, and q is The exponent number of ARCH, and determine optimal p, q parameter value；

3rd step：The delay of next ACK times reached is entered using ARCH (p, the q) forecast model set up in second step Row prediction, and record actual value, is compared according to the ratio between the size of congestion window and this prediction lag value with current bandwidth, The speed that transmitting terminal sends packet is dynamically adjusted according to comparative result；

4th step：Processing to response timeout：When the response time correctly ordered exceeding a certain threshold value, now regard as Response timeout, can cause the loss of packet, and transmitting terminal receives and that packet of loss can be carried out again after timeout signal Send, and record transmission times, illustrate to occur in that congestion phenomenon in network when number of times is more than 3 times, solution is to reduce to gather around The size for filling in window is original half；Otherwise, according to the delay variance of calculating come the size of dynamic regulation congestion window；

5th step：Whether there is bandwidth available enough in detection network, if so, increase packet that then can be suitably Transmission rate makes full use of existing bandwidth.

The 3rd specific regulating strategy of step is, when the ratio between size and this prediction lag value of congestion window are less than current net In network during available bandwidth, illustrate there is bandwidth available enough in network, the transmission rate of increase packet that now can be suitably, When the ratio between the size of congestion window and this prediction lag value are more than current network available bandwidth, illustrate that network now occur gathers around Plug, reduces the size of transmitting terminal congestion window, and using the actual value of record as setting up GARCH (p, q) forecast model next time Data.

According to time series X_tGARCH (p, q) forecast model is set up, process is specifically set up as follows：

By X_tSequence is expressed as

X_t=β x_t+ε_t

Wherein ε_tFor noise, β is coefficient, meets ε_t~N (0,1) independent same distribution, and σ_tStandard deviation, meets

To ensure varianceTherefore there is factor alpha_i>=0, i=0,1 ..., p, and α₁+α₂+…+α_p<1；

If Ω_t-1=σ (X_t-1,X_t-2...) representIn t- The set of 1 moment all available information, wherein h_tFor conditional variance, α_iFor coefficient；

GARCH (p, q) model is expressed as follows：

The features of the present invention and beneficial effect are：

1st, the present invention such as can be because network congestion by ordering arrival time to identify the loss type of packet Caused is also due to that network linking is unstable caused, more accurately to implement regulation and control.

2nd, the present invention sends the transmission speed of end data according to the dynamic adjustment that predicts the outcome by real-time estimate network delay Rate, preferably to utilize existing bandwidth.

3rd, the present invention can be sent in the packet that automatically reduced when sending more than existing bandwidth bearing capacity of packet Speed is to avoid the generation of network congestion phenomenon.

4th, the present invention can improve the stability that packet is transmitted in a network.

Brief description of the drawings：

Fig. 1 accesses wireless network schematic for terminal device in the present invention；

The time series for the time interval that Fig. 2 reaches for correct order；

Fig. 3 is the present invention according to prediction lag dynamic regulation congestion window size flow chart.

Embodiment

In order to solve the above problems, the present invention is the congestion avoidance algorithm based on transmitting terminal.The algorithm is with GARCH (broad sense ARCH) based on algorithm, the time interval reached using the correct order that transmitting terminal is received is a time sequence Row, set up GARCH (p, q) (generilized auto regressive conditional heteroskedastic, the autoregressive exponent numbers of p, q is the exponent number of ARCH) autoregression bar Part Singular variance forecast model, is predicted to the time delay that transmitting terminal sends packet, obtains predicted value.Then packet is used Size than this upper predicted value, this obtained ratio can be for weighing the bandwidth now in network.If actual bandwidth is big In this value, then illustrate that transmission now is not in congestion phenomenon, otherwise, now occur in that congestion phenomenon in network, then It is accomplished by being adjusted correspondingly the size of the data transmission window (congestion window) of transmitting terminal.If a certain moment, send The time for holding the order that receives correctly to respond has exceeded a certain preset value, i.e. response timeout, be now considered as data-bag lost, it is necessary to Lost data packets are resend, and reduce the size of congestion window, so that packet can successfully be transmitted in existing network.

When network congestion causes data-bag lost, the handling capacity of network can promptly reduce, and two-way time can increase rapidly Greatly, a very big delay variance can now be brought, it is therefore desirable to reduce the size of congestion window to alleviate this congestion phenomenon.

When the loss of packet be due to wireless network link it is unstable caused by, and there is not congestion phenomenon in network, Bandwidth now is more stable, can bring a smaller delay variance, therefore need not adjust the size of congestion window.

When in certain time, when the network bandwidth can use enough, the mechanism is dynamically adjusted according to the delay of real-time estimate The transmission rate of packet, sufficiently to utilize the network bandwidth.

The present invention have devised a kind of congestion control mechanism for being based on ARCH (ARCH) algorithm, the mechanism Comprise the following steps：

The first step：It is determined that the time that correct ACK (command to answer correctly) instructions are reached, then obtains sound according to this time The time interval sequence answered, such as Fig. 2 is the time series schematic diagram for the time interval that right instructions are reached.

Second step：This time series produced according to the first step sets up ARCH (p, q), and (ARCH, p is certainly The exponent number of recurrence, q is the exponent number of ARCH), and determine optimal p, q parameter value.

3rd step：Using the ARCH (p, q) set up in second step, (q is for ARCH, p autoregressive exponent numbers The exponent number of ARCH) delay of the forecast model to next ACK (command to answer correctly) times reached is predicted, and record true Real value.Be compared according to the ratio between the size of congestion window and this prediction lag value with current bandwidth, according to comparative result come Dynamic adjustment transmitting terminal sends the speed of packet.Specific regulating strategy is, when size and this prediction of congestion window are prolonged When the ratio between value is less than available bandwidth in current network late, illustrate there is bandwidth available enough in network, increasing that now can be appropriate The transmission rate of big data bag.When the ratio between the size of congestion window and this prediction lag value are more than current network available bandwidth, Illustrate network congestion now occur, reduce the size of transmitting terminal congestion window.And using the actual value of record as setting up next time The data of GARCH (p, q) (generilized auto regressive conditional heteroskedastic, the autoregressive exponent numbers of p, q is the exponent number of ARCH) forecast model, Forecast model so can be dynamically set up, more accurate forecast model can be obtained.

4th step：Processing to response timeout.When the response time correctly ordered exceeding a certain threshold value, now regard as Response timeout, can cause the loss of packet.Transmitting terminal receives and that packet of loss can be carried out again after timeout signal Send, and record transmission times.Illustrate to occur in that congestion phenomenon in network when number of times is more than 3 times, solution is to reduce to gather around The size for filling in window is original half；Otherwise, according to the delay variance of calculating come the size of dynamic regulation congestion window.

5th step：Whether there is bandwidth available enough in detection network, if so, increase packet that then can be suitably Transmission rate makes full use of existing bandwidth.

Below in conjunction with accompanying drawing, embodiment of the present invention is described further in detail.

As Fig. 1 be terminal device link wireless network access point schematic diagram, when the quantity of mobile terminal is more and more, often The bandwidth of individual link can be reduced, if easily causing network congestion phenomenon if in the case that transmission packet rate is constant, Cause mass data packet loss.The transfer rate of packet can not only be influenceed but also the online experience of user can be influenceed.Need Take measures to solve this problem.

First, record data bag t transmitting terminal transmission time DSt and the reception time DRt of receiving terminal, then according to hair Send time DSt and receive the time interval sequence X t=DRt-DSt that time DRt obtains ACK, as shown in Figure 2.

Then this time series X obtained according to calculating_tSet up GARCH (p, q) (generilized auto regressive conditional heteroskedastic, p Autoregressive exponent number, q is the exponent number of ARCH) forecast model, specifically set up process as follows：

By X_tSequence is expressed as

By X_tSequence is expressed as

X_t=β x_t+ε_t

Wherein ε_tFor noise, β is coefficient, meets ε_t~N (0,1) independent same distribution, and σ_tStandard deviation, meets

To ensure varianceTherefore there is factor alpha_i>=0, i=0,1 ..., p, and α₁+α₂+…+α_p<1；

If Ω_t-1=σ (X_t-1,X_t-2...) representIn t- The set of 1 moment all available information, wherein h_tFor conditional variance, α_iFor coefficient；

GARCH (p, q) (generilized auto regressive conditional heteroskedastic, the autoregressive exponent numbers of p, q is the exponent number of ARCH) model table Show as follows：

Wherein, σ_t ²For variance, the autoregressive exponent numbers of p, q is the exponent number of ARCH, α_iAnd β_iFor coefficient, X_tFor time series.

ACK (command to answer correctly) is predicted according to GARCH forecast models established above, then dynamic adjustment is gathered around Fill in the size of window and then make full use of the network bandwidth.The detailed process for adjusting window is as follows：

Next ACK (command to answer correctly) is carried out first after a new ACK (command to answer correctly) is received Prediction, obtains prediction lag p_delay, then calculates available bandwidth, calculation formula by prediction lag p_delay For：

Bandwidth=packet_size/p_delay, wherein packet_size are the size of packet, Bandwidth is available bandwidth, and p_delay is prediction lag.If the band of this prediction is wider than actual bandwidth, now very Congestion phenomenon is occurred in that in possible network, transmitting terminal can reduce the size of congestion window；If the bandwidth of this prediction is less than real Border bandwidth, then illustrate there be available bandwidth in now network, and now transmitting terminal needs the transmission speed of appropriate increase packet Make full use of the network bandwidth.

Then, if the actual delay that receiving terminal is measured is more than a certain value of setting, now there is response timeout and show Caused by being due to network congestion as, the phenomenon, data-bag lost is caused, so now needing to resend packet, and is remembered Record repeats the number of times sent.If when packet is sent successfully, the number of times of transmission is not reaching to three times, then is only needed to congestion The size of window is reduced to 2, and enters next ACK (command to answer correctly) reception and prediction；If sent in packet During success, the number of times of transmission is more than or equal to three times, then needs the delay variance in calculating networkAnd the measurement write down now is prolonged Slow size m_delay.

Next need to be compared.Judge whetherAnd m_delay/p_delay>2, Wherein window_size is window size, and m_delay postpones for measurement,For variance.If a condition is unsatisfactory for, then say There is larger delay variance in bright network, now need to be reduced to congestion window size now into original half；If two Person meets, then illustrates do not occur congestion phenomenon in network, and the size to congestion window need not make any adjustments.It is whole dynamic The flow of state adjustment is as shown in Figure 3.

Claims (4)

1. a kind of jamming control method based on GARCH time series algorithms, it is characterized in that, with generilized auto regressive conditional heteroskedastic Based on GARCH algorithms, the time interval of the correct order arrival received using transmitting terminal sets up GARCH as a time series (p, q) ARCH forecast model, the autoregressive exponent numbers of wherein p, q is the exponent number of ARCH, and transmitting terminal is sent The time delay of packet is predicted, and obtains predicted value；Then with the size of packet than this upper predicted value, obtain this One ratio, if actual bandwidth is more than this value, can illustrate that transmission now will not for weighing the bandwidth in now network There is congestion phenomenon, otherwise, now occur in that congestion phenomenon in network, then be accomplished by sending window to the data of transmitting terminal congestion The size of mouth is adjusted correspondingly, if a certain moment, the time that the order that transmitting terminal is received correctly is responded has exceeded a certain Preset value, i.e. response timeout, are now considered as data-bag lost, it is necessary to resend lost data packets, and reduce congestion window Size, so that packet can successfully be transmitted in existing network.

2. the jamming control method as claimed in claim 1 based on GARCH time series algorithms, it is characterized in that, specific steps It is further refined as：

The first step：It is determined that the time that correct command to answer correctly ACK instructions are reached, then according to this time obtain response when Between intervening sequence；

Second step：This time series produced according to the first step sets up ARCH (p, q), and the autoregressive exponent numbers of p, q is ARCH Exponent number, and determine optimal p, q parameter value；

3rd step：The delay of next ACK times reached is carried out using ARCH (p, the q) forecast model set up in second step pre- Survey, and record actual value, be compared according to the ratio between the size of congestion window and this prediction lag value with current bandwidth, according to Comparative result come dynamically adjust transmitting terminal send packet speed；

4th step：Processing to response timeout：When the response time correctly ordered exceeding a certain threshold value, response is now regarded as Time-out, can cause the loss of packet, and transmitting terminal receives and that packet of loss can be sent again after timeout signal, And transmission times is recorded, illustrate to occur in that congestion phenomenon in network when number of times is more than 3 times, solution is to reduce congestion window Size be original half；Otherwise, according to the delay variance of calculating come the size of dynamic regulation congestion window；

5th step：Whether there is bandwidth available enough in detection network, if so, the transmission of increase packet that then can be suitably Speed makes full use of existing bandwidth.

3. the jamming control method as claimed in claim 2 based on GARCH time series algorithms, it is characterized in that, the 3rd step tool The regulating strategy of body is, when the ratio between size and this prediction lag value of congestion window are less than available bandwidth in current network, Illustrate there is bandwidth available enough in network, the transmission rate of increase packet that now can be suitably, the size of congestion window When being more than current network available bandwidth with the ratio between this prediction lag value, illustrate network congestion now occur, reduction transmitting terminal is gathered around The size of window is filled in, and regard the actual value of record as the data for setting up GARCH (p, q) forecast model next time.

4. the jamming control method as claimed in claim 1 based on GARCH time series algorithms, it is characterized in that, according to the time Sequence X_tGARCH (p, q) forecast model is set up, process is specifically set up as follows：

By X_tSequence is expressed as

X_t=β x_t+ε_t

Wherein ε_tFor noise, β is coefficient, meets ε_t~N (0,1) independent same distribution, and σ_tStandard deviation, meets

${\mathrm{\&sigma;}}_t^2={\mathrm{\&alpha;}}_0+\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{\&alpha;}}_i{X}_{t-i}^2$

To ensure varianceTherefore there is factor alpha_i>=0, i=0,1 ..., p, and α₁+α₂+…+α_p<1；

If Ω_t-1=σ (X_t-1,X_t-2...) representIn t-1 Carve the set of all available information, wherein h_tFor conditional variance, α_iFor coefficient；

GARCH (p, q) model is expressed as follows：

${\mathrm{\&sigma;}}_t^2={\mathrm{\&alpha;}}_0+\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{\&beta;}}_i{\mathrm{\&sigma;}}_{t-i}^2+\underset{j=1}{\overset{q}{\mathrm{\&Sigma;}}}{\mathrm{\&alpha;}}_j{X}_{t-j}^2.$