CN104320805A - Method for estimating link quality of wireless sensor network through a small amount of packets - Google Patents

Abstract

The invention discloses a method for estimating link quality of a wireless sensor network through a small amount of packets. The method comprises the following steps of receiving statistics on a signal to noise ratio (SNR), a link quality index (LQI) and a packet reception rate (PRR) of each packet; calculating regression parameters and selecting models; modifying an estimated link quality expression after estimating the link quality; giving a specific mode of execution. By means of the method, the packet reception rate in the future can be estimated according to parameters of the small amount of received packets under the environment of dynamic changes of the link quality. In addition, the method has the advantages of being capable of distinguishing different transitional links of the PRR, measuring the SNR and the LQI of the links more accurately and having self-repairing capability, and easy to implement.

Description

The method of radio sensing network link-quality is estimated by low volume data bag

Technical field

The present invention relates to a kind of method of signal to noise ratio according to low volume data bag, link-quality exponential sum packet reception rate estimation radio sensing network link-quality, belong to technology of wireless sensing network field.

Background technology

Under true environment, by freely declining, the impact of the factor such as multipath fading and shadow fading, the receives data packets rate of radio sensing network can be variant, generally be divided into stable connecting-type (receives data packets rate >=90%), transiens (10%≤receives data packets rate≤90%) and low acceptance rate type (acceptance rate is fixing lower) according to link-quality, the link-quality of a large amount of nodes belongs to transiens, its receives data packets rate (hereinafter referred to as PRR) 10% dynamic change between 90%.

PRR can the most directly react link-quality, and obtaining PRR needs to add up the transmitting-receiving situation of mass data bag, but transmitting-receiving mass data bag can cause a series of problem such as energy ezpenditure and communication delay.

Signal to noise ratio (hereinafter referred to as SNR) and link-quality index (hereinafter referred to as LQI) are two other indexs weighing link-quality, SNR can by the parameter obtained in radio frequency chip by necessarily calculating, and LQI directly can be obtained by radio frequency chip.For CC2420, this chip has a built-in received signal strength indicator register RSSI_VAL, read this value to deduct 45dB again and can obtain received signal power, received signal power time idle (not receiving packet) is exactly noise power, and received signal power when receiving packet deducts noise power can obtain SNR.The corresponding relation of monotone increasing is had between SNR and PRR, link-quality index (LQI) indicates the error rate of the packet received, CC2420 chip provides an average correlation value Corr, LQI can by (Corr a) b calculate that (a and b is constant, 0 between 255), usually, the LQI calculated 50 between 110, the monotonic increase relation between the average of LQI and PRR is better.

At present, existing employing SNR and LQI estimates that link-quality has following shortcoming: adopt single SNR to estimate link-quality, SNR 3.4 change between 7.3dB, can cause PRR 10% acute variation between 90%, therefore, this parameter Estimation link-quality of signal to noise ratio is only adopted to there is " resolution " too low (change that SNR is very little will cause PRR acute variation) of link-quality, be difficult to distinguish the problem of the different transiens link of PRR; Because CC2420 only records received signal strength and the LQI of the packet received, therefore, when adopting SNR, LQI or both Estimation of Mean link-qualities, SNR and LQI of record is more higher than physical link, thus cause SNR PRR curve and LQI PRR curve have deviation.

Summary of the invention

For the existing deficiency adopting PRR, SNR or LQI to estimate link-quality existence, the present invention proposes a kind of by PRR, SNR and PRR fusion, just can carry out the method for link quality estimation by means of only low volume data bag.

Estimated a method for radio sensing network link-quality by low volume data bag, step comprises: receive the statistics of packet SNR, LQI and PRR, regression parameter calculates and link-quality expression formula is estimated in Model Selection, link quality estimation and amendment;

Step 1): the statistics receiving packet SNR, LQI and PRR,

After the position of transmitter and receiver is fixing, transmitter sends StartNum StartTransmitMsg message, and the time interval of message is t ₁, StartNum>=100, comprise message SN in StartTransmitMsg message; Then, transmitter without time space to send m*w length be the packet of Length, m, w be respectively 10 20 natural number, Length be 20 30bytes, the time interval of packet is t ₂every m data bag is designated as a collection of, after receiver receives first StartTransmitMsg message, for the 1st batch data bag, according to the 1st, the time point of m data bag in time of the StartTransmitMsg message received and sequence number respectively first packet of sending of transmitter computes, by SNR before the 1st time point _totaland LQI _totalbe set to 0, the packet that receiver/transmitter sends between the 1st and the 2nd time point, the method calculating the signal to noise ratio of each bag is as follows: received signal power when receiver take 60ms as cycle timing sampling channel idle, as the sample value of a noise floor and the entry upgraded in noise floor table, by the mean value calculation noise floor now of 8 entries after every 8 samplings, the signal to noise ratio of the packet received deducts noise floor by the received signal power of this packet and obtains; The value of SNR and LQI of the packet received is added up, and statistics receives the number m ' of packet, at the 2nd time point, the accumulated value of SNR and LQI is designated as SNR respectively _totaland LQI _total, by the average signal-to-noise ratio SNR of following first packet of formulae discovery _avg1, first packet average link quality index LQI _avg1with the receives data packets rate PRR of first packet ₁:

${\mathrm{PRR}}_1=\frac{m^{\′}}{m}---\left(1\right)$

${\mathrm{SNR}}_{\mathrm{Avg}1}=\frac{{\mathrm{SNR}}_{\mathrm{total}1}}{m^{\′}}\·{\mathrm{PRR}}_1---\left(2\right)$

${\mathrm{LQI}}_{\mathrm{Avg}1}=\frac{{\mathrm{LQI}}_{\mathrm{total}1}}{m^{\′}}\·{\mathrm{PRR}}_1---\left(3\right)$

For transmitter sends the 2nd, 3 ... w batch data bag, receiver also carries out similar statistics and computing, obtains SNR respectively _avg2, LQI _avg2, PRR ₂..., SNR _avgw, LQI _avgw, PRR _w, remove the data point of PRR=1, after supposing removal, only have p group data point SNR _avg1, LQI _avg1, PRR ₁..., SNR _avgp, LQI _avgp, PRR _p;

Step 2): model parameter calculation and Model Selection,

2.1) model used

In link quality estimation, step 1) in the acceptance rate of m data bag of every a collection of PRR<1 and the SNR of packet _avgiand LQI _avgimeet one of following monotonically increasing model expression:

Model one,

${\mathrm{PRR}}_i=A_1\&CenterDot;{\mathrm{SNR}}_{\mathrm{Avgi}}^2+B_1\&CenterDot;{\mathrm{LQI}}_{\mathrm{Avgi}}^2---\left(4\right),$

Model two,

PRR _i＝A ₂·SNR _Avgi+B ₂·LQI _Avgi+C ₂ ⑸，

Wherein, i=1,2 ..., p, A ₁, B ₁, A ₂, B ₂, C ₂it is parameter;

2.2) model parameter calculation

Respectively by above-mentioned 2.1) in the expression formula of two kinds of models to p data point (SNR _avg1, LQI _avg1, PRR ₁) ..., (SNR _avgp, LQI _avgp, PRR _p) carrying out linear regression, circular is as follows:

The first model, adopts binary linear regression to above-mentioned p data point, obtains parameter A ₁and B ₁value computing formula as follows:

$A_1=\frac{(\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}^2\&CenterDot;{\mathrm{PRR}}_i)\left(\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{LQI}}_{\mathrm{Avgi}}^4\right)-(\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{LQI}}_{\mathrm{Avgi}}^2\&CenterDot;{\mathrm{PRR}}_i)(\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}^2\&CenterDot;{\mathrm{PRR}}_i)}{\left(\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}^4\right)\left(\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{LQI}}_{\mathrm{Avgi}}^4\right)-{(\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{LQI}}_{\mathrm{Avgi}}^2\&CenterDot;{\mathrm{SNR}}_{\mathrm{Avgi}}^2)}^2}---\left(6\right),$

$B_1=\frac{(\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}^2\&CenterDot;{\mathrm{PRR}}_i)(\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{LQI}}_{\mathrm{Avgi}}^2\&CenterDot;{\mathrm{SNR}}_{\mathrm{Avgi}}^2)-(\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{LQI}}_{\mathrm{Avgi}}^2\&CenterDot;{\mathrm{PRR}}_i)\left(\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}S{\mathrm{NR}}_{\mathrm{Avgi}}^4\right)}{{(\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{LQI}}_{\mathrm{Avgi}}^2\&CenterDot;{\mathrm{SNR}}_{\mathrm{Avgi}}^2)}^2-\left(\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}^4\right)\left(\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{LQI}}_{\mathrm{Avgi}}^4\right)}---\left(7\right),$

Obtain thus formula (5) in parameter A ₁, B ₁, the then expression formula of Confirming model one; Again by p data point to (SNR _avg1, LQI _avg1) ..., (SNR _avgp, LQI _avgp) bring the expression formula of model one into, p the receives data packets rate PRR calculated can be obtained _cal1, PRR _cal2..., PRR _calp, note then Pearson correlation coefficient R ₁obtained by following formula

$R_1=\frac{\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}({\mathrm{PRR}}_i-\stackrel{\&OverBar;}{\mathrm{PRR}})({\mathrm{PRR}}_{\mathrm{cali}}-\stackrel{\&OverBar;}{{\mathrm{PRR}}_{\mathrm{cal}}})}{\sqrt{\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}({\mathrm{PRR}}_i-\stackrel{\&OverBar;}{\mathrm{PRR}})}\&CenterDot;\sqrt{\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}({\mathrm{PRR}}_{\mathrm{cali}}-\stackrel{\&OverBar;}{{\mathrm{PRR}}_{\mathrm{cal}}})}}---\left(8\right),$

Second model, when p data point coincidence formula model (5), adopts the method for binary linear regression, obtains parameter A ₂, B ₂and C ₂value computing formula as follows:

$A_2=\frac{D_1}{D}---\left(9\right),$

$B_2=\frac{D_2}{D}---\left(10\right),$

$C_2=\frac{D_3}{D}---\left(11\right),$

Wherein D, D ₁, D ₂, D ₃computational methods as follows:

$D=\left|\begin{array}{ccc}\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}^2& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}\&CenterDot;{\mathrm{LQI}}_{\mathrm{Avgi}}& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}\\ \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}\&CenterDot;{\mathrm{LQI}}_{\mathrm{Avgi}}& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{LQI}}_{\mathrm{Avgi}}^2& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{LQI}}_{\mathrm{Avgi}}\\ \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{LQI}}_{\mathrm{Avgi}}& -2p\end{array}\right|---\left(12\right)$

$D_1=\left|\begin{array}{ccc}\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}\&CenterDot;{\mathrm{PRR}}_i& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}\&CenterDot;{\mathrm{LQI}}_{\mathrm{Avgi}}& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}\\ \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{LQI}}_{\mathrm{Avgi}}\&CenterDot;{\mathrm{PRR}}_i& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{LQI}}_{\mathrm{Avgi}}^2& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{LQI}}_{\mathrm{Avgi}}\\ \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{PRR}}_i& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{LQI}}_{\mathrm{Avgi}}& -2p\end{array}\right|---\left(13\right)$

$D_2=\left|\begin{array}{ccc}\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}^2& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}\&CenterDot;{\mathrm{PRR}}_i& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}\\ \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}\&CenterDot;{\mathrm{LQI}}_{\mathrm{Avgi}}& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{LQI}}_{\mathrm{Avgi}}\&CenterDot;{\mathrm{PRR}}_i& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{LQI}}_{\mathrm{Avgi}}\\ \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{PRR}}_i& -2p\end{array}\right|---\left(14\right)$

$D_3=\left|\begin{array}{ccc}\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}^2& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}\&CenterDot;{\mathrm{LQI}}_{\mathrm{Avgi}}& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}\&CenterDot;{\mathrm{PRR}}_i\\ \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}\&CenterDot;{\mathrm{LQI}}_{\mathrm{Avgi}}& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{LQI}}_{\mathrm{Avgi}}^2& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{LQI}}_{\mathrm{Avgi}}\&CenterDot;{\mathrm{PRR}}_i\\ \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{LQI}}_{\mathrm{Avgi}}& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{PRR}}_i\end{array}\right|---\left(15\right)$

Obtain thus formula (5) in parameter A ₂, B ₂, C ₂, and then the expression formula of Confirming model two, then by p data point to (SNR _avg1, LQI _avg1) ..., (SNR _avgp, LQI _avgp) bring the expression formula of model two into, obtain p the receives data packets rate PRR calculated _cal1, PRR _cal2..., PRR _calp, note then Pearson correlation coefficient R ₂can be obtained by following formula

$R_2=\frac{\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}({\mathrm{PRR}}_i-\stackrel{\&OverBar;}{\mathrm{PRR}})({\mathrm{PRR}}_{\mathrm{cali}}-\stackrel{\&OverBar;}{{\mathrm{PRR}}_{\mathrm{cal}}})}{\sqrt{\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}({\mathrm{PRR}}_i-\stackrel{\&OverBar;}{\mathrm{PRR}})}\&CenterDot;\sqrt{\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}({\mathrm{PRR}}_{\mathrm{cali}}-\stackrel{\&OverBar;}{{\mathrm{PRR}}_{\mathrm{cal}}})}}---\left(16\right),$

Step 2.3): Model Selection,

When only having one to be greater than 0.7 in R1 and R2, Pearson correlation coefficient is greater than the PRR change of the model calculation value energy accurate response reality of 0.7, estimates link-quality with this model expression;

When R1 and R2 is greater than 0.7, these two model calculation value can accurate response reality PRR change, therefore these two models all can be estimated link-quality;

When R1 and R2 is less than 0.7, these two model calculation value can not the PRR change of accurate response reality, therefore repeats the first step and second step, until select the model that Pearson correlation coefficient is greater than 0.7, estimates link-quality with this model expression;

Step 3): link quality estimation,

Transmitter send q length be the packet of Length, q be 10 20, SNR and LQI of the packet received adds up by receiver respectively, obtains SNR _sumand LQI _sum, go out SNR' by following formulae discovery _avgand LQI' _avg:

$P{\mathrm{RR}}^{\&prime;}=\frac{v}{q}---\left(17\right),$

${\mathrm{SNR}}_{\mathrm{Avg}}^{\&prime;}=\frac{{\mathrm{SNR}}_{\mathrm{sum}}}{v}\&CenterDot;{\mathrm{PRR}}^{\&prime;}---\left(18\right),$

${\mathrm{LQI}}_{\mathrm{Avg}}^{\&prime;}=\frac{{\mathrm{LQI}}_{\mathrm{sum}}}{v}\&CenterDot;{\mathrm{PRR}}^{\&prime;}---\left(19\right),$

By the SNR' calculated _avgand LQI' _avgsubstituting into step 2.3) expression formula of model selected can estimate the packet reception rate PRR of this link during this period of time _present;

Transmitter when and then to send length in 5 seconds be Length packet,

PRR _prediction＝PRR _present，

Step 4): after link quality estimation, link-quality expression formula is estimated in amendment,

In step 3) in, if it is the packet of Length that transmitter have sent q length for the 1st time, the PRR (20) calculated by expression formula _predictionbe designated as PRR _prediction1, and then have sent the packet that multiple length is Length in 5 seconds, adding up the receives data packets rate obtained is PRR ₁, the relative error of link quality estimation is

${\mathrm{\&delta;}}_1=\frac{{\mathrm{PRR}}_{\mathrm{prediction}1}-{\mathrm{PRR}}_1}{{\mathrm{PRR}}_1}\&times;100\%---\left(20\right),$

In like manner can obtain the 2nd, 3,4 ..., have sent packet that q length is Length for r time and after measured data packet reception rate, the relative error of link quality estimation is δ ₂, δ ₃, δ ₄..., δ _r, when the number of times that the value of relative error is greater than 10% has 3 times, just start to revise the process estimating link-quality expression formula, namely repeat step 1), step 2), step 3).

Described step 2.3): when R1 and R2 is greater than 0.7, preferably, with the model expression that Pearson correlation coefficient is larger, link-quality is estimated.

Beneficial effect: the present invention has compared as follows several a little based on SNR, LQI with the radio sensing network link quality estimation method of PRR with more existing:

By means of only low volume data bag SNR, LQI and PRR just can to link in the future PRR in a short time estimate (namely link-quality being estimated), could estimate link-quality without the need to the acceptance rate by adding up a large amount of packet;

(2), owing to adding the information of SNR and LQI simultaneously, so overcome the too low problem of the resolution that only has the link quality estimation of SNR information to exist, further, the transiens link that PRR is different can be distinguished;

(3) consider the factor of PRR, SNR and LQI overcoming the record existed when only adopting SNR, LQI or both Estimation of Mean link-qualities, than actual higher problem, more accurately can measure SNR and LQI of link;

(4) radio sensing network link quality estimation method of the present invention can switch between different link models, overcome fixing link quality estimation model and easily cause shortcoming compared with big error, and, after the situation that the link quality estimation value error obtained is larger exceedes certain number of times, can reselect and calculate link quality estimation model, namely radio sensing network link quality estimation method of the present invention has the ability of " self-regeneration ";

(5) each step of link quality estimation method of the present invention all can directly realize with computer program, and therefore link quality estimation method is easy to implement.

Embodiment

The method of estimation radio sensing network link-quality of the present invention comprises: receive the statistics of packet SNR, LQI and PRR, regression parameter calculates and several steps such as link-quality expression formula are estimated in Model Selection, link quality estimation and amendment.

The first step: the statistics receiving packet SNR, LQI and PRR.

After the position of transmitter and receiver is fixing, transmitter sends StartNum StartTransmitMsg message, and the time interval of message is t ₁(StartNum>=100, during to ensure that link-quality between transmitter and receiver is very poor, receiver also can receive this message), comprise message SN in StartTransmitMsg message.Then, transmitter sends packet that m*w length is Length (the desirable less value of m, the representative value that the representative value of m, w is respectively 20,10, Length is 30bytes) without time space, and the time interval of packet is t ₂every m data bag is designated as a collection of, after receiver receives first StartTransmitMsg message, for the 1st batch data bag, according to the 1st, the time point of m data bag in time of the StartTransmitMsg message received and sequence number respectively first packet of sending of transmitter computes, by SNR before the 1st time point _totaland LQI _totalbe set to 0, the packet that receiver/transmitter sends between the 1st and the 2nd time point, the method calculating the signal to noise ratio of each bag is as follows: received signal power when receiver take 60ms as cycle timing sampling channel idle, as the sample value of a noise floor and the entry upgraded in noise floor table, by the mean value calculation noise floor now (change of the more realistic noise of noise floor obtained like this of 8 entries after every 8 samplings, rapider to noisy response), the signal to noise ratio of the packet received deducts noise floor by the received signal power of this packet and obtains.The value of SNR and LQI of the packet received is added up, and statistics receives the number m ' of packet, at the 2nd time point, the accumulated value of SNR and LQI is designated as SNR respectively _totaland LQI _total, by following formulae discovery SNR _avg1, LQI _avg1and PRR ₁:

${\mathrm{PRR}}_1=\frac{m^{\&prime;}}{m}---\left(1\right)$

${\mathrm{SNR}}_{\mathrm{Avg}1}=\frac{{\mathrm{SNR}}_{\mathrm{total}1}}{m^{\&prime;}}\&CenterDot;{\mathrm{PRR}}_1---\left(2\right)$

${\mathrm{LQI}}_{\mathrm{Avg}1}=\frac{{\mathrm{LQI}}_{\mathrm{total}1}}{m^{\&prime;}}\&CenterDot;{\mathrm{PRR}}_1---\left(3\right)$

For transmitter sends the 2nd, 3 ... w batch data bag, receiver also carries out similar statistics and computing, obtains SNR respectively _avg2, LQI _avg2, PRR ₂..., SNR _avgw, LQI _avgw, PRR _wbecause PRR has saturation value 1, PRR and LQI and SNR has obvious positive correlation, under normal circumstances, as SNR>13 or LQI>90, PRR just reaches 1, SNR and LQI continues to increase, PRR is still 1 (saturation value), so will remove the data point of PRR=1 during model of fit, only has p group data point SNR after supposing removal _avg1, LQI _avg1, PRR ₁..., SNR _avgp, LQI _avgp, PRR _p.

Second step: model parameter calculation and Model Selection.

(1) the model used

In link quality estimation, the acceptance rate of m data bag of every a collection of PRR<1 and the SNR of packet in the first step _avgiand LQI _avgimeet one of following monotonically increasing model expression:

1. model one

${\mathrm{PRR}}_i=A_1\&CenterDot;{\mathrm{SNR}}_{\mathrm{Avgi}}^2+B_1\&CenterDot;{\mathrm{LQI}}_{\mathrm{Avgi}}^2---\left(4\right)$

2. model two

PRR _i＝A ₂·SNR _Avgi+B ₂·LQI _Avgi+C ₂ ⑸

Wherein, i=1,2 ..., p, A ₁, B ₁, A ₂, B ₂, C ₂it is parameter.

(2) model parameter calculation

Press the expression formula of two kinds of models respectively to p data point (SNR _avg1, LQI _avg1, PRR ₁) ..., (SNR _avgp, LQI _avgp, PRR _p) carry out linear regression, following computational methods can be obtained

1. the first model

To above-mentioned p data point adopt binary linear regression (as formula (4)), obtain parameter A ₁and B ₁value computing formula as follows:

$A_1=\frac{(\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}^2\&CenterDot;{\mathrm{PRR}}_i)\left(\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{LQI}}_{\mathrm{Avgi}}^4\right)-(\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{LQI}}_{\mathrm{Avgi}}^2\&CenterDot;{\mathrm{PRR}}_i)(\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}^2\&CenterDot;{\mathrm{PRR}}_i)}{\left(\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}^4\right)\left(\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{LQI}}_{\mathrm{Avgi}}^4\right)-{(\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{LQI}}_{\mathrm{Avgi}}^2\&CenterDot;{\mathrm{SNR}}_{\mathrm{Avgi}}^2)}^2}---\left(6\right)$

$B_1=\frac{(\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}^2\&CenterDot;{\mathrm{PRR}}_i)(\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{LQI}}_{\mathrm{Avgi}}^2\&CenterDot;{\mathrm{SNR}}_{\mathrm{Avgi}}^2)-(\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{LQI}}_{\mathrm{Avgi}}^2\&CenterDot;{\mathrm{PRR}}_i)\left(\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}S{\mathrm{NR}}_{\mathrm{Avgi}}^4\right)}{{(\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{LQI}}_{\mathrm{Avgi}}^2\&CenterDot;{\mathrm{SNR}}_{\mathrm{Avgi}}^2)}^2-\left(\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}^4\right)\left(\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{LQI}}_{\mathrm{Avgi}}^4\right)}---\left(7\right)$

Like this, can obtain formula (5) in parameter A ₁, B ₁, can the expression formula of Confirming model one.Again by p data point to (SNR _avg1, LQI _avg1) ..., (SNR _avgp, LQI _avgp) bring the expression formula of model one into, p the receives data packets rate PRR calculated can be obtained _cal1, PRR _cal2..., PRR _calp, note then Pearson correlation coefficient R ₁can be obtained by following formula

$R_1=\frac{\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}({\mathrm{PRR}}_i-\stackrel{\&OverBar;}{\mathrm{PRR}})({\mathrm{PRR}}_{\mathrm{cali}}-\stackrel{\&OverBar;}{{\mathrm{PRR}}_{\mathrm{cal}}})}{\sqrt{\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}({\mathrm{PRR}}_i-\stackrel{\&OverBar;}{\mathrm{PRR}})}\&CenterDot;\sqrt{\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}({\mathrm{PRR}}_{\mathrm{cali}}-\stackrel{\&OverBar;}{{\mathrm{PRR}}_{\mathrm{cal}}})}}---\left(8\right)$

2. second model

When p data point coincidence formula model (5), adopt the method for binary linear regression, obtain parameter A ₂, B ₂and C ₂value computing formula as follows:

$A_2=\frac{D_1}{D}---\left(9\right)$

$B_2=\frac{D_2}{D}---\left(10\right)$

$C_2=\frac{D_3}{D}---\left(11\right)$

Wherein D ₁, D ₂, D ₃computational methods as follows:

$D=\left|\begin{array}{ccc}\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}^2& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}\&CenterDot;{\mathrm{LQI}}_{\mathrm{Avgi}}& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}\\ \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}\&CenterDot;{\mathrm{LQI}}_{\mathrm{Avgi}}& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{LQI}}_{\mathrm{Avgi}}^2& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{LQI}}_{\mathrm{Avgi}}\\ \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{LQI}}_{\mathrm{Avgi}}& -2p\end{array}\right|---\left(12\right)$

$D_1=\left|\begin{array}{ccc}\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}\&CenterDot;{\mathrm{PRR}}_i& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}\&CenterDot;{\mathrm{LQI}}_{\mathrm{Avgi}}& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}\\ \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{LQI}}_{\mathrm{Avgi}}\&CenterDot;{\mathrm{PRR}}_i& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{LQI}}_{\mathrm{Avgi}}^2& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{LQI}}_{\mathrm{Avgi}}\\ \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{PRR}}_i& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{LQI}}_{\mathrm{Avgi}}& -2p\end{array}\right|---\left(13\right)$

$D_2=\left|\begin{array}{ccc}\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}^2& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}\&CenterDot;{\mathrm{PRR}}_i& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}\\ \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}\&CenterDot;{\mathrm{LQI}}_{\mathrm{Avgi}}& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{LQI}}_{\mathrm{Avgi}}\&CenterDot;{\mathrm{PRR}}_i& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{LQI}}_{\mathrm{Avgi}}\\ \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{PRR}}_i& -2p\end{array}\right|---\left(14\right)$

$D_3=\left|\begin{array}{ccc}\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}^2& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}\&CenterDot;{\mathrm{LQI}}_{\mathrm{Avgi}}& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}\&CenterDot;{\mathrm{PRR}}_i\\ \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}\&CenterDot;{\mathrm{LQI}}_{\mathrm{Avgi}}& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{LQI}}_{\mathrm{Avgi}}^2& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{LQI}}_{\mathrm{Avgi}}\&CenterDot;{\mathrm{PRR}}_i\\ \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{LQI}}_{\mathrm{Avgi}}& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{PRR}}_i\end{array}\right|---\left(15\right)$

So just can obtain formula (5) in parameter A ₂, B ₂, C ₂, and then can the expression formula of Confirming model two, then by p data point to (SNR _avg1, LQI _avg1) ..., (SNR _avgp, LQI _avgp) bring the expression formula of model two into, p the receives data packets rate PRR calculated can be obtained _cal1, PRR _cal2..., PRR _calp, note then Pearson correlation coefficient R ₂can be obtained by following formula

$R_2=\frac{\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}({\mathrm{PRR}}_i-\stackrel{\&OverBar;}{\mathrm{PRR}})({\mathrm{PRR}}_{\mathrm{cali}}-\stackrel{\&OverBar;}{{\mathrm{PRR}}_{\mathrm{cal}}})}{\sqrt{\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}({\mathrm{PRR}}_i-\stackrel{\&OverBar;}{\mathrm{PRR}})}\&CenterDot;\sqrt{\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}({\mathrm{PRR}}_{\mathrm{cali}}-\stackrel{\&OverBar;}{{\mathrm{PRR}}_{\mathrm{cal}}})}}---\left(16\right)$

(3) Model Selection

When there being one to be greater than 0.7 in R1 and R2, should can comparing the PRR change of accurate response reality by (Pearson correlation coefficient be greater than 0.7) model calculation value, therefore can estimate link-quality with this model expression;

When R1 and R2 is greater than 0.7, these two model calculation value can compare the PRR change of accurate response reality, therefore these two models all can be estimated link-quality, preferably, estimate link-quality with the model expression that Pearson correlation coefficient is larger;

When R1 and R2 is less than 0.7, these two model calculation value can not accurate response reality PRR change, therefore need the repetition first step and second step, until select the model that Pearson correlation coefficient is greater than 0.7, with this model expression, link-quality is estimated.

3rd step: link quality estimation.

Transmitter sends the packet (the desirable less value of q, representative value is 10) that q length is Length, SNR and LQI of the packet received adds up by receiver respectively, obtains SNR _sumand LQI _sum, go out SNR' by following formulae discovery _avgand LQI' _avg:

$P{\mathrm{RR}}^{\&prime;}=\frac{v}{q}---\left(17\right)$

${\mathrm{SNR}}_{\mathrm{Avg}}^{\&prime;}=\frac{{\mathrm{SNR}}_{\mathrm{sum}}}{v}\&CenterDot;{\mathrm{PRR}}^{\&prime;}---\left(18\right)$

${\mathrm{LQI}}_{\mathrm{Avg}}^{\&prime;}=\frac{{\mathrm{LQI}}_{\mathrm{sum}}}{v}\&CenterDot;{\mathrm{PRR}}^{\&prime;}---\left(19\right)$

By the SNR' calculated _avgand LQI' _avgthe expression formula of bringing the model selected in the Model Selection step of second step into can estimate the packet reception rate PRR of this link during this period of time _present.

Transmitter sends length within the back to back short time (representative value is 5 seconds) when being Length packet, because change in link quality is slower, so receiver a period of time in future receives data packets rate with during this period of time in the receives data packets rate of this link equal, therefore

PRR _prediction＝PRR _present

4th step: after link quality estimation, link-quality expression formula is estimated in amendment.

In the 4th step, if it is the packet of Length that transmitter have sent q length for the 1st time, the PRR (20) calculated by expression formula _predictionbe designated as PRR _prediction1, and then have sent the packet that several length are Length in the short time, adding up the receives data packets rate obtained is PRR ₁, the relative error of link quality estimation is

${\mathrm{\&delta;}}_1=\frac{{\mathrm{PRR}}_{\mathrm{prediction}1}-{\mathrm{PRR}}_1}{{\mathrm{PRR}}_1}\&times;100\%---\left(20\right)$

In like manner can obtain the 2nd, 3,4 ..., have sent packet that q length is Length for r time and after measured data packet reception rate, the relative error of link quality estimation is δ ₂, δ ₃, δ ₄..., δ _r, when the number of times that the value of relative error is greater than 10% has 3 times, just start to revise the process estimating link-quality expression formula, namely repeat the first step of the present invention, second step and the 3rd step.

Claims (2)

1. estimated the method for radio sensing network link-quality by low volume data bag for one kind, it is characterized in that, step comprises: receive the statistics of packet SNR, LQI and PRR, regression parameter calculates and link-quality expression formula is estimated in Model Selection, link quality estimation and amendment;

Step 1): the statistics receiving packet SNR, LQI and PRR,

After the position of transmitter and receiver is fixing, transmitter sends StartNum StartTransmitMsg message, and the time interval of message is t ₁, StartNum>=100, comprise message SN in StartTransmitMsg message; Then, transmitter without time space to send m*w length be the packet of Length, m, w be respectively 10 20 natural number, Length be 20 30bytes, the time interval of packet is t ₂every m data bag is designated as a collection of, after receiver receives first StartTransmitMsg message, for the 1st batch data bag, according to the 1st, the time point of m data bag in time of the StartTransmitMsg message received and sequence number respectively first packet of sending of transmitter computes, by SNR before the 1st time point _totaland LQI _totalbe set to 0, the packet that receiver/transmitter sends between the 1st and the 2nd time point, the method calculating the signal to noise ratio of each bag is as follows: received signal power when receiver take 60ms as cycle timing sampling channel idle, as the sample value of a noise floor and the entry upgraded in noise floor table, by the mean value calculation noise floor now of 8 entries after every 8 samplings, the signal to noise ratio of the packet received deducts noise floor by the received signal power of this packet and obtains; The value of SNR and LQI of the packet received is added up, and statistics receives the number m ' of packet, at the 2nd time point, the accumulated value of SNR and LQI is designated as SNR respectively _totaland LQI _total, by the average signal-to-noise ratio SNR of following first packet of formulae discovery _avg1, first packet average link quality index LQI _avg1with the receives data packets rate PRR of first packet ₁:

${\mathrm{PRR}}_1=\frac{m^{\&prime;}}{m}---\left(1\right),$

${\mathrm{SNR}}_{\mathrm{Avg}1}=\frac{{\mathrm{SNR}}_{\mathrm{total}1}}{m^{\&prime;}}\&CenterDot;{\mathrm{PRR}}_1---\left(2\right),$

${\mathrm{LQI}}_{\mathrm{Avg}1}=\frac{{\mathrm{LQI}}_{\mathrm{total}1}}{m^{\&prime;}}\&CenterDot;{\mathrm{PRR}}_1---\left(3\right),$

For transmitter sends the 2nd, 3 ... w batch data bag, receiver also carries out similar statistics and computing, obtains SNR respectively _avg2, LQI _avg2, PRR ₂..., SNR _avgw, LQI _avgw, PRR _w, remove the data point of PRR=1, after supposing removal, only have p group data point SNR _avg1, LQI _avg1, PRR ₁..., SNR _avgp, LQI _avgp, PRR _p;

Step 2): model parameter calculation and Model Selection,

2.1) model used

In link quality estimation, step 1) in the acceptance rate of m data bag of every a collection of PRR<1 and the SNR of packet _avgiand LQI _avgimeet one of following monotonically increasing model expression:

Model one,

${\mathrm{PRR}}_i=A_1\&CenterDot;{\mathrm{SNR}}_{\mathrm{Avgi}}^2+B_1\&CenterDot;{\mathrm{LQI}}_{\mathrm{Avgi}}^2---\left(4\right),$

Model two,

PRR _i＝A ₂·SNR _Avgi+B ₂·LQI _Avgi+C ₂ ⑸，

Wherein, i=1,2 ..., p, A ₁, B ₁, A ₂, B ₂, C ₂it is parameter;

2.2) model parameter calculation

Respectively by above-mentioned 2.1) in the expression formula of two kinds of models to p data point (SNR _avg1, LQI _avg1, PRR ₁) ..., (SNR _avgp, LQI _avgp, PRR _p) carrying out linear regression, circular is as follows:

The first model,

Binary linear regression is adopted to above-mentioned p data point, obtains parameter A ₁and B ₁value computing formula as follows:

$A_1=\frac{(\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}^2\&CenterDot;{\mathrm{PRR}}_i)\left(\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{LQI}}_{\mathrm{Avgi}}^4\right)-(\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{LQI}}_{\mathrm{Avgi}}^2\&CenterDot;{\mathrm{PRR}}_i)(\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNRR}}_{\mathrm{Avgi}}^2\&CenterDot;{\mathrm{PRR}}_i)}{\left(\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}^4\right)\left(\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{LQI}}_{\mathrm{Avgi}}^4\right)-{(\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{LQI}}_{\mathrm{Avgi}}^2\&CenterDot;{\mathrm{SNR}}_{\mathrm{Avgi}}^2)}^2}---\left(6\right),$

$B_1=\frac{(\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}^2\&CenterDot;{\mathrm{PRR}}_i)(\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{LQI}}_{\mathrm{Avgi}}^2\&CenterDot;{\mathrm{SNR}}_{\mathrm{Avgi}}^2)-(\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{LQI}}_{\mathrm{Avgi}}^2\&CenterDot;{\mathrm{PRR}}_i)(\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNRR}}_{\mathrm{Avgi}}^4)}{{(\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{LQI}}_{\mathrm{Avgi}}^2\&CenterDot;{\mathrm{SNR}}_{\mathrm{Avgi}}^2)}^2\left(\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}^4\right)\left(\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{LQI}}_{\mathrm{Avgi}}^4\right)}---\left(6\right),$

Obtain thus formula (5) in parameter A ₁, B ₁, the then expression formula of Confirming model one; Again by p data point to (SNR _avg1, LQI _avg1) ..., (SNR _avgp, LQI _avgp) bring the expression formula of model one into, p the receives data packets rate PRR calculated can be obtained _cal1, PRR _cal2..., PRR _calp, note then Pearson correlation coefficient R ₁obtained by following formula

$R_1=\frac{\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}({\mathrm{PRR}}_i-\stackrel{\&OverBar;}{\mathrm{PRR}})({\mathrm{PRR}}_{\mathrm{cali}}-\stackrel{\&OverBar;}{{\mathrm{PRR}}_{\mathrm{cal}}})}{\sqrt{\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}({\mathrm{PRR}}_i-\stackrel{\&OverBar;}{\mathrm{PRR}})}\&CenterDot;\sqrt{\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}({\mathrm{PRR}}_{\mathrm{cali}}-\stackrel{\&OverBar;}{{\mathrm{PRR}}_{\mathrm{cal}}})}}---\left(8\right),$

Second model,

When p data point coincidence formula model (5), adopt the method for binary linear regression, obtain parameter A ₂, B ₂and C ₂value computing formula as follows:

$A_2=\frac{D_1}{D}---\left(9\right),$

$B_2=\frac{D_2}{D}---\left(10\right),$

$C_2=\frac{D_3}{D}---\left(11\right),$

Wherein D, D ₁, D ₂, D ₃computational methods as follows:

$D=\left|\begin{array}{ccc}\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}^2& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}\&CenterDot;{\mathrm{LQI}}_{\mathrm{Avgi}}& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}\\ \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}\&CenterDot;{\mathrm{LQI}}_{\mathrm{Avgi}}& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{LQI}}_{\mathrm{Avgi}}^2& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{LQI}}_{\mathrm{Avgi}}\\ \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{LQI}}_{\mathrm{Avgi}}& -2p\end{array}\right|---\left(12\right)$

$D_1=\left|\begin{array}{ccc}\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}\&CenterDot;{\mathrm{PRR}}_i& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}\&CenterDot;{\mathrm{LQI}}_{\mathrm{Avgi}}& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}\\ \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{LQI}}_{\mathrm{Avgi}}\&CenterDot;{\mathrm{PRR}}_i& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{LQI}}_{\mathrm{Avgi}}^2& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{LQI}}_{\mathrm{Avgi}}\\ \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{PRR}}_i& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{LQI}}_{\mathrm{Avgi}}& -2p\end{array}\right|---\left(13\right)$

$D_2=\left|\begin{array}{ccc}\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}^2& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}\&CenterDot;{\mathrm{PRR}}_i& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}\\ \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}\&CenterDot;{\mathrm{LQI}}_{\mathrm{Avgi}}& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{LQI}}_{\mathrm{Avgi}}\&CenterDot;{\mathrm{PRR}}_i& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{LQI}}_{\mathrm{Avgi}}\\ \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{PRR}}_i& -2p\end{array}\right|---\left(14\right)$

$D_3=\left|\begin{array}{ccc}\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}^2& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}\&CenterDot;{\mathrm{LQI}}_{\mathrm{Avgi}}& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}\&CenterDot;{\mathrm{PRR}}_i\\ \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}\&CenterDot;{\mathrm{LQI}}_{\mathrm{Avgi}}& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{LQI}}_{\mathrm{Avgi}}^2& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{LQI}}_{\mathrm{Avgi}}\&CenterDot;{\mathrm{PRR}}_i\\ \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{SNR}}_{\mathrm{Avgi}}& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{LQI}}_{\mathrm{Avgi}}& \underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{PRR}}_i\end{array}\right|---\left(15\right)$

Obtain thus formula (5) in parameter A ₂, B ₂, C ₂, and then the expression formula of Confirming model two, then by p data point to (SNR _avg1, LQI _avg1) ..., (SNR _avgp, LQI _avgp) bring the expression formula of model two into, obtain p the receives data packets rate PRR calculated _cal1, PRR _cal2..., PRR _calp, note then Pearson correlation coefficient R ₂can be obtained by following formula

$R_2=\frac{\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}({\mathrm{PRR}}_i-\stackrel{\&OverBar;}{\mathrm{PRR}})({\mathrm{PRR}}_{\mathrm{cali}}-\stackrel{\&OverBar;}{{\mathrm{PRR}}_{\mathrm{cal}}})}{\sqrt{\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}({\mathrm{PRR}}_i-\stackrel{\&OverBar;}{\mathrm{PRR}})}\&CenterDot;\sqrt{\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}({\mathrm{PRR}}_{\mathrm{cali}}-\stackrel{\&OverBar;}{{\mathrm{PRR}}_{\mathrm{cal}}})}}---\left(16\right),$

Step 2.3): Model Selection,

When only having one to be greater than 0.7 in R1 and R2, Pearson correlation coefficient is greater than the PRR change of the model calculation value energy accurate response reality of 0.7, estimates link-quality with this model expression;

When R1 and R2 is greater than 0.7, these two model calculation value can accurate response reality PRR change, therefore these two models all can be estimated link-quality;

When R1 and R2 is less than 0.7, these two model calculation value can not the PRR change of accurate response reality, therefore repeats the first step and second step, until select the model that Pearson correlation coefficient is greater than 0.7, estimates link-quality with this model expression;

Step 3): link quality estimation,

Transmitter sends the packet that q length is Le ngth, q be 10 20, SNR and LQI of the packet received adds up by receiver respectively, obtains SNR _sumand LQI _sum, go out SNR' by following formulae discovery _avgand LQI' _avg:

${\mathrm{PRR}}^{\&prime;}=\frac{v}{q}---\left(17\right),$

${\mathrm{SNR}}_{\mathrm{Avg}}^{\&prime;}=\frac{{\mathrm{SNR}}_{\mathrm{sum}}}{v}\&CenterDot;{\mathrm{PRR}}^{\&prime;}---\left(18\right),$

${\mathrm{LQI}}_{\mathrm{Avg}}^{\&prime;}=\frac{{\mathrm{LQI}}_{\mathrm{sum}}}{v}\&CenterDot;{\mathrm{PRR}}^{\&prime;}---\left(19\right),$

By the SNR' calculated _avgand LQI' _avgsubstituting into step 2.3) expression formula of model selected can estimate the packet reception rate PRR of this link during this period of time _present;

Transmitter when and then to send length in 5 seconds be Length packet,

PRR _prediction＝PRR _present，

Step 4): after link quality estimation, link-quality expression formula is estimated in amendment,

In step 3) in, if it is the packet of Length that transmitter have sent q length for the 1st time, the PRR (20) calculated by expression formula _predictionbe designated as PRR _prediction1, and then have sent the packet that multiple length is Length in 5 seconds, adding up the receives data packets rate obtained is PRR ₁, the relative error of link quality estimation is

${\mathrm{\&delta;}}_1=\frac{{\mathrm{PRR}}_{\mathrm{predictioon}1}-{\mathrm{PRR}}_1}{{\mathrm{PRR}}_1}\&times;100\%---\left(20\right)$

In like manner can obtain the 2nd, 3,4 ..., have sent packet that q length is Length for r time and after measured data packet reception rate, the relative error of link quality estimation is δ ₂, δ ₃, δ ₄..., δ _r, when the number of times that the value of relative error is greater than 10% has 3 times, just start to revise the process estimating link-quality expression formula, namely repeat step 1), step 2), step 3).

2. method according to claim 1, is characterized in that, described step 2.3): when R1 and R2 is greater than 0.7, preferably, with the model expression that Pearson correlation coefficient is larger, link-quality is estimated.