CN104579569A - Full-duplex wireless network data transmission method based on mode and rate adaptive selection - Google Patents

Abstract

The invention discloses a full-duplex wireless network data transmission method based on mode and rate adaptive selection. The method is characterized by comprising the steps of determining an optimal combination comprising modulation and encoding schemes, multi-input multi-output modes, and duplex modes before data of a next frame are transmitted, and transmitting the data of the next frame according to parameters determined by the optimal combination. The method is small in searching space, short in algorithm convergence time, high in spectrum effectiveness and large in system capacity.

Description

Based on the full duplex radio network data transmission method that pattern and rate adaptation are selected

Technical field

The invention belongs to wireless network data transmission technique field, particularly a kind of full duplex radio network data transmission method selected based on pattern and rate adaptation.

Background technology

Traditional full duplex receiver is virtual full-duplex communication, namely must send and Received signal strength in different time or frequency.Therefore, frequency spectrum does not make full use of, and communicates if realize transmitting-receiving bidirectional on identical time with frequency simultaneously, namely with frequently simultaneously full-duplex communication, then and can by spectrum efficiency and power system capacity be maximum doubles.

Current, just started starting with the research of full duplex radio network simultaneously frequently, the achievement in research published is considerably less.In July, 2013, primary study project that U.S.'s Natural Science Fund set up " Foudations ofHierarchical Full-Duplex WirelessNetworks ", this is first full duplex radio network research problem setting up of U.S.'s Natural Science Fund up to now, problem member is by Ohio State University, related researcher's composition of University of California in Los Angeles and Rice University, the Sabharwal professor of the leader Rice University of problem has made lot of research in Full-duplex wireless communications field, achieve the full duplex radio network based on Wi-Fi communication system.In addition, the research team of Sabharwal professor, according to the feature of Wi-Fi network, research achieves a kind of new full duplex radio Medium Access Control Protocols: FD-MAC.FD-MAC solves the compatibility issue with conventional half duplex terminal, improves MAC protocol, the chance of Full-duplex wireless communications in maximization network.Full-duplex wireless communications can make single-channel message capacity double, and Full-duplex wireless communications is applied in wireless network, and the throughput of network but not necessarily can be made double.Xinyu professor Zhang in University of Wisconsin-Madison pungent university wheat enlightening grandson branch school has carried out the theory analysis of wireless network capacitance to full duplex Wi-Fi network while of frequently, result of study shows that full duplex radio network capacity is even lower than conventional half duplex network capacity if can not design effective full duplex MAC protocol.This result of study also shows further, by the application of Full-duplex wireless communications technology in wireless network, from the angle of network system, must will apply in a flexible way with Full-duplex wireless communications technology while of frequently.

Rate adaptation is the important transmission optimization mechanism of wireless network link layer, its according to time the channel status that becomes be that each message to be sent selects optimum physical layer transmission rate, i.e. modulation and encoding scheme (modulation and codingscheme, be called for short MCS), and then optimized transmission performance.At present, rate adaptation mechanism mainly can be divided into two classes: closed loop and open loop.In closed loop rate adaptation mechanism, its receive channel quality or optimum reception speed explicitly are fed back to transmit leg by recipient, and transmit leg is according to the feedback dynamic conditioning transmission rate received; In open-loop rate adaptation mechanism, transmit leg estimates current best transmission rate according to the channel quality of detection or historical statistical information.Closed loop rate adaptation mechanism needs receiving-transmitting sides close cooperation, is difficult to compatible between the Wireless Communication Equipment of different production firm each other.Therefore in actual use, what extensively adopt is all open-loop rate adaptation mechanism.

The adaptive problem of combining of speed and pattern (dual-mode and MIMO mode) is one and typically explores and Utilizing question.Wireless channel has the time-varying characteristics of height, does not have stable probability statistics feature under normal conditions, is not therefore suitable for classical static Multi-Armed Bandit learning method.

Therefore, prior art Problems existing is: because message transmission rate, MIMO mode and dual-mode are combined, adaptively selected solution search volume is large, the algorithmic statement time is long, causes that spectrum efficiency is low and power system capacity is little.

Summary of the invention

The object of the present invention is to provide a kind of full duplex radio network data transmission method selected based on pattern and rate adaptation, effectively promote frequency spectrum resource utilization rate, increase network system capacity.

The technical solution realizing the object of the invention is: a kind of full duplex radio network data transmission method selected based on pattern and rate adaptation, it is characterized in that, before next frame transfer of data, determine to comprise the best of breed modulated with encoding scheme, MIMO mode, dual-mode, the parameter that described next frame data are determined according to described best of breed is transmitted.

The present invention compared with prior art, its remarkable advantage:

Result of study (the Journal of Software in early stage, 2015,26 (1): 98-108) show, the adaptively selected problem of IEEE 802.11n medium-rate, MIMO mode and channel width can be solved based on non-static Multi-Armed Bandit learning method well.On the basis of this achievement in research, first the inventive method devises the reward function of MCS, MIMO mode and dual-mode combination, and devise MCS, MIMO mode and dual-mode prediction algorithm respectively based on post-class processing, solve speed, MIMO mode and dual-mode further and combine the problem that adaptively selected solution space is comparatively large, convergence rate is slower, improve practicality and the feasibility of algorithm.

Below in conjunction with the drawings and specific embodiments, the present invention is described in further detail.

Accompanying drawing explanation

Fig. 1 is the full duplex radio network data transmission method fundamental diagram that the pattern that the present invention is based on and rate adaptation are selected.

Embodiment

As shown in Figure 1, the present invention is based on the full duplex radio network data transmission method that pattern and rate adaptation are selected, before next frame transfer of data, determine to comprise modulation and encoding scheme (modulation and coding scheme, be called for short MCS), MIMO mode (Multiple Input Multiple Output, be called for short MIMO), the best of breed of dual-mode, the parameter that described next frame data are determined according to described best of breed is transmitted.

Describedly determine that best of breed step comprises:

(10) frame transmission result in acquisition: obtain the transmission result of previous frame data under its running parameter;

Before transmission first frame data, need to system initialization: namely for all m, s, w; , setup parameter μ _{m, s, w}=0, f _{m, s, w}=0, θ _{m, s, w}=0, ρ _{m, s, w}=0.

In described (10) acquisition, frame transmission result comprises: send totalframes f _{m, s, w}, message success sending probability in frame

ρ _{m, s, w}successfully receive message total θ _{m, s, w}.Wherein, f _{m, s, w}represent under MCS m, MIMO mode s and mode of operation w condition, send totalframes, ρ _{m, s, w}represent message success sending probability in frame, θ _{m, s, w}represent and successfully receive message total, MCS is modulation and encoding scheme, and MIMO is multiple-input, multiple-output, and M represents whole available MCS sets { M|m ₁< ... < m ₁, m is wherein element, and S represents MIMO mode set { empty point of many collection, space division multiplexing }, and s is wherein element, and W represents dual-mode set { half-duplex, full duplex }, and w is wherein element.

(20) the lower frame transformation parameter of prediction: according to upper frame transmission result, prediction next frame data transmission parameters;

Described (20) predict that this frame transformation parameter step comprises:

(21) MIMO mode and dual-mode is predicted: according to previous frame data at the message success sending probability of different MIMO mode and dual-mode with successfully receive message total, the MIMO mode adopted when predicting next frame transfer of data and dual-mode;

Described (21) prediction MIMO mode and dual-mode step comprise:

(211) θ is calculated _{s, w}and ρ _{s, w}:

${\mathrm{\&theta;}}_{s,w}=\underset{m=1}{\overset{R}{\mathrm{\&Sigma;}}}{\mathrm{\&theta;}}_{m,s,w}/R$

${\mathrm{\&rho;}}_{s,w}=\underset{m=1}{\overset{R}{\mathrm{\&Sigma;}}}{\mathrm{\&rho;}}_{m,s,w}/R,$

In formula, θ _{s, w}for the average successfully reception message number under MIMO mode s and dual-mode w, ρ _{s, w}for the average message success sending probability under MIMO mode s and dual-mode w, R is total MCS quantity.

(212) by θ _{s, w}and ρ _{s, w}the MIMO mode of classical CART method structure and the post-class processing of dual-mode carry out classifying and identifying, draws best MIMO mode s* and the dual-mode w* of prediction;

(213) at s ^*and w ^*the frame number of lower transmission and successfully send message number and adjust, that is:

$f_{{m,s}^{*},w}=f_{{m,s}^{*},w}+1,{\mathrm{\&theta;}}_{{m,s}^{*},w}={\mathrm{\&theta;}}_{{m,s}^{*},w}+{\mathrm{\&rho;}}_{{m,s}^{*},w}\&times;N_{{m,s}^{*},w}$

$f_{m,s,w^{*}}=f_{m,s,w^{*}}+1,{\mathrm{\&theta;}}_{m,s,w^{*}}={\mathrm{\&theta;}}_{m,s,w^{*}}+{\mathrm{\&rho;}}_{m,s,w^{*}}\&times;N_{m,s,w^{*}}$

In formula, f _{m, s*, w}, ρ _{m, s*, w}, θ _{m, s*, w}and N _{m, s*, w}then be illustrated respectively in the message number sent in the successful sending probability of message in the transmission totalframes under MIMO mode s*, frame and every frame, and f _{m, s, w*}, ρ _{m, s, w*}, θ _{m, s, w*}and N _{m, s, w*}then be illustrated respectively in the message number sent in the successful sending probability of message in the transmission totalframes under MIMO mode w*, frame and every frame

The set sizes of MIMO mode and channel width is 2, and search volume is less, and therefore, first the present invention completes the prediction to MIMO mode and dual-mode.Adopt in document [15] the classical CART method proposed to build the post-class processing of MIMO mode and dual-mode, its attribute is < m, w, s, θ _{m, s, w}, ρ _{m, s, w}>.This predicted method is according to the average theta of different MIMO mode and dual-mode _{s, w}and ρ _{s, w}, the MIMO mode adopted during the transmission of prediction next frame and dual-mode.

(22) predict MCS: the MIMO mode predicted according to step (21) and dual-mode, and sent the successful reception message total of Frame and the successful probability of acceptance, the best modulation of prediction and encoding scheme MCS;

Described (22) prediction MCS step comprises:

(221) computational prediction with that is:

${\mathrm{\&rho;}}_{m,s*,w*}\left(i\right)={\mathrm{\&sigma;\&rho;}}_{m,s*,w*}(i-1)+(1-\mathrm{\&sigma;})\frac{\underset{k=2}{\overset{\mathrm{win}}{\mathrm{\&Sigma;}}}{\mathrm{\&rho;}}_{m,s*,w*}(i-k)}{\mathrm{win}}$

θ _m，s*，w*(i)＝θ _m，s*，w*(i-1)+ρ _m，s*，w*(i)×N _m，s*，w*，

In formula, win is moving average window size; σ is smoothing factor, and σ value is larger, and recent obtained θ and ρ impact when predicting MCS is larger;

(222) by θ _{m, s, w}(i) and ρ _{m, s, w}i () carries out classifying and identifying on the MCS post-class processing built, draw the best MCS m* of prediction;

(223) to m ^*the frame number of lower transmission adjusts, that is:

$f_{m^{*},s^{*},w^{*}}=f_{m^{*},s^{*},w^{*}}+1.$

(23) predict this frame transformation parameter: the best tlv triple comprehensively drawing prediction, upgrade the remuneration of the best tlv triple of prediction.

Described (23) predict that this frame transformation parameter step comprises:

(231) comprehensively draw the best tlv triple < m* of prediction, s*, w* >, and upgrade tlv triple < m* according to following formula, s*, w* >'s

θ _m*，s*，w*＝γ×(θ _m*，s*，w*+ρ _m*，s*，w*(i)×N _m*，s*，w*)

(232) will with substitute into remuneration computing function and upgrade the best tlv triple < m* of prediction, the remuneration of s*, w* >.

(30) prediction lower frame transmission result: according to next frame transformation parameter, prediction next frame transfer of data result;

The lower frame transmission result step of described (30) prediction is specially:

θ _m*，s*，w*＝γ×(θ _m*，s*，w*+ρ _m*，s*，w*(i)×N _m*，s*，w*)，

${\mathrm{\&rho;}}_{m,s*,w*}\left(i\right)={\mathrm{\&sigma;\&rho;}}_{m,s*,w*}(i-1)+(1-\mathrm{\&sigma;})\frac{\underset{k=2}{\overset{\mathrm{win}}{\mathrm{\&Sigma;}}}{\mathrm{\&rho;}}_{m,s*,w*}(i-k)}{\mathrm{win}}$

θ _m，s*，w*(i)＝θ _m，s*，w*(i-1)+ρ _m，s*，w*(i)×N _m，s*，w*，

${\mathrm{\&theta;}}_{s,w}=\underset{m=1}{\overset{R}{\mathrm{\&Sigma;}}}{\mathrm{\&theta;}}_{m,s,w}/R$

${\mathrm{\&rho;}}_{s,w}=\underset{m=1}{\overset{R}{\mathrm{\&Sigma;}}}{\mathrm{\&rho;}}_{m,s,w}/R,$

In formula, ρ _{m*, s*, w*}i () is that the i-th frame transmission adopts tlv triple < m ^*, s ^*, w ^*the prediction message success sending probability of >, γ is discount factor, and γ < 1, R is total MCS quantity, N _{m, s, w}then represent the message amount that this frame can send in time T, θ _{s, w}successfully message number is received, ρ for average _{s, w}for average message success sending probability.

(40) best tlv triple is determined: according to next frame transformation parameter and the transfer of data result of prediction, calculate the reward function of the tlv triple be made up of modulation and encoding scheme, MIMO mode, dual-mode, the tlv triple selecting remuneration the highest combining from all modulation and encoding scheme, MIMO mode, dual-mode is best of breed, and its speed determined and mode of operation are combined as the parameter of next frame transfer of data.

Described (40) determine that best tlv triple step is specially:

(41) reward function is calculated as follows:

${\mathrm{\&mu;}}_{m,s,w}=\frac{{\mathrm{\&theta;}}_{m,s,w}}{f_{m,s,w}}+B\sqrt{\frac{\mathrm{\&alpha;}\log ({\mathrm{\&Sigma;}}_{m^{\&prime;}\&Element;M,s^{\&prime;}\&Element;S,w^{\&prime;}\&Element;W}{f}_{m^{\&prime;},s^{\&prime;},w^{\&prime;}}\&times;N_{m^{\&prime;},s^{\&prime;},w^{\&prime;}})}{f_{m,s,w}\&times;N_{m,s,w}}}$

Reward function represented by above-mentioned formula utilizes factor two parts to form by exploring Summing Factor.θ _{m, s, w}/ f _{m, s, w}represent up to the present, under MCSm, MIMO mode s and dual-mode w condition, the average throughput of acquisition.Log (∑ _{m ' ∈ M, s ' ∈ S, w ' ∈ W}f _{m ', s ', w '}× N _{m ', s ', w '})/(f _{m, s, w}× N _{m, s, w}) reflect be current tlv triple utilize the factor, numerical value is larger, and throughput is higher, and transmit leg more tends to select this tlv triple.

(42) tlv triple the highest for remuneration is defined as best tlv triple, that is:

〈m，s，w〉＝argmaxμ _m，s，w(i)，

This formula represents that the tlv triple selecting remuneration the highest from all triple combination is best tlv triple.Before next frame transfer of data, determine to comprise the best of breed modulated with encoding scheme, MIMO mode, dual-mode, the parameter that described next frame data are determined according to described best of breed is transmitted, as described in above formula.

The inventive method devises a kind of reward function, and is that each tlv triple < m, s, w > distribute a remuneration according to this function.When transmitting data frame, the tlv triple that the inventive method selects remuneration the highest, it represents best MCS, MIMO mode and dual-mode combination known at present;

For shortening convergence time, the post-class processing CART method respectively based on classics builds MIMO mode and the post-class processing of dual-mode and the post-class processing of MCS; The inventive method first dopes best MIMO mode and dual-mode combines, on basis, dope best MCS again; Combine according to the best MCS predicted, MIMO mode and dual-mode, then by reward function, calculate the remuneration of this tlv triple, and select this MCS, MIMO mode and dual-mode to combine.The parameter that next frame data are then determined according to described best of breed is transmitted.

The inventive method solves speed, MIMO mode and dual-mode and combines the problem that adaptively selected solution space is comparatively large, convergence rate is slower, improves practicality and the feasibility of algorithm.Effectively can promote frequency spectrum resource utilization rate, increase network system capacity.

Claims (9)

1. the full duplex radio network data transmission method selected based on pattern and rate adaptation, it is characterized in that, before next frame transfer of data, determine to comprise the best of breed modulated with encoding scheme, MIMO mode, dual-mode, the parameter that described next frame data are determined according to described best of breed is transmitted.

2. wireless network data transmission method according to claim 1, is characterized in that, describedly determines that best of breed step comprises:

(10) frame transmission result in acquisition: obtain the transmission result of previous frame data under its running parameter;

(20) the lower frame transformation parameter of prediction: according to upper frame transmission result, prediction next frame data transmission parameters;

(30) prediction lower frame transmission result: according to lower frame transformation parameter, prediction next frame transfer of data result;

(40) best tlv triple is determined: according to next frame transformation parameter and the transfer of data result of prediction, calculate the reward function of the tlv triple be made up of modulation and encoding scheme, MIMO mode, dual-mode, the tlv triple selecting remuneration the highest combining from all modulation and encoding scheme, MIMO mode, dual-mode is best of breed, and its speed determined and mode of operation are combined as the parameter of next frame transfer of data.

3. wireless network data transmission method according to claim 2, is characterized in that, in described (10) acquisition, frame transmission result comprises: send totalframes f _{m, s, w}, message success sending probability ρ in frame _{m, s, w}successfully receive message total θ _{m, s, w};

Wherein, f _{m, s, w}represent under MCS m, MIMO mode s and mode of operation w condition, send totalframes, ρ _{m, s, w}represent message success sending probability in frame, θ _{m, s, w}represent and successfully receive message total, MCS is modulation and encoding scheme, and MIMO is multiple-input, multiple-output, and M represents whole available MCS sets { M|m ₁< ... < m ₁, m is wherein element, and S represents MIMO mode set { empty point of many collection, space division multiplexing }, and s is wherein element, and W represents dual-mode set { half-duplex, full duplex }, and w is wherein element.

4. wireless network data transmission method according to claim 3, is characterized in that, described (20) predict that this frame transformation parameter step comprises:

(21) MIMO mode and dual-mode is predicted: according to previous frame data at the message success sending probability of different MIMO mode and dual-mode with successfully receive message total, the MIMO mode adopted when predicting next frame transfer of data and dual-mode;

(22) predict MCS: the MIMO mode predicted according to step (21) and dual-mode, and sent the successful reception message total of Frame and the successful probability of acceptance, the best modulation of prediction and encoding scheme MCS;

(23) the lower frame transformation parameter of prediction: the best tlv triple comprehensively drawing prediction, upgrades the remuneration of the best tlv triple of prediction.

5. wireless network data transmission method according to claim 4, is characterized in that, described (21) prediction MIMO mode and dual-mode step comprise:

(211) θ is calculated _{s, w}and ρ _{s, w}:

${\mathrm{\&theta;}}_{s,w}=\underset{m=1}{\overset{R}{\mathrm{\&Sigma;}}}{\mathrm{\&theta;}}_{m,s,w}/R$

${\mathrm{\&rho;}}_{s,w}=\underset{m=1}{\overset{R}{\mathrm{\&Sigma;}}}{\mathrm{\&rho;}}_{m,s,w}/R,$

In formula, θ _{s, w}for the average successfully reception message number under MIMO mode s and dual-mode w, ρ _{s, w}for the average message success sending probability under MIMO mode s and dual-mode w, R is total MCS quantity.

(212) by θ _{s, w}and ρ _{s, w}the MIMO mode of classical CART method structure and the post-class processing of dual-mode carry out classifying and identifying, draws the best MIMO mode s of prediction ^*with dual-mode w ^*;

(213) at s ^*and w ^*the frame number of lower transmission and successfully send message number and adjust, that is:

$f_{{m,s}^{*},w}=f_{{m,s}^{*},w}+1,{\mathrm{\&theta;}}_{{m,s}^{*},w}={\mathrm{\&theta;}}_{{m,s}^{*},w}+{\mathrm{\&rho;}}_{{m,s}^{*},w}\&times;N_{{m,s}^{*},w}$

$f_{m,s,w^{*}}=f_{m,s,w^{*}}+1,{\mathrm{\&theta;}}_{m,s,w^{*}}={\mathrm{\&theta;}}_{m,s,w^{*}}+{\mathrm{\&rho;}}_{m,s,w^{*}}\&times;N_{m,s,w^{*}},$

In formula, with then be illustrated respectively in MIMO mode s ^*under transmission totalframes, the message number that sends in message success sending probability and every frame in frame, and with then be illustrated respectively in the message number sent in the successful sending probability of message in the transmission totalframes under MIMO mode w*, frame and every frame.

6. wireless network data transmission method according to claim 5, is characterized in that, described (22) prediction MCS step comprises:

(221) computational prediction with that is:

${\mathrm{\&rho;}}_{m,s^{*},w^{*}}\left(i\right)={\mathrm{\&sigma;\&rho;}}_{m,s^{*},w^{*}}(i-1)+(1-\mathrm{\&sigma;})\frac{\underset{k=2}{\overset{\mathrm{win}}{\mathrm{\&Sigma;}}}{\mathrm{\&rho;}}_{m,s^{*},w^{*}}(i-k)}{\mathrm{win}}$

${\mathrm{\&theta;}}_{m,s^{*},w^{*}}\left(i\right)={\mathrm{\&theta;}}_{m,s^{*},w^{*}}(i-1)+{\mathrm{\&rho;}}_{m,s^{*},w^{*}}\left(i\right)\&times;N_{m,s^{*},w^{*}}$

In formula, win is moving average window size; σ is smoothing factor, and σ value is larger, and recent obtained θ and ρ impact when predicting MCS is larger;

(222) by θ _{m, s, w}(i) and ρ _{m, s, w}i () carries out classifying and identifying on the MCS post-class processing built, draw the best MCSm of prediction ^*;

(223) to m ^*the frame number of lower transmission adjusts, that is:

$f_{m^{*},s^{*},w^{*}}=f_{m^{*},s^{*},w^{*}}+1.$

7. wireless network data transmission method according to claim 5, is characterized in that, described (23) predict that this frame transformation parameter step comprises:

(231) the best tlv triple < m of prediction is comprehensively drawn ^*, s ^*, w ^*>, and upgrade tlv triple < m according to following formula ^*, s ^*, w ^*>'s

${\mathrm{\&theta;}}_{m^{*},s^{*},w^{*}}=\mathrm{\&gamma;}\&times;({\mathrm{\&theta;}}_{m^{*},s^{*},w^{*}}+{\mathrm{\&rho;}}_{m^{*},s^{*},w^{*}}\left(i\right)\&times;N_{m^{*},s^{*},w^{*}})$

(232) will with substitute into remuneration computing function and upgrade the best tlv triple < m of prediction ^*, s ^*, w ^*the remuneration of >.

8. wireless network data transmission method according to claim 7, is characterized in that, described (30) predict that this frame transmission result step is specially:

${\mathrm{\&theta;}}_{m^{*},s^{*},w^{*}}=\mathrm{\&gamma;}\&times;({\mathrm{\&theta;}}_{m^{*},s^{*},w^{*}}+{\mathrm{\&rho;}}_{m^{*},s^{*},w^{*}}\left(i\right)\&times;N_{m^{*},s^{*},w^{*}}),$

${\mathrm{\&rho;}}_{m,s^{*},w^{*}}\left(i\right)={\mathrm{\&sigma;\&rho;}}_{m,s^{*},w^{*}}(i-1)+(1-\mathrm{\&sigma;})\frac{\underset{k=2}{\overset{\mathrm{win}}{\mathrm{\&Sigma;}}}{\mathrm{\&rho;}}_{m,s^{*},w^{*}(i-k)}}{\mathrm{win}}$

${\mathrm{\&theta;}}_{m,s^{*},w^{*}}\left(i\right)={\mathrm{\&theta;}}_{m,s^{*},w^{*}}(i-1)+{\mathrm{\&rho;}}_{m,s^{*},w^{*}}\left(i\right)\&times;N_{m,s^{*},w^{*},}$

${\mathrm{\&theta;}}_{s,w}=\underset{m=1}{\overset{R}{\mathrm{\&Sigma;}}}{\mathrm{\&theta;}}_{m,s,w}/R$

${\mathrm{\&rho;}}_{s,w}=\underset{m=1}{\overset{R}{\mathrm{\&Sigma;}}}{\mathrm{\&rho;}}_{m,s,w}/R,$

In formula, be that the i-th frame transmission adopts tlv triple < m ^*, s ^*, w ^*the prediction message success sending probability of >, γ is discount factor, and γ < 1, R is total MCS quantity, N _{m, s, w}then represent the message amount that this frame can send in time T, θ _{s, w}successfully message number is received, ρ for average _{s, w}for average message success sending probability.

9. wireless network data transmission method according to claim 8, is characterized in that, described (40) determine that best tlv triple step is specially:

(41) reward function is calculated as follows:

${\mathrm{\&mu;}}_{m,s,w}=\frac{{\mathrm{\&theta;}}_{m,s,w}}{f_{m,s,w}}+B\sqrt{\frac{\mathrm{\&alpha;}\log ({\mathrm{\&Sigma;}}_{m^{\&prime;}\&Element;M,s^{\&prime;}\&Element;S,w^{\&prime;}\&Element;W}{f}_{m^{\&prime;},s^{\&prime;},w^{\&prime;}}\&times;N_{m^{\&prime;},s^{\&prime;},w^{\&prime;}})}{f_{m,s,w}\&times;N_{m,s,w}}}$

(42) tlv triple the highest for remuneration is defined as best tlv triple, that is:

〈m，s，w〉＝argmaxμ _m，s，w(i)。