CN106454941A - Optimal congestion control method for wireless network - Google Patents

Abstract

The invention aims at a problem that the congestion control of the wireless network is difficult to achieve the dynamic and high-efficiency management. The method achieves the optimal utilization of resources of the wireless network through building an optimal congestion control rate model and a congestion optimal control mechanism.

Description

A kind of optimum jamming control method of wireless network

Technical field

The present invention relates to communication network field, particularly relate to queueing theory, and optimum theory.

Background technology

Radio network technique always is research and the focus of application.Numerous experts and researcher are to wireless network skill Constantly analyzing and research of art so that it is obtained rapid lasting development, the 802.11X standard of more and more species is suggested, Constantly being updated, the wireless network architecture of various novelties is suggested, and is also constantly carrying out their realization, to them Dispose and application is also being carried out in a deep going way.

When there is too much message in a network, the performance of network can decline, and this phenomenon is referred to as congestion, in network Congestion derives from Internet resources and the lack of uniformity of network traffics distribution.Congestion will not disappear with the raising of network throughput Remove.By network congestion Producing reason it is recognized that while the generation of congestion comes from shortage of resources, but unilaterally increase resource can not Avoiding the generation of congestion, sometimes even increasing the weight of Congestion Level SPCC, a kind of typical congestion control mechanism is as shown in Figure 1.For example, increase Add gateway caches and can increase the delay by gateway for the message, when packet through long queue up complete to forward when they early oneself Time-out, and source thinks that these packets have been dropped, and starts to retransmit, but these packets still transmit in a network, on the contrary Waste Internet resources and increased the weight of network congestion.And for example, the processing speed of processor may cause network congestion too slowly, but single The performance of pure this processor of raising, the bottleneck of network can be transferred to again other places, cause not mating of system components.Always It, there is the difficulty of following several respects in the design of congestion avoidance algorithm：

(1) distributivity of algorithm：The realization of congestion avoidance algorithm is distributed in different network layers and multiple network joint In point.Use distributed congestion avoidance algorithm can reduce the process complexity of individual node, improve the sane of network simultaneously Property.

(2) extensibility of algorithm：In network, performance everywhere has very big difference, for different network conditions, as The change of network size, the change of bandwidth, the change of chain circuit transmission time delay, different end system situations, and there is multiple number During according to stream, congestion avoidance algorithm all should have relatively good performance indications.

(3) performance requirement of algorithm：Congestion avoidance algorithm has very high requirement to performance, including the efficiency of algorithm, justice Property, stability, robustness and convergence.Common congestion control policy can only achieve the performance requirement of part, therefore to these Index needs to consider.

(4) easy implementation of algorithm：The design of congestion avoidance algorithm is simple as far as possible with realization, not only to reduce as far as possible Additional network traffics, and the complexity of feedback signal to be reduced, the design of congestion avoidance algorithm must also be as far as possible simultaneously Reduce the complexity of the amount of calculation at network node for this algorithm and realization.

(5) complexity of congestion：Communication network oneself develop into a huge complex system, its complexity is net The complexity of network topological structure, the complexity of network data flow such as self similarity, self-organizing is critical and congestion phase transition phenomena etc..

Therefore, it is the transmission enhancing efficiency realizing wireless mesh network, it is necessary to set up efficient congestion control mechanism.

Content of the invention

The technical problem to be solved is：During by setting up terminal wake-up mechanism, data encryption mechanism and realizing Clock Synchronous fluorimetry, it is achieved the safety encryption certification of wireless network.

The present invention solves that the technical scheme that above-mentioned technical problem is used comprises the following steps, as shown in Figure 2：

A, the optimum congestion control Rate Models of foundation；

B, set up congestion optimum control mechanism.

In described step A, set up optimum congestion control Rate Models：

0≤p_u≤1,0<q_v<1

Wherein source node s to destination node t ∈ T_sTransmission path be A, A and logical links I_uThe overlapping transmission road of=(f, H) Footpath collection is combined into B,For in B through the business flow rate of service convergence node h, T_sFor the purpose of node set,Save for origin to source Put s and the business flow rate through logical links (f, H), x_sFor the business flow rate of the session that source node s initiates, r_u,v For logical links I_uThe transfer rate of ∈ (f, J), p_uIt is u-th logical links I_uWeight coefficient, q_vIt is v-th transmitting scene Weight coefficient,S is source node set, and logical links (f, H) uses broadcast, H For Business Stream aggregation node set, tr (u) is logical links I_uTransmitting terminal,For logic-based link I_uThe industry of session s Business flow velocity rate, ε_fFor the primary power of node f, e_u,vFor logical links I_uAverage transmission energy,T_fFor node f's Life span, N is the node set in network, x_f,sFor decision variable, if the source node that node f is session, then x_f,s=x_sIf, Node f is the sink node of session, then x_f,s=-x_sIf node f is not for source node or sink node, the then x of session_f,s=0, L For logical links set, λ, λ₁And λ₂It is balance weight coefficient, W₁And W₂It is respectively λ₁And λ₂Regulatory factor, F is set of node Close.

In described step B, carry out fuzzy learning optimization, be specially：A. initialization q value, q [i, j]=0,1≤i≤N, 1≤ j≤J；B., behavior trigger policy is set：When triggering probability and being 1-ε, a_i=argmax_kQ [i, k], when triggering probability and being ε, a_i =random{a_k, k=1,2 ..., J}；C. the global behavior parameter of fuzzy logic control is obtainedD. count CalculateE. behavioral parameters is usedAnd enter NextState s (t+1)； F. new state amount is obtainedG. error signal Δ Q=r (t+1)+γ V_t(s (t), a), γ is weight coefficient to (s (t+1))-Q；H. rule q [i, a are used_i]⁺=q [i_,a_i]+η·ΔQ·α_i(s(t)) Update q value；I. repeat the above steps, and constantly update s (t), when obtaining the convergence domain of optimal solution, then stop calculating, wherein α_iS () is a_iWeight coefficient, Q (s (t), a) for state be s (t), iterations is t and fuzzy control behavioral parameters is a's Q functional value, α_iS () is the degree of belief of fuzzy logic ordination i, r (t+1) is that forward strengthens signal, and η is learning rate, and i is fuzzy Logic rules number, j is fuzzy logic ordination correction number.

In described step B, the fuzzy rule of congestion control is set, is specially：Fuzzy control unit is by fuzzy rule base, mould Stick with paste processing unit, ambiguity solution processing unit and interference engine composition；Fuzzy Processing unit receives R_ijWith HO margin, and produce Corresponding controlling behavior, as it is shown on figure 3, be specially：(1). work as R_ijWhen being H for lack of balance (i → j) and HO margin, Fuzzy Control Behavior processed is EL；(2). work as R_ijWhen being M for lack of balance (i → j) and HO margin, fuzzy control behavior is VL；(3). work as R_ij When being L for lack of balance (i → j) and HO margin, fuzzy control behavior is L；(4). work as R_ijBe equilibrium and HO margin be H When, fuzzy control behavior is N；(5). work as R_ijWhen being M for equilibrium with HO margin, fuzzy control behavior is N；(6). work as R_ijFor When equilibrium and HO margin are L, fuzzy control behavior is N；(7). work as R_ijWhen being L for lack of balance (i → j) and HO margin, Fuzzy control behavior is EH；(8). work as R_ijWhen being M for lack of balance (i → j) and HO margin, fuzzy control behavior is VH；Wherein L is rudimentary, and M is middle rank, and H is senior, and EL is ultralow level, and VL is very rudimentary, and N is empty set, and VH is very senior, and EH is superelevation level, R_ijRepresent Business Stream by i to j, R for lack of balance (i → j)_ijRepresent between i and j, there is not Business Stream transmission, R for equilibrium_ij= CBR_i-CBR_j, as shown in Figure 4.

In described step B, setting up optimum Congestion Control Model, it is by wireless network, fuzzy control unit and fuzzy learning Optimize unit composition, as it is shown in figure 5, fuzzy learning optimizes unit receives the CDR that the status information from wireless network obtains_iWith CDR_jInformation, and export Δ HO margin₁, fuzzy control unit receives to the HO margin and (CBR of wireless network_i- CBR_j) information and fuzzy learning optimize the output Δ HO margin of unit₁, and export Δ HO margin₂It is used for wireless network The congestion control ability of regulation wireless network, whereinFor call blocking rate,For call dropping rates, N_blockedFor number of calls blocked in call access control, N_acceptedFor connecing Receive control in received call number, N_droppedFor the number of calls being dropped in call access control, N_succFor call access control becomes Work(completes the number of calls of call.

Brief description

A kind of typical congestion control mechanism schematic diagram of Fig. 1

The optimum congestion control schematic flow sheet of Fig. 2 wireless network

Fig. 3 fuzzy control cellular construction schematic diagram

Fig. 4 fuzzy control rule

The optimum congestion control architecture schematic diagram of Fig. 5

Detailed description of the invention

For reaching above-mentioned purpose, technical scheme is as follows：

The first step, sets up optimum congestion control Rate Models：

0≤p_u≤1,0<q_v<1

Wherein source node s to destination node t ∈ T_sTransmission path be A, A and logical links I_uThe overlapping transmission road of=(f, H) Footpath collection is combined into B,For in B through the business flow rate of service convergence node h, T_sFor the purpose of node set,Save for origin to source Put s and the business flow rate through logical links (f, H), x_sFor the business flow rate of the session that source node s initiates, r_u,v For logical links I_uThe transfer rate of ∈ (f, J), p_uIt is u-th logical links I_uWeight coefficient, q_vIt is v-th transmitting scene Weight coefficient,S is source node set, and logical links (f, H) uses broadcast, H For Business Stream aggregation node set, tr (u) is logical links I_uTransmitting terminal,For logic-based link I_uThe industry of session s Business flow velocity rate, ε_fFor the primary power of node f, e_u,vFor logical links I_uAverage transmission energy,T_fFor node f's Life span, N is the node set in network, x_f,sFor decision variable, if the source node that node f is session, then x_f,s=x_sIf, Node f is the sink node of session, then x_f,s=-x_sIf node f is not for source node or sink node, the then x of session_f,s=0, L For logical links set, λ, λ₁And λ₂It is balance weight coefficient, W₁And W₂It is respectively λ₁And λ₂Regulatory factor, F is set of node Close.

Second step, carries out fuzzy learning optimization, is specially：A. q value, q [i, j]=0,1≤i≤N, 1≤j≤J are initialized； B., behavior trigger policy is set：When triggering probability and being 1-ε, a_i=argmax_kQ [i, k], when triggering probability and being ε, a_i= random{a_k, k=1,2 ..., J}；C. the global behavior parameter of fuzzy logic control is obtainedD. calculateE. behavioral parameters is usedAnd enter NextState s (t+1)；f. Obtain new state amountG. error signal Δ Q=r (t+1)+γ V_t (s (t), a), γ is weight coefficient to (s (t+1))-Q；H. rule q [i, a are used_i]⁺=q [i, a_i]+η·ΔQ·α_i(s (t)) more New q value；I. repeat the above steps, and constantly update s (t), when obtaining the convergence domain of optimal solution, then stop calculating, wherein α_i S () is a_iWeight coefficient, (s (t), a) for being s (t), the Q that iterations is t and fuzzy control behavioral parameters is a in state for Q Functional value, α_iS () is the degree of belief of fuzzy logic ordination i, r (t+1) is that forward strengthens signal, and η is learning rate, and i is fuzzy Logic rules number, j is fuzzy logic ordination correction number.

3rd step, arranges congestion control rule, is specially：Fuzzy control unit by fuzzy rule base, Fuzzy Processing unit, Ambiguity solution processing unit and interference engine composition；Fuzzy Processing unit receives R_ijWith HO margin, and produce corresponding control row For being specially：(1). work as R_ijWhen being H for lack of balance (i → j) and HO margin, fuzzy control behavior is EL；(2). work as R_ijFor When lack of balance (i → j) and HO margin are M, fuzzy control behavior is VL；(3). work as R_ijFor lack of balance (i → j) and HO When margin is L, fuzzy control behavior is L；(4). work as R_ijWhen being H for equilibrium with HO margin, fuzzy control behavior is N； (5). work as R_ijWhen being M for equilibrium with HO margin, fuzzy control behavior is N；(6). work as R_ijBe equilibrium and HO margin be L When, fuzzy control behavior is N；(7). work as R_ijWhen being L for lack of balance (i → j) and HO margin, fuzzy control behavior is EH； (8). work as R_ijWhen being M for lack of balance (i → j) and HO margin, fuzzy control behavior is VH；Wherein L is rudimentary, and M is middle rank, H For senior, EL is ultralow level, and VL is very rudimentary, and N is empty set, and VH is very senior, and EH is superelevation level, R_ijFor lack of balance (i → j) table Show Business Stream by i to j, R_ijRepresent between i and j, there is not Business Stream transmission, R for equilibrium_ij=CBR_i-CBR_j.

4th step, sets up optimum Congestion Control Model, and it is optimized single by wireless network, fuzzy control unit and fuzzy learning Unit's composition, fuzzy learning optimizes unit and receives the CDR that the status information from wireless network obtains_iAnd CDR_jInformation, and export Δ HO margin₁, fuzzy control unit receives to the HO margin and (CBR of wireless network_i-CBR_j) information and fuzzy learning optimization The output Δ HO margin of unit₁, and export Δ HO margin₂To wireless network for regulating the congestion control of wireless network Ability, whereinFor call blocking rate,For call dropping rates, N_blockedFor number of calls blocked in call access control, N_acceptedFor received call number, N in call access control_dropped For the number of calls being dropped in call access control, N_succFor call access control is successfully completed the number of calls of call.

The present invention proposes the optimum jamming control method of a kind of wireless network, by setting up optimum congestion control speed mould Type and congestion optimum control mechanism, it is achieved the optimizing network resource utilization of wireless network.

Claims (5)

1. an optimum jamming control method for wireless network, by setting up optimum congestion control Rate Models and the optimum control of congestion Making mechanism, it is achieved the optimizing network resource utilization of wireless network, comprises the steps：

A, the optimum congestion control Rate Models of foundation；

B, set up congestion optimum control mechanism.

2. method according to claim 1, is characterized in that for described step A：Set up optimum congestion control Rate Models：

$\begin{array}{cc}\min & \mathrm{\&mu;}\left(\mathrm{\&lambda;}\right)={\mathrm{\&lambda;}}_1\frac{\underset{s\&Element;S}{\&Sigma;}\log \left(x_s\right)}{W_1}-{\mathrm{\&lambda;}}_2\frac{\underset{f\&Element;N}{\&Sigma;}{\mathrm{\&phi;}}_f^2}{W_2}\\ s.t.& \underset{\left\{H\right|(f,H)\&Element;L\}}{\&Sigma;}\underset{h\&Element;H}{\&Sigma;}{f}_{fHh}^{st}-\underset{h\&Element;N}{\&Sigma;}\underset{\left\{f\right|(h,F)\&Element;L,f\&Element;F\}}{\&Sigma;}{f}_{hFf}^{st}\&GreaterEqual;x_{f,s},\&ForAll;f\&Element;N,s\&Element;S,t\&Element;T_s\end{array}$

$\underset{\{u:tr\left(u\right)=f\}}{\&Sigma;}\underset{v=1}{\overset{n}{\&Sigma;}}{p}_u{e}_{u,v}{q}_v\&le;{\mathrm{\&phi;}}_f{\mathrm{\&epsiv;}}_f$

$\underset{h\&Element;H}{\&Sigma;}{f}_{fHh}^{st}\&le;g_{fH}^s$

$\underset{s\&Element;S}{\&Sigma;}{g}_u^s\&le;\underset{v=1}{\overset{m}{\&Sigma;}}{p}_u{r}_{u,v}{q}_v$

$\underset{u=1}{\overset{m}{\&Sigma;}}{p}_u=1,\underset{v=1}{\overset{n}{\&Sigma;}}{q}_v=1$

0≤p_u≤1,0<q_v<1

Wherein source node s to destination node t ∈ T_sTransmission path be A, A and logical links I_uThe overlapping transmission path of=(f, H) Collection is combined into B,For in B through the business flow rate of service convergence node h, T_sFor the purpose of node set,Save for origin to source Put s and the business flow rate through logical links (f, H), x_sFor the business flow rate of the session that source node s initiates, r_u,v For logical links I_uThe transfer rate of ∈ (f, J), p_uIt is u-th logical links I_uWeight coefficient, q_vIt is v-th transmitting scene Weight coefficient,S is source node set, and logical links (f, H) uses broadcast, H For Business Stream aggregation node set, tr (u) is logical links I_uTransmitting terminal,For logic-based link I_uThe business of session s Flow velocity rate, ε_fFor the primary power of node f, e_u,vFor logical links I_uAverage transmission energy,T_fLife for node f Depositing the time, N is the node set in network, x_f,sFor decision variable, if the source node that node f is session, then x_f,s=x_sIf, joint Point f is the sink node of session, then x_f,s=-x_sIf node f is not for source node or sink node, the then x of session_f,s=0, L are Logical links set, λ, λ₁And λ₂It is balance weight coefficient, W₁And W₂It is respectively λ₁And λ₂Regulatory factor, F is node set.

3. method according to claim 1, is characterized in that for described step B：Carry out fuzzy learning optimization, be specially：a. Initialize q value, q [i, j]=0,1≤i≤N, 1≤j≤J；B., behavior trigger policy is set：When triggering probability and being 1-ε, a_i= arg max_kQ [i, k], when triggering probability and being ε, a_i=random{a_k, k=1,2 ..., J}；C. fuzzy logic control is obtained Global behavior parameterD. calculateE. behavioral parameters is usedAnd enter NextState s (t+1)；F. new state amount is obtained G. error signal Δ Q=r (t+1)+γ V_t(s (t), a), γ is weight coefficient to (s (t+1))-Q；H. rule q is used [i,a_i]⁺=q [i, a_i]+η·ΔQ·α_i(s (t)) updates q value；I. repeat the above steps, and constantly update s (t), work as acquisition During the convergence domain of optimal solution, then stop calculating, wherein α_iS () is a_iWeight coefficient, (s (t), a) for being s (t), repeatedly in state for Q The Q functional value that generation number is t and fuzzy control behavioral parameters is a, α_iS () is the degree of belief of fuzzy logic ordination i, r (t+1) is Forward strengthens signal, and η is learning rate, and i is fuzzy logic ordination number, and j is fuzzy logic ordination correction number.

4. method according to claim 1, is characterized in that for described step B：Congestion control rule is set, is specially：Mould Stick with paste control unit to be made up of fuzzy rule base, Fuzzy Processing unit, ambiguity solution processing unit and interference engine；Fuzzy Processing unit Receive R_ijWith HO margin, and produce corresponding controlling behavior, be specially：(1). work as R_ijFor lack of balance (i → j) and HO When margin is H, fuzzy control behavior is EL；(2). work as R_ijWhen being M for lack of balance (i → j) and HO margin, fuzzy control Behavior is VL；(3). work as R_ijWhen being L for lack of balance (i → j) and HO margin, fuzzy control behavior is L；(4). work as R_ijFor all When weighing apparatus and HO margin are H, fuzzy control behavior is N；(5). work as R_ijWhen being M for equilibrium with HO margin, fuzzy control row For for N；(6). work as R_ijWhen being L for equilibrium with HO margin, fuzzy control behavior is N；(7). work as R_ijFor lack of balance (i → j) When being L with HO margin, fuzzy control behavior is EH；(8). work as R_ijWhen being M for lack of balance (i → j) and HO margin, fuzzy Controlling behavior is VH；Wherein L is rudimentary, and M is middle rank, and H is senior, and EL is ultralow level, and VL is very rudimentary, and N is empty set, and VH is for very Senior, EH is superelevation level, R_ijRepresent Business Stream by i to j, R for lack of balance (i → j)_ijRepresent do not exist between i and j for equilibrium Business Stream transmits, R_ij=CBR_i-CBR_j, HOMargin is Edge position control weights.

5. method according to claim 1, is characterized in that for described step B：Setting up optimum Congestion Control Model, it is by nothing Gauze network, fuzzy control unit and fuzzy learning optimize unit composition, and fuzzy learning optimizes unit and receives the shape from wireless network The CDR of state information acquisition_iAnd CDR_jInformation, and export Δ HO margin₁, fuzzy control unit receives to the HO of wireless network Margin and (CBR_i-CBR_j) information and fuzzy learning optimize the output Δ HO margin of unit₁, and export Δ HO margin₂ To wireless network for regulating the congestion control ability of wireless network, whereinFor call congestion Rate,For call dropping rates, N_blockedFor number of calls blocked in call access control, N_acceptedFor Received call number in call access control, N_droppedFor the number of calls being dropped in call access control, N_succFor in call access control It is successfully completed the number of calls of call.