CN108391272A - Dynamically distribute the spectrum service model of recycling and centralized method for allocating dynamic frequency spectrums - Google Patents
Dynamically distribute the spectrum service model of recycling and centralized method for allocating dynamic frequency spectrums Download PDFInfo
- Publication number
- CN108391272A CN108391272A CN201810177017.2A CN201810177017A CN108391272A CN 108391272 A CN108391272 A CN 108391272A CN 201810177017 A CN201810177017 A CN 201810177017A CN 108391272 A CN108391272 A CN 108391272A
- Authority
- CN
- China
- Prior art keywords
- spectrum
- business
- user
- frequency spectrum
- bandwidth
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W16/00—Network planning, e.g. coverage or traffic planning tools; Network deployment, e.g. resource partitioning or cells structures
- H04W16/02—Resource partitioning among network components, e.g. reuse partitioning
- H04W16/10—Dynamic resource partitioning
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W16/00—Network planning, e.g. coverage or traffic planning tools; Network deployment, e.g. resource partitioning or cells structures
- H04W16/14—Spectrum sharing arrangements between different networks
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W16/00—Network planning, e.g. coverage or traffic planning tools; Network deployment, e.g. resource partitioning or cells structures
- H04W16/22—Traffic simulation tools or models
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Telephonic Communication Services (AREA)
Abstract
The invention discloses a kind of spectrum service model dynamically distributing recycling and centralized method for allocating dynamic frequency spectrums.The model is:Communication request from the user reaches center spectrum manager according to random order, center spectrum manager is that user distributes frequency spectrum progress real-time Communication for Power, current frequency spectrum allocation result can influence the satisfaction performance of future customer, and frequency spectrum is discharged after sign off, realize frequency spectrum recycling.Method is:Off-line training first obtains the optimal spectrum allocation strategy under different system state;Then real time spectrum distributes, and when customer service reaches, center spectrum manager carries out real time spectrum distribution according to the result of off-line training;Final online adjusts, and when system satisfaction performance is less than preset threshold value, starts on-line tuning and reacquires optimal spectrum allocation strategy.The present invention is based on offline Q to learn to carry out centralized dynamic frequency spectrum deployment, improves the availability of frequency spectrum and system performance.
Description
Technical field
The invention belongs to wireless communication technology field, especially a kind of spectrum service model and concentration dynamically distributing recycling
Formula method for allocating dynamic frequency spectrums.
Background technology
Frequency spectrum resource is a kind of non-renewable resources.Since existing frequency spectrum distribution mechanism is asked there are the availability of frequency spectrum is low
Topic, with the surge of mobile terminal device quantity, frequency spectrum resource shortage problem is on the rise.Therefore the availability of frequency spectrum is improved to compel
The eyebrows and eyelashes.Since dynamic frequency spectrum deployment technology can effectively improve the availability of frequency spectrum, it is concerned in recent years.
Have the researchs largely shared for the dynamic spectrum based on database intelligent decision at present, be divided into centralization and divide
Two kinds of algorithms of cloth.In distributed algorithm, needs the user communicated to access database first and obtain usable spectrum list, then
By analyze quick obtaining environmental characteristic, finally by game compete mode carry out resource allocation (bibliography Chen X,
Gong X,Yang L,et al.A social group utility maximization framework with
applications in database assisted spectrum access[C]//INFOCOM,2014Proceedings
IEEE.IEEE,2014:1959-1967.).In centralized algorithm, dynamic frequency spectrum deployment system includes a center spectrum pipe
Device is managed, the user of communication reports communication request to center spectrum manager first, and then center spectrum manager is inquired in real time
Database obtains usable spectrum information, and carries out frequency spectrum distribution according to current usable spectrum information.But at present for collection
The research of Chinese style algorithm has following deficiency:1, it is assumed that communication requirement that all users send out while reaching, this is in cognitive radio
It is unpractical in network, does not meet the randomness of customer service arrival;2, inquire data next time in center spectrum manager
Before library, it is assumed that user occupies always current frequency range, this does not meet randomness (the bibliography Yiyang that customer service is left away
Pei,Yugang Ma,Peh,E.C.Y.,Ser Wah Oh,Ming-HungTao,Dynamic spectrum assignment
for white space devices with dynamic and heterogeneous bandwidth
requirements,IEEE Wireless Communications and Networking Conference(WCNC),
2015.)。
The final purpose of dynamic frequency spectrum deployment is to improve the availability of frequency spectrum and optimization system performance, therefore, how to frequency spectrum
Resource, which carries out effectively distribution, has important research significance.The existing centralization dynamic frequency being directed to based on database intelligent decision
Compose the research of distribution, it is believed that database obtains radio environment information by periodically perceiving, although center spectrum manager
(Spectrum Manager, SM) can access database and obtain usable spectrum information, but the usable spectrum information in real time
It is not corresponded to current spectrum environment, its occupied frequency spectrum can not be timely after this will lead to cognitive user sign off
It withdraws, the availability of frequency spectrum is low.
Invention content
The purpose of the present invention is to provide a kind of spectrum service models dynamically distributing recycling, and are learnt based on offline Q
Centralized method for allocating dynamic frequency spectrums.
Realize that the technical solution of the object of the invention is:A kind of spectrum service model dynamically distributing recycling, the model
Reaching and leaving away and be handled as follows to business:The service request of all users in network reaches center according to random order
Spectrum manager;After sign off, the frequency spectrum of user's release busy realizes recycling and frequency spectrum resource being total in sequential of frequency spectrum
It enjoys;The arrival of customer service and leave away independently of each other, do not interfere with each other.
A kind of centralized method for allocating dynamic frequency spectrums for the spectrum service model dynamically distributing recycling, includes the following steps:
Step 1, off-line training step:According to spectrum environment variation priori rule and customer service type distribution, in
Heart spectrum manager realizes the convergence of dynamic frequency spectrum deployment algorithm using Q study optimization dynamic frequency spectrum deployment strategies;
Step 2, real time spectrum distribution:When service request reaches, center spectrum manager accesses frequency spectrum data library obtains
Then currently available spectral slice list carries out dynamic frequency spectrum deployment according to the result of Q study optimizations;
Step 3, on-line tuning stage:Pass through studying history spectrum information and historic demand information obtaining system statistic
Value starts on-line tuning if the value for adding up average satisfaction in system is less than threshold value.
Further, off-line training step described in step 1, it is specific as follows:
Dynamic frequency spectrum deployment problem is modeled as markoff process, optimization aim is that system adds up average satisfaction most
Greatly:
Wherein, k indicates the number of communication request, ukIt is the satisfaction of k-th of communication request, ak∈ { 0,1 ... n ... ND },
N number of usable spectrum segment is shared in system, the maximum possible distribution condition number of all spectral slices is D, ak=0 indicates the request
It is rejected;ak=n indicates that current request is assigned to theA spectral slice;
Spectrum allocation strategy is obtained using off-line training, is as follows:
(1.1) it initializes:Iterations k=1, action select probability vector are setQk(sk,ak),ak∈ { 0,1 ..., ND }, wherein skFor system mode,sk∈ S, S are system
State set;System mode includes two parts,It is usable spectrum segment and its bandwidth, TkIt is current
The type of service of request;akIt is numbered for spectral slice, Qk(sk,ak) it is k moment skA is selected under statekCorresponding Q values, Q when action
To act utility function, for evaluating the quality for taking some to act in a particular state;
(1.2) when k-th of user asks to reach, system mode sk, user is according to action selection strategySelection
Act ak;
(1.3) according to the action of selection, user obtains a stochastic return rk;
If the type of service of user is rigid business, T is enabledk=1, return value at this time is:
Wherein, bkIt is allocated to the bandwidth of k-th of business, BthBandwidth threshold value;
If the type of service of user is stream business, T is enabledk=2, return value at this time is:
Wherein, bkIt is allocated to the bandwidth of k-th of business, α and β are the constants for determining extent function shape;
If the type of service of user is elastic business, T is enabledk=3, return value at this time is:
Wherein, bkIt is allocated to the bandwidth of k-th of business, BthBandwidth threshold value, γ1It is scale constant;
(1.4) Q values are updated to all states, more new formula is as follows:
Wherein, akIt is learning rate, γ2It is discount factor;
(1.5) to all state update action select probability vectors, more new formula is as follows:
Wherein, γ3> 0 is temperature coefficient, is used for detecting and utilizing during Schistosomiasis control;
(1.6) for all system modes, acting in select probability vector that there are one elements to be more than 0.99, cycle terminates,
Otherwise return to step (1.2).
Further, real time spectrum described in step 2 distributes, specific as follows:
(2.1) reporting of user communication request;
(2.2) center spectrum manager accesses database obtains currently available frequency spectrum bands;
(2.3) center spectrum manager carries out real time spectrum distribution according to the result of off-line training or on-line tuning, specifically
It realizes as follows:
When k-th of request reaches, the frequency spectrum state that center spectrum manager accesses frequency spectrum data library obtains isSystem mode is denoted as s at this timek,sk∈ S, S are system
State set;System mode includes two parts,It is usable spectrum segment and its bandwidth, TkIt is current
The type of service of request;
Minimal frequency bandwidth allocation is set as Δ, if user is assigned to n-th of spectral slice, which can distribute
Bandwidth beIn one kind;
It is set in a certain decision moment, N number of usable spectrum segment, the maximum bandwidth of n-th of spectral slice are shared in system
ForAnd the maximum possible distribution condition number of all spectral slices is D, then for each user, the knot of frequency spectrum distribution
Fruit shares ND+1 kinds, wherein containing the case where being rejected;When user asks to reach, center spectrum manager is according to system shape
State selects an action ak, wherein ak∈ { 0,1 ... n ... ND }, ak=0 indicates that the request is rejected, the bandwidth being assigned at this time
bk=0;ak=n indicates that current request is assigned to theA spectral slice, the bandwidth b being assigned at this timekFor:
Each spectral slice is at best able to accessA user;
It sets the arrival of user and leaves away and all obey Poisson distribution, that is, the time interval for reaching and leaving away obeys index point
Cloth, primary user arrival time interval and the time interval left away be respectively λpAnd μp, the arrival time of secondary user interval and leaves away
Time interval be respectively λcAnd μc, when+1 request of kth reaches, frequency spectrum state is:
Wherein,
Wherein,
Further, the type of service includes rigid business, flows three kinds of business and elastic business, specific as follows:
(1) rigid business
Rigid business is real time business, includes the data transmission of video calling, tele-medicine, security risk prevention and control;If
It is unsatisfactory for basic communication requirement, the business of the type does not allow to access network, the mathematic(al) representation of the type business satisfaction
As follows, when meeting communication requirement, satisfaction 1;Otherwise satisfaction is 0:
Wherein, bkIt is allocated to the bandwidth of k-th of business, BthBandwidth threshold value;
(2) business is flowed
Stream business is also a kind of real time business, including online audio, Online Video, game on line;The type business is satisfied with
The mathematic(al) representation of degree is as follows:
Wherein, bkIt is allocated to the bandwidth of k-th of business, α and β are the constants for determining extent function shape;
(3) elastic business
Elastic business is non-real-time service, includes the transmission of data or picture;The mathematical expression of the type business satisfaction
Formula is as follows:
Wherein, bkIt is allocated to the bandwidth of k-th of business, BthBandwidth threshold value, γ are scale constants.
Compared with prior art, the present invention its remarkable advantage is:(1) actual communication scenes have been fully considered, with electricity
The dynamic allocation of white frequency spectrum are regarded as background, based on frequency spectrum resource pond, database real-time perception wireless environment is considered, recycles in time
The occupied frequency spectrum of historical user realizes the frequency spectrum share of different user in time;(2) it is directed to the small-scale net of number of users
The drawbacks of network scene is designed the centralized method for allocating dynamic frequency spectrums learnt based on offline Q, overcomes distributed method is improved
The availability of frequency spectrum and system satisfaction performance.
Description of the drawings
Fig. 1 is the dynamic frequency spectrum deployment flow chart based on database.
Fig. 2 is the centralized dynamic frequency spectrum deployment flow diagram learnt based on offline Q.
Fig. 3 is Q value convergent schematic diagrames.
Fig. 4 is the system satisfaction performance comparison schematic diagram of model method and existing model method in the embodiment of the present invention.
Specific implementation mode
The present invention is using the dynamic allocation of the white frequency spectrum of TV as background, based on frequency spectrum resource pond, in order to improve spectrum utilization
Rate sets database real-time perception wireless environment, recycles historical user in time, that is, communicate the occupied frequency of the user terminated
Spectrum realizes the frequency spectrum share of different user in time, and the service request of user reaches center spectrum according to random order
Manager.
The present invention dynamically distributes the spectrum service model of recycling, which is characterized in that the model reaches and leaves away to business
It is handled as follows:The service request of all users in network reaches center spectrum manager according to random order;Sign off
Afterwards, the frequency spectrum of user's release busy realizes that the recycling of frequency spectrum and frequency spectrum resource are shared in sequential;The arrival of customer service and
It leaves away independently of each other, does not interfere with each other.
Fig. 1 is the dynamic frequency spectrum deployment flow chart based on database.When a certain user generates communication requirement, first in
Heart spectrum manager accesses the acquisition of frequency spectrum data library currently after reporting communication request, center spectrum manager to receive user's request
Usable spectrum fragment list, frequency spectrum distribution is then carried out according to current usable spectrum fragment list, refusal current business is asked
Current service request is sought or received, if receiving current request, specific frequency point and bandwidth are distributed for current request;Logical
After letter, the occupied frequency spectrum of user release realizes the timely recycling of frequency spectrum.Entire communication process is divided into following step
Suddenly:
1) user for generating communication requirement reports service request to center spectrum manager, which includes service class
Type information;
2) after receiving service request from the user, access database obtains currently available center spectrum manager
Spectral slice list;
3) center spectrum manager carries out frequency spectrum decision according to the type of service that current spectrum environment and user ask;
4) result of decision is fed back to user by center spectrum manager, completes dynamic frequency spectrum deployment;
5) sign off, user discharge occupied frequency spectrum, realize the timely recycling of frequency spectrum.
Since user's request reaches database according to random order, current frequency spectrum allocation result will influence future customer
Satisfaction, be based on this, the present invention propose it is a kind of dynamically distribute recycling spectrum service model centralized dynamic frequency spectrum deployment
Method, detailed process is as shown in Fig. 2, include the following steps:
Step 1, off-line training step:According to spectrum environment variation priori rule and customer service type distribution, in
Heart spectrum manager optimizes dynamic frequency spectrum deployment strategy using Q study, realizes the Fast Convergent of dynamic frequency spectrum deployment algorithm;
Step 2, real time spectrum distribution:When service request reaches, center spectrum manager accesses frequency spectrum data library obtains
Then currently available spectral slice list carries out dynamic frequency spectrum deployment according to the result of Q study;
Step 3, on-line tuning stage:Pass through studying history spectrum information and historic demand information obtaining system statistic
Value starts on-line tuning if the value for adding up average satisfaction in system is less than threshold value.
The specific implementation of the present invention is as follows:
Step 1, off-line training step:
Dynamic frequency spectrum deployment problem is modeled as markoff process, optimization aim is that system adds up average satisfaction most
Greatly:
Wherein, k indicates the number of communication request, ukIt is the satisfaction of k-th of communication request, ak∈ { 0,1 ... n ... ND },
N number of usable spectrum segment is shared in system, the maximum possible distribution condition number of all spectral slices is D, ak=0 indicates the request
It is rejected;ak=n indicates that current request is assigned to theA spectral slice;
Using off-line training, the spectrum allocation strategy stablized under specific environment feature is obtained, is as follows:
(1.1) it initializes:Iterations k=1, action select probability vector are setQk(sk,ak),ak∈ { 0,1 ..., ND }, wherein skFor system mode,sk∈ S, S are system
State set;System mode includes two parts,It is usable spectrum segment and its bandwidth, TkIt is current
The type of service of request;akIt is numbered for spectral slice, Qk(sk,ak) it is k moment skA is selected under statekCorresponding Q values, Q when action
To act utility function, for evaluating the quality for taking some to act in a particular state;
(1.2) when k-th of user asks to reach, system mode sk, user is according to action selection strategySelection
Act ak;
(1.3) according to the action of selection, user obtains a stochastic return rk;
If the type of service of user is rigid business, T is enabledk=1, return value at this time is:
Wherein, bkIt is allocated to the bandwidth of k-th of business, BthBandwidth threshold value;
If the type of service of user is stream business, T is enabledk=2, return value at this time is:
Wherein, bkIt is allocated to the bandwidth of k-th of business, α and β are the constants for determining extent function shape;
If the type of service of user is elastic business, T is enabledk=3, return value at this time is:
Wherein, bkIt is allocated to the bandwidth of k-th of business, BthBandwidth threshold value, γ1It is scale constant;
(1.4) Q values are updated to all states, more new formula is as follows:
Wherein, akIt is learning rate, γ2It is discount factor;
(1.5) to all state update action select probability vectors, more new formula is as follows:
Wherein, γ3> 0 is temperature coefficient, is used for detecting and utilizing during Schistosomiasis control;
(1.6) it for all system modes, acts in select probability vector and is substantial access to 1 there are one element, such as 0.99, it follows
Ring terminates, otherwise return to step (1.2).
Step 2, real time spectrum distribution, are as follows:
(2.1) reporting of user communication request;
(2.2) center spectrum manager accesses database obtains currently available frequency spectrum bands;
(2.3) center spectrum manager carries out real time spectrum distribution according to the result of off-line training or on-line tuning, specifically
It realizes as follows:
When k-th of request reaches, the frequency spectrum state that center spectrum manager accesses frequency spectrum data library obtains isSystem mode is denoted as s at this timek,sk∈ S, S are system
State set;System mode includes two parts,It is usable spectrum segment and its bandwidth, TkIt is current
The type of service of request;
Minimal frequency bandwidth allocation is set as Δ, if user is assigned to n-th of spectral slice, which can distribute
Bandwidth beIn one kind;
It is set in a certain decision moment, N number of usable spectrum segment, the maximum bandwidth of n-th of spectral slice are shared in system
ForAnd the maximum possible distribution condition number of all spectral slices is D, then for each user, the knot of frequency spectrum distribution
Fruit shares ND+1 kinds, wherein containing the case where being rejected;When user asks to reach, center spectrum manager is according to system shape
State selects an action ak, wherein ak∈ { 0,1 ... n ... ND }, ak=0 indicates that the request is rejected, the bandwidth being assigned at this time
bk=0;ak=n indicates that current request is assigned to theA spectral slice, the bandwidth b being assigned at this timekFor:
Each spectral slice is at best able to accessA user;
It sets the arrival of user and leaves away and all obey Poisson distribution, that is, the time interval for reaching and leaving away obeys index point
Cloth, primary user arrival time interval and the time interval left away be respectively λpAnd μp, the arrival time of secondary user interval and leaves away
Time interval be respectively λcAnd μc, when+1 request of kth reaches, frequency spectrum state is:
Wherein,
Wherein,
Further, the type of service includes rigid business, flows three kinds of business and elastic business, specific as follows:
(1) rigid business
Rigid business is the data transmission of real time business, including video calling, tele-medicine, security risk prevention and control etc.;Such as
Fruit is unsatisfactory for basic communication requirement, and the business of the type does not allow to access network, the mathematical expression of the type business satisfaction
Formula is as follows, when meeting communication requirement, satisfaction 1;Otherwise satisfaction is 0:
Wherein, bkIt is allocated to the bandwidth of k-th of business, BthBandwidth threshold value;
(2) business is flowed
Stream business is also a kind of real time business, including online audio, Online Video, game on line etc.;The type business is full
The mathematic(al) representation of meaning degree is as follows:
Wherein, bkIt is allocated to the bandwidth of k-th of business, α and β are the constants for determining extent function shape;
(3) elastic business
Elastic business is non-real-time service, includes the transmission of data or picture;The mathematical expression of the type business satisfaction
Formula is as follows:
Wherein, bkIt is allocated to the bandwidth of k-th of business, BthBandwidth threshold value, γ are scale constants.
When k-th of communication request reaches, system mode sk, at this time center spectrum manager must select one action ak,
And the user for sending communication request will obtain a return value (being satisfied with angle value), therefore dynamic frequency spectrum deployment problem can be modeled
For markov decision process.In view of the final purpose of dynamic frequency spectrum deployment be the accumulative average satisfaction of realization system most
Greatly, so dynamic frequency spectrum deployment problem optimization aim can be set as to accumulative average satisfaction maximum.
Step 3, on-line tuning stage:Center spectrum manager judges whether spectrum environment is become according to historical information
Change, starts on-line tuning, the same step of specific steps (1.2)-step (1.5) if spectrum environment is changed.
Embodiment 1
The specific embodiment of the present invention is described below, and system emulation uses Matlab softwares, parameter setting not to influence
It is general.The embodiment verifies the convergence and validity of put forward model and method, as shown in Figure 3 and Figure 4.When emulation starts,
Usable spectrum segments in network is N=4, and bandwidth is respectivelyΔ=1 is set,
The sum of then D=4, everything are 4 × 4+1=17;The collection of type of service is combined intoThe corresponding probability that reaches is
{0.4,0.3,0.3};The service parameter of all primary users is identical, and the service parameter of all secondary users is identical;Set λp=0.5, up
=0.8, λc=0.2, uc=0.5.
The present invention dynamically distributes the centralized method for allocating dynamic frequency spectrums of the spectrum service model of recycling, and detailed process is such as
Under:
Step 1, off-line training step:According to spectrum environment variation priori rule and customer service type distribution, in
Heart spectrum manager optimizes dynamic frequency spectrum deployment strategy using Q study, realizes the Fast Convergent of dynamic frequency spectrum deployment algorithm, tool
Steps are as follows for body:
Step 1.1, initialization.Iterations k=1, action select probability vector are setQk(sk,ak),ak∈{0,1,…,ND}。
Step 1.2, when k-th user asks to reach, system mode sk, user is according to action selection strategyChoosing
Select action ak。
Step 1.3, the action according to selection, user obtain a stochastic return rk。
If the type of service of user is rigid business, T is enabledk=1, return value calculation formula at this time is formula (2);
If the type of service of user is stream business, T is enabledk=2, return value calculation formula at this time is formula (3);
If the type of service of user is elastic business, T is enabledk=3, return value calculation formula at this time is formula (4);
Step 1.4 updates all states Q values, and more new formula is formula (5);
Step 1.5, to all state update action select probability vectors, more new formula is formula (6);
Step 1.6, for all system modes, act in select probability vector and be substantial access to 1 there are one element, such as
0.99, cycle terminates, otherwise return to step 1.2.
Step 2, real time spectrum distribution:When service request reaches, center spectrum manager accesses frequency spectrum data library obtains
Then currently available spectral slice list carries out dynamic frequency spectrum deployment according to the result of Q study.
Step 3, on-line tuning stage:Pass through studying history spectrum information and historic demand information obtaining system statistic
Value starts on-line tuning if the value for adding up average satisfaction in system is less than threshold value, and specific steps are walked with step 1.2-
Rapid 1.5.
Fig. 3 is Q value convergent schematic diagrames.From figure 3, it can be seen that by off-line training step, the convergence of Q values, for
Each system mode center spectrum manager can select stable spectrum allocation strategy.Offline Q learning algorithms can be fast
Speed convergence, approximately passes through 60 iteration, Q values can restrain.In addition, we can also obtain from Fig. 3, in current system shape
Under state, the frequency spectrum allocation result of the user is second spectral slice, bandwidth 5.Fig. 4 illustrates institute's extracting method model and tradition
The comparison of method model.From fig. 4, it can be seen that the accumulative average satisfaction of institute's extracting method is 0.7 or so, greedy algorithm adds up
Average satisfaction is 0.5 or so, and institute's extracting method about promotes 20% compared with greedy algorithm, system satisfaction performance.
To sum up, the present invention proposes the spectrum service model for dynamically distributing recycling, and essence has been carried out to the Behavior law of user
True description;Consider the small-scale network of number of users, in order to realize better system satisfaction performance, proposes the frequency of centralization
Compose allocation model;Meanwhile in order to solve the problems, such as that user asks the dynamic frequency spectrum deployment that is reached according to random order, propose to be based on from
The centralized dynamic frequency spectrum deployment scheme of line Q study, this method can Fast Convergent, and satisfaction performance is better than conventional method.
Claims (5)
1. it is a kind of dynamically distribute recycling spectrum service model, which is characterized in that the model to business reach and progress of leaving away
Following processing:The service request of all users in network reaches center spectrum manager according to random order;After sign off,
The frequency spectrum of user's release busy realizes that the recycling of frequency spectrum and frequency spectrum resource are shared in sequential;The arrival of customer service and from
It goes independently of each other, not interfereing with each other.
2. a kind of centralized method for allocating dynamic frequency spectrums for the spectrum service model dynamically distributing recycling, which is characterized in that including
Following steps:
Step 1, off-line training step:According to the distribution of the priori rule and customer service type of spectrum environment variation, center frequency
Manager is composed using Q study optimization dynamic frequency spectrum deployment strategies, realizes the convergence of dynamic frequency spectrum deployment algorithm;
Step 2, real time spectrum distribution:When service request reaches, center spectrum manager accesses frequency spectrum data library obtains current
Then available spectral slice list carries out dynamic frequency spectrum deployment according to the result of Q study optimizations;
Step 3, on-line tuning stage:By the value of studying history spectrum information and historic demand information obtaining system statistic,
If the value for adding up average satisfaction in system is less than threshold value, start on-line tuning.
3. the centralized method for allocating dynamic frequency spectrums of the spectrum service model according to claim 2 for dynamically distributing recycling,
It is characterized in that, off-line training step described in step 1, specific as follows:
Dynamic frequency spectrum deployment problem is modeled as markoff process, optimization aim is that system adds up average satisfaction maximum:
Wherein, k indicates the number of communication request, ukIt is the satisfaction of k-th of communication request, ak∈ { 0,1 ... n ... ND }, system
In share N number of usable spectrum segment, the maximum possible distribution condition numbers of all spectral slices is D, ak=0 indicates that the request is refused
Absolutely;ak=n indicates that current request is assigned to theA spectral slice;
Spectrum allocation strategy is obtained using off-line training, is as follows:
(1.1) it initializes:Iterations k=1, action select probability vector are setQk(sk,ak),ak
∈ { 0,1 ..., ND }, wherein skFor system mode,sk∈ S, S are the state set of system;
System mode includes two parts,It is usable spectrum segment and its bandwidth, TkIt is the business of current request
Type;akIt is numbered for spectral slice, Qk(sk,ak) it is k moment skA is selected under statekCorresponding Q values when action, Q are action effectiveness
Function, for evaluating the quality for taking some to act in a particular state;
(1.2) when k-th of user asks to reach, system mode sk, user is according to action selection strategySelection acts
ak;
(1.3) according to the action of selection, user obtains a stochastic return rk;
If the type of service of user is rigid business, T is enabledk=1, return value at this time is:
Wherein, bkIt is allocated to the bandwidth of k-th of business, BthBandwidth threshold value;
If the type of service of user is stream business, T is enabledk=2, return value at this time is:
Wherein, bkIt is allocated to the bandwidth of k-th of business, α and β are the constants for determining extent function shape;
If the type of service of user is elastic business, T is enabledk=3, return value at this time is:
Wherein, bkIt is allocated to the bandwidth of k-th of business, BthBandwidth threshold value, γ1It is scale constant;
(1.4) Q values are updated to all states, more new formula is as follows:
Wherein, akIt is learning rate, γ2It is discount factor;
(1.5) to all state update action select probability vectors, more new formula is as follows:
Wherein, γ3> 0 is temperature coefficient, is used for detecting and utilizing during Schistosomiasis control;
(1.6) for all system modes, act in select probability vector that there are one elements to be more than 0.99, cycle terminates, otherwise
Return to step (1.2).
4. the centralized method for allocating dynamic frequency spectrums of the spectrum service model according to claim 2 for dynamically distributing recycling,
It is characterized in that, real time spectrum described in step 2 distributes, it is specific as follows:
(2.1) reporting of user communication request;
(2.2) center spectrum manager accesses database obtains currently available frequency spectrum bands;
(2.3) center spectrum manager carries out real time spectrum distribution, specific implementation according to the result of off-line training or on-line tuning
It is as follows:
When k-th of request reaches, the frequency spectrum state that center spectrum manager accesses frequency spectrum data library obtains isSystem mode is denoted as s at this timek,sk∈ S, S are system
State set;System mode includes two parts,It is usable spectrum segment and its bandwidth, TkIt is current
The type of service of request;
Minimal frequency bandwidth allocation is set as Δ, if user is assigned to n-th of spectral slice, the band which can distribute
Width isIn one kind;
It is set in a certain decision moment, shares N number of usable spectrum segment in system, the maximum bandwidth of n-th of spectral slice isAnd the maximum possible distribution condition number of all spectral slices is D, then for each user, the result of frequency spectrum distribution
Shared ND+1 kinds, wherein containing the case where being rejected;When user asks to reach, center spectrum manager is according to system mode
One action a of selectionk, wherein ak∈ { 0,1 ... n ... ND }, ak=0 indicates that the request is rejected, the bandwidth b being assigned at this timek
=0;ak=n indicates that current request is assigned to theA spectral slice, the bandwidth b being assigned at this timekFor:
Each spectral slice is at best able to accessA user;
It sets the arrival of user and leaves away and all obey Poisson distribution, that is, the time interval for reaching and leaving away obeys exponential distribution, main
The arrival time interval of user is respectively λ with the time interval left awaypAnd μp, secondary user arrival time interval and leave away when
Between interval be respectively λcAnd μc, when+1 request of kth reaches, frequency spectrum state is:
Wherein,
Wherein,
5. the centralized dynamic frequency spectrum deployment side of the spectrum service model according to claim 3 or 4 for dynamically distributing recycling
Method, which is characterized in that the type of service includes rigid business, flows three kinds of business and elastic business, specific as follows:
(1) rigid business
Rigid business is real time business, includes the data transmission of video calling, tele-medicine, security risk prevention and control;If discontented
The basic communication requirement of foot, the business of the type do not allow to access network, and the mathematic(al) representation of the type business satisfaction is as follows
It is shown, when meeting communication requirement, satisfaction 1;Otherwise satisfaction is 0:
Wherein, bkIt is allocated to the bandwidth of k-th of business, BthBandwidth threshold value;
(2) business is flowed
Stream business is also a kind of real time business, including online audio, Online Video, game on line;The type business satisfaction
Mathematic(al) representation is as follows:
Wherein, bkIt is allocated to the bandwidth of k-th of business, α and β are the constants for determining extent function shape;
(3) elastic business
Elastic business is non-real-time service, includes the transmission of data or picture;The mathematic(al) representation of the type business satisfaction is such as
Shown in lower:
Wherein, bkIt is allocated to the bandwidth of k-th of business, BthBandwidth threshold value, γ are scale constants.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810177017.2A CN108391272A (en) | 2018-03-04 | 2018-03-04 | Dynamically distribute the spectrum service model of recycling and centralized method for allocating dynamic frequency spectrums |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810177017.2A CN108391272A (en) | 2018-03-04 | 2018-03-04 | Dynamically distribute the spectrum service model of recycling and centralized method for allocating dynamic frequency spectrums |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108391272A true CN108391272A (en) | 2018-08-10 |
Family
ID=63069724
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810177017.2A Pending CN108391272A (en) | 2018-03-04 | 2018-03-04 | Dynamically distribute the spectrum service model of recycling and centralized method for allocating dynamic frequency spectrums |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108391272A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112804694A (en) * | 2019-11-14 | 2021-05-14 | 中国移动通信集团重庆有限公司 | Method, system, computing device and storage medium for configuring communication network bandwidth |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102724113A (en) * | 2012-05-31 | 2012-10-10 | 北京邮电大学 | Method for reconstructing frequency slot resources |
CN103220751A (en) * | 2013-05-08 | 2013-07-24 | 哈尔滨工业大学 | Heterogeneous network access control method based on Q learning resource allocation strategy |
-
2018
- 2018-03-04 CN CN201810177017.2A patent/CN108391272A/en active Pending
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102724113A (en) * | 2012-05-31 | 2012-10-10 | 北京邮电大学 | Method for reconstructing frequency slot resources |
CN103220751A (en) * | 2013-05-08 | 2013-07-24 | 哈尔滨工业大学 | Heterogeneous network access control method based on Q learning resource allocation strategy |
Non-Patent Citations (1)
Title |
---|
TONG LI等: "Database-based Dynamic Spectrum Assignment with Heterogeneous Requirements in 5G Networks", 《2016 25TH WIRELESS AND OPTICAL COMMUNICATION CONFERENCE》 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112804694A (en) * | 2019-11-14 | 2021-05-14 | 中国移动通信集团重庆有限公司 | Method, system, computing device and storage medium for configuring communication network bandwidth |
CN112804694B (en) * | 2019-11-14 | 2022-12-16 | 中国移动通信集团重庆有限公司 | Method, system, computing device and storage medium for configuring communication network bandwidth |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109947545B (en) | Task unloading and migration decision method based on user mobility | |
CN111277437B (en) | Network slice resource allocation method for smart power grid | |
CN111246586B (en) | Method and system for distributing smart grid resources based on genetic algorithm | |
Chen et al. | Qoe-aware decentralized task offloading and resource allocation for end-edge-cloud systems: A game-theoretical approach | |
CN108834080B (en) | Distributed cache and user association method based on multicast technology in heterogeneous network | |
Wu et al. | An efficient offloading algorithm based on support vector machine for mobile edge computing in vehicular networks | |
CN106100907B (en) | A kind of MEC server selection method based on fairness | |
CN109639377A (en) | Dynamic spectrum resource management method based on deeply study | |
CN111866601B (en) | Cooperative game-based video code rate decision method in mobile marginal scene | |
CN110719641B (en) | User unloading and resource allocation joint optimization method in edge computing | |
CN111711666B (en) | Internet of vehicles cloud computing resource optimization method based on reinforcement learning | |
CN109982434B (en) | Wireless resource scheduling integrated intelligent control system and method and wireless communication system | |
CN101945369A (en) | Auction and satisfaction model-based dynamic frequency spectrum allocation method | |
CN110233755A (en) | The computing resource and frequency spectrum resource allocation method that mist calculates in a kind of Internet of Things | |
CN110717300A (en) | Edge calculation task allocation method for real-time online monitoring service of power internet of things | |
CN110191489B (en) | Resource allocation method and device based on reinforcement learning in ultra-dense network | |
CN109905864A (en) | A kind of cross-layer Resource Allocation Formula towards electric power Internet of Things | |
CN107105455A (en) | It is a kind of that load-balancing method is accessed based on the user perceived from backhaul | |
CN108449149B (en) | Energy acquisition small base station resource allocation method based on matching game | |
CN106792995A (en) | The user access method of content low time delay transmission is ensured in a kind of following 5G networks | |
Shome et al. | Deep Q-learning for 5G network slicing with diverse resource stipulations and dynamic data traffic | |
CN114885422A (en) | Dynamic edge computing unloading method based on hybrid access mode in ultra-dense network | |
Chen et al. | DDPG-based radio resource management for user interactive mobile edge networks | |
CN108391272A (en) | Dynamically distribute the spectrum service model of recycling and centralized method for allocating dynamic frequency spectrums | |
CN110139281A (en) | A method of the full duplex D2D sub-clustering resource allocation based on K-means |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180810 |
|
RJ01 | Rejection of invention patent application after publication |