CN117676596A - Time-frequency resource dynamic allocation method based on task driving - Google Patents
Time-frequency resource dynamic allocation method based on task driving Download PDFInfo
- Publication number
- CN117676596A CN117676596A CN202311558633.XA CN202311558633A CN117676596A CN 117676596 A CN117676596 A CN 117676596A CN 202311558633 A CN202311558633 A CN 202311558633A CN 117676596 A CN117676596 A CN 117676596A
- Authority
- CN
- China
- Prior art keywords
- node
- traffic
- time
- network
- nodes
- 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
- 238000000034 method Methods 0.000 title claims abstract description 41
- 238000007726 management method Methods 0.000 claims abstract description 53
- 238000013468 resource allocation Methods 0.000 claims abstract description 26
- 238000004891 communication Methods 0.000 claims abstract description 13
- 230000003993 interaction Effects 0.000 claims abstract description 5
- 230000005540 biological transmission Effects 0.000 claims description 38
- 238000013528 artificial neural network Methods 0.000 claims description 14
- 238000010606 normalization Methods 0.000 claims description 12
- 238000012549 training Methods 0.000 claims description 12
- 238000010276 construction Methods 0.000 claims description 10
- 239000011159 matrix material Substances 0.000 claims description 9
- 238000012545 processing Methods 0.000 claims description 8
- 230000008569 process Effects 0.000 claims description 4
- 238000001228 spectrum Methods 0.000 claims description 4
- 238000001514 detection method Methods 0.000 claims description 3
- 230000009191 jumping Effects 0.000 claims description 3
- 238000012546 transfer Methods 0.000 claims description 3
- 230000002265 prevention Effects 0.000 claims description 2
- 238000012163 sequencing technique Methods 0.000 claims description 2
- 230000008447 perception Effects 0.000 abstract description 5
- 238000004364 calculation method Methods 0.000 abstract description 4
- 238000010586 diagram Methods 0.000 description 4
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 230000006870 function Effects 0.000 description 2
- 230000006855 networking Effects 0.000 description 2
- 101000931590 Homo sapiens Prostaglandin F2 receptor negative regulator Proteins 0.000 description 1
- 102100020864 Prostaglandin F2 receptor negative regulator Human genes 0.000 description 1
- 230000006978 adaptation Effects 0.000 description 1
- 230000003042 antagnostic effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000013480 data collection Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- UREBDLICKHMUKA-CXSFZGCWSA-N dexamethasone Chemical compound C1CC2=CC(=O)C=C[C@]2(C)[C@]2(F)[C@@H]1[C@@H]1C[C@@H](C)[C@@](C(=O)CO)(O)[C@@]1(C)C[C@@H]2O UREBDLICKHMUKA-CXSFZGCWSA-N 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 238000013439 planning Methods 0.000 description 1
Abstract
The invention discloses a time-frequency resource dynamic allocation method based on task driving, which comprises the following steps: constructing a hierarchical clustering network; wherein the hierarchical clustering network is composed of M subnets, each subnet comprising N nodes: 1 network management node, K gateway nodes and N-1-K common nodes; each node has a unique identity; the data interaction is carried out among the sub-networks through the gateway node; the network management node is used for carrying out time-frequency resource allocation; judging the current node type according to the node identity; and when the current node is determined to be the network management node, calling a time-frequency resource dynamic allocation module to allocate the time-frequency resource. The method constructs a new network resource dynamic allocation scheme for future combat scenes, solves the contradiction between differentiated service requirements in a high-dynamic self-organizing combat environment and limited multi-dimensional resources such as communication, calculation, perception and the like in a military network, and realizes the self-adaptive matching of the network multi-dimensional resources and various service requirements.
Description
Technical Field
The invention belongs to the technical field of wireless networks and communication, and particularly relates to a time-frequency resource dynamic allocation method based on task driving.
Background
In future integrated, informationized communication scenarios, networks will have highly dynamic, strong antagonistic, self-organizing features. In this scenario, the weapon equipment and platform are no longer an independent battle individual, and the various platforms and battlefield elements need to be integrated into an organic whole through the communication network to jointly perform and complete various types of tasks.
In a certain communication scene, in order to ensure the manageability and controllability of a transmission process, particularly transmission delay, a TDMA mode is first adopted abroad to support network access and data transmission of multiple users in a wireless channel. The node obtains network transmission resources which are determined by the fact that the node can be allocated to the transmission time slot in the transmission process, and the optimal allocation of the transmission resources in the network is equivalent to the dynamic optimal allocation of the transmission time slot. In a tactical data Link11, the network adopts a master-slave mode to group networks, the nodes acquire available transmission time slots in a polling mode, and the use efficiency of transmission resources is low. In Link16, although multi-hop flexible networking between nodes can be realized, the time slot resource allocation between nodes often adopts a relatively fixed allocation mode, and is difficult to adaptively and flexibly adjust according to the change of business service requirements.
In order to meet the QoS service requirements of differentiated transport services, it is very necessary to dynamically, optimally, and on-demand configuration management of time slot resources. The current time slot allocation algorithms can be broadly classified into 3 categories, a fixed time slot allocation algorithm, a dynamic time slot allocation algorithm, and a hybrid time slot allocation algorithm combining fixed and dynamic. The TDMA protocols based on the dynamic allocation algorithm can be divided into centralized and distributed according to the implementation mode of the algorithm; the distributed dynamic TDMA protocol can be further subdivided into topology dependent and topology transparent 2 types depending on whether topology information is required at the time of slot allocation.
Chenxi Zhu and m.scott Corson published an FPRP protocol for TDMA ad hoc networks that simultaneously solved the channel access and broadcast scheduling problems for nodes that guaranteed collision-free broadcast slot allocation through a contention-based five-step reservation mechanism. Rhee et al propose a distributed random time slot scheduling algorithm (DRAND) which uses a completely random coin-thrown time slot reservation algorithm to limit the probability of each node sending a reservation request at each reservation time slot, thereby reducing the probability of nodes sending requests at the same reservation time slot at the same time in the two-hop neighbor range to some extent.
Many contradictions exist between differentiated service requirements in a high-dynamic self-organizing battlefield environment and limited communication, calculation, perception and other multidimensional resources in a military network, and how to realize the self-adaptive matching of the network multidimensional resources and various service requirements is one of the problems to be solved urgently by those skilled in the art
Disclosure of Invention
The technical solution of the invention is as follows: the task-driven time-frequency resource dynamic allocation method has the advantages that the shortcomings of the prior art are overcome, a new network resource dynamic allocation scheme oriented to future combat scenes is constructed, contradiction between differentiated service requirements in a high-dynamic self-organizing battlefield environment and limited communication, calculation, perception and other multidimensional resources in a military network is solved, self-adaptive matching of network multidimensional resources and various service requirements is realized, and the advantages of rapid adjustment, high-efficiency transmission and the like are achieved.
In order to solve the technical problems, the invention discloses a time-frequency resource dynamic allocation method based on task driving, which comprises the following steps:
constructing a hierarchical clustering network; wherein the hierarchical clustering network is composed of M subnets, each subnet comprising N nodes: 1 network management node, K gateway nodes and N-1-K common nodes; each node has a unique identity; the data interaction is carried out among the sub-networks through the gateway node; the network management node is used for carrying out time-frequency resource allocation; m is more than or equal to 1, K is more than or equal to 1, and N is more than or equal to 3;
judging the current node type according to the node identity;
and when the current node is determined to be the network management node, calling a time-frequency resource dynamic allocation module to allocate the time-frequency resource.
In the above task-driven time-frequency resource dynamic allocation method, the time-frequency resource dynamic allocation module includes: the system comprises an LSTM prediction module, a time slot resource allocation module, a general computing resource management module and a frequency hopping pattern allocation module.
In the above method for dynamically allocating time-frequency resources based on task driving, when determining that the current node is a network management node, invoking a time-frequency resource dynamic allocation module to allocate time-frequency resources comprises:
when the current node is determined to be a network management node, a general sense computing resource management module is called, and local link flow characteristics of other nodes are collected and used as historical service flow; wherein the historical traffic characteristics are characteristics of traffic transmitted in the past;
calling an LSTM prediction module, and predicting the local link flow characteristics in the next period according to the local link flow characteristics of other nodes collected by the general computing resource management module to obtain a prediction result;
a time slot resource allocation module is called, and time slot resource allocation in the sub-network is carried out according to the prediction result output by the LSTM prediction module and the link construction requirement information issued by the network layer; the link construction requirement information issued by the network layer is the characteristic of the service to be transmitted in the next period, namely future service flow;
and calling a frequency hopping pattern distribution module, and distributing the frequency hopping patterns in the sub-network according to the spectrum interference results of all the common nodes in the whole network, which are summarized by the network management nodes.
In the task-driven time-frequency resource dynamic allocation method, the historical traffic and the future traffic each comprise: source node, destination node, service priority and traffic size; the formats are as follows: < source node, destination node, traffic priority, traffic size >.
In the time-frequency resource dynamic allocation method based on task driving, a plurality of task phases exist in the flying process, and each task phase is regarded as a period; wherein, the task stage comprises: a navigation stage, a collision prevention stage, an enemy attack stage and a detection stage.
In the task-driven time-frequency resource dynamic allocation method, when the LSTM prediction module predicts the local link traffic characteristics in the next period, the method includes:
the historical traffic of the link e covered by the t-time slice s is expressed as { traffic } e,s (t) -a }; for { traffic } es (t) } performing normalization processing to obtain normalization result { traffic }, and es (t)} tr :
where t=1, 2,..h, h represents the time length of the current period;represents { traffic } e,s An average value of (t);
then, time window T window The normalized result of future traffic within the length is expressed as { traffic } e,s (j)} tr The method comprises the following steps:
{traffic e,s (j)} tr ={traffic e,s (j),traffic e,s (j+T window ),...,traffic e,s (j+D·T window )} tr
where j=1, 2,.. window ;
Will { traffic } e,s (j)} tr Taking the previous step D of the (1) as an input value of the LSTM neural network, taking the step D+1 as a comparison value of the minimum mean square error, and dividing the data into T window Obtaining training data;
training the LSTM neural network according to the training data to obtain a prediction model;
will transfer e,s (j) Takes the D-step data of the step (C) as the input of a prediction model, and outputs a predicted value traffic 'through the prediction model' e,s (j+(D+1)·T window ) The method comprises the steps of carrying out a first treatment on the surface of the Taking time window T window Predicted values within the length to obtain the predicted sequence { traffic' e,s (1+(D+1)·T window ),...,traffic′ e,s (T window +(D+1)·T window )};
Performing inverse normalization processing on the predicted sequence to obtain predicted traffic in the next period;
repeating the steps to obtain the predicted traffic flow in the next period of all links in the network.
In the task-driving-based time-frequency resource dynamic allocation method, the Loss function Loss of the LSTM neural network is as follows:
wherein o is j Representing the output value of the j-th pair of training data in the LSTM neural network.
In the above method for dynamically allocating time-frequency resources based on task driving, the time slot resource allocation module includes: and filling the micro time slots required by service transmission among nodes in a frame completely according to the prediction result output by the LSTM prediction module and the link construction requirement information.
In the above task-driven time-frequency resource dynamic allocation method, when the frequency hopping pattern allocation module allocates frequency hopping patterns in the subnet, the method includes:
according to the irreducible polynomial x of stage 9 9 +x 4 +1 generates an m-sequence f_p and performs mod (f_num) operation on the m-sequence f_p; where mod represents the remainder, and f_num represents the total number of bins;
carrying out narrow-band judgment on frequency points used by N nodes at any moment;
if |f_ (p+1) -f_p| <15, f_ (p+1) = [ f_ (p+1) +30] mod (500), obtaining the frequency hopping pattern of N nodes in the sub-network; where p=1, 2, …, N.
The task-driving-based time-frequency resource dynamic allocation method further comprises the following steps: calling a general sense computing resource management module to manage resources:
ordering the node service transmission demand from small to large, and dividing the busyness of the node into 4 grades according to the ordering size: the node of N/4 before sequencing is a first gear, the node of N/4 to N/2 is a second gear, the node of N/2 to 3N/4 is a third gear, and the node of 3N/4 to N is a fourth gear;
initialization matrix g=adjacency matrix x p,q ]Wherein x is p,q =1 means that node p communicates with node q, x p,q =0 indicates that node p is not in communication with node q; q+.p, q=1, 2, …, N;
initializing all nodes to let x p =0,n p =0; wherein x is p =1 means node i is selected as the sensing node, x p =0 means that the node is not selected; n is n p Indicating that the sense node i will be separated by n p Transmitting primary sensing data by a frame;
the following operations are sequentially carried out on the nodes of the first gear, the second gear, the third gear and the fourth gear:
a) For the same gear internal node, according to the degree D i From big to small, the node with the largest degree is preferentially selected;
b) If the current node is a common node, selecting the current node as a sensing node to collect the burst traffic flow characteristics, namely updating the x of the current node i =1; and then updates matrix G: setting the row and column of the current node to 0;
c) Judging whether all elements in G are 0, if so, ending; if not, continuing to operate the a) to b) on other nodes in the same gear;
d) If all nodes in the same gear are traversed, jumping to the next gear, and performing the same operation until all elements in G are 0; if the selected sensing node is a node in the first or second gear, the sensing data transmission frequency is 2 frames transmitted once, namely, n of the node is updated p =2; if the selected sensing node is a node in the third gear or the fourth gear, the sensing data transmission frequency is 4 frames to be transmitted once, namely, the node n is updated p =4;
f) Outputting all the selected sensing nodes and the transmission frequency of sensing data of each selected sensing node, namely outputting x p And n p 。
The invention has the following advantages:
(1) The invention discloses a time-frequency resource dynamic allocation method based on task driving, which provides a clustering hierarchical network topology architecture, wherein network management nodes are adopted in clusters to manage the whole network, and communication is carried out among the clusters through the network management nodes, so that the large-scale aircraft ad hoc network is realized.
(2) The invention discloses a time-frequency resource dynamic allocation method based on task driving, which adopts an elastic network protocol system, wherein a transmission side is divided into a physical layer, a link layer and a network layer, so that the time-frequency resource can be dynamically allocated according to task requirements, intelligent adaptation of the task is realized, communication delay can be reduced, and communication efficiency is improved.
Drawings
FIG. 1 is a schematic diagram of a hierarchical clustering network in accordance with an embodiment of the present invention;
FIG. 2 is a schematic diagram of an overall scheme of time-frequency dynamic allocation according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of an LSTM neural network in accordance with an embodiment of the present invention;
fig. 4 is a schematic diagram of a slot structure in an embodiment of the present invention.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the present invention more apparent, the embodiments of the present invention disclosed herein will be described in further detail with reference to the accompanying drawings.
Aiming at the task driving adjustment time-frequency resource requirement in a large-scale wireless networking scene, the invention provides a time-frequency resource dynamic allocation method based on task driving, which comprises the following steps:
and 1, constructing a hierarchical clustering network.
In this embodiment, as shown in fig. 1, the network architecture of the hierarchical clustering network is composed of M subnets, where each subnet includes N nodes: 1 network management node, K gateway nodes and N-1-K common nodes. Each node has a unique identity. The data interaction is carried out among the sub-networks through the gateway node; the network management node is used for managing node resources in the sub-network and distributing time-frequency resources of the nodes in the sub-network. Wherein M is more than or equal to 1, K is more than or equal to 1, and N is more than or equal to 3.
And 2, judging the current node type according to the node identity.
In this embodiment, first, the type of the current node needs to be determined, that is, whether the current node is a gateway node, a network management node, or a common node is determined. The functional modules which are required to be responsible for different nodes have variability, only the network management node is responsible for time-frequency resource allocation in the subnetwork, and only the gateway node can communicate with other subnetwork nodes.
And step 3, when the current node is determined to be the network management node, calling a time-frequency resource dynamic allocation module to allocate the time-frequency resources.
In this embodiment, when it is determined that the current node is a network management node, the time-frequency resource dynamic allocation module is invoked to perform time-frequency resource allocation, as shown in fig. 2, where the time-frequency resource dynamic allocation module includes: the system comprises an LSTM prediction module, a time slot resource allocation module, a general computing resource management module and a frequency hopping pattern allocation module. The LSTM prediction module, the time slot resource allocation module, the general sense computing resource management module and the frequency hopping pattern allocation module are positioned on a link layer, so that interaction with a network layer and a physical layer is realized.
Since there are multiple mission phases in the flight, such as a navigation phase, a collision avoidance phase, an enemy attack phase, and a detection phase, each mission phase is considered as one cycle (note that the duration of each cycle is not fixed or predictable). And constructing the demand information through the received link of the next period, and generating a time-frequency resource allocation table of a traffic channel in the subnet based on the historical traffic flow and the perceived data transmission demand provided by the general computation resource management module by the network management node.
The LSTM prediction module is used for predicting the point-to-point burst service in the next period.
The time slot resource allocation module is used for allocating time slot resources of a transmission channel in the subnet based on the link construction demand information of the next period and the point-to-point burst service flow predicted by the LSTM module.
The frequency hopping pattern distribution module collects the frequency spectrum result perceived by each node, and the network management node redistributes the frequency hopping patterns to obtain a frequency hopping pattern set of all nodes in the sub-network, each node corresponds to a unique frequency hopping pattern, and each frequency hopping pattern is orthogonal as far as possible, so that the maximum transmission efficiency is ensured.
The general computing resource management module is used for collecting the characteristic data of the burst traffic flow in the next task stage and is used as the input of the LSTM prediction module in the next task stage. In the task adjustment stage, the network management node performs the following steps according to the point-to-point service transmission requirement, the network topology structure, the node available computing resources, the node position information and the like in the next stage: 1) The selection of sensing nodes, namely which nodes in the subnet collect the burst service flow characteristic data of the next stage and transmit the burst service flow characteristic data to the network management node; 2) Sensing the transmission frequency of data, namely sensing nodes need to transmit the currently collected burst service flow characteristic data once every several frames; 3) And the perception data calculation indication is that whether the node needs to preprocess the data after collecting the perception data or not is transmitted to the network management node. Further, the general sense computing resource management module comprises two parts, namely a general sense module and a general sense module, wherein the general sense module is used for determining the selection and the data transmission frequency of the sensing node, and the general sense module is used for making a decision on whether the sensing data needs to be subjected to local operation or not based on the result of the general sense module, the available computing resources of the node and the transmission capacity of the node. Further, the general sense computing resource management module performs burst service flow characteristic data collection planning in the next task stage according to the point-to-point transmission requirement of the next stage, generates the data transmission requirement of the next stage, and transmits the data transmission requirement to the time slot resource allocation module; and the time slot resource allocation module performs dynamic allocation of time slot resources according to the transmission requirement of the next stage and the point-to-point burst service information predicted by the LSTM module to form a time slot table of a new stage.
In addition, when it is determined that the current node is not the network management node, the flow is ended.
In this embodiment, when it is determined that the current node is a network management node, a specific flow of calling the time-frequency resource dynamic allocation module to perform time-frequency resource allocation is as follows:
and step 31, when the current node is determined to be the network management node, calling a general computing resource management module, and collecting the local link flow characteristics of other nodes as historical service flow. Wherein, the historical traffic flow characteristics refer to the characteristics of the traffic transmitted in the past, and include: information such as source node, destination node, service priority, traffic volume, etc., and the format is generally: < source node, destination node, traffic priority, traffic size >.
And step 32, calling an LSTM prediction module, and predicting the local link flow characteristics in the next period according to the local link flow characteristics of other nodes collected by the general computing resource management module to obtain a prediction result.
In this embodiment, the prediction flow of the LSTM prediction module is as follows:
first, data processing is performed. Data processing refers to a process of performing normalization to convert input data (traffic) into a series of values acceptable to LSTM neural networks. The LSTM neural network model is shown in fig. 3. The specific process of normalization is as follows:
the historical traffic of the link e covered by the t-time slice s is expressed as { traffic } e,s (t) }; for { traffic } e,s (t) } performing normalization processing to obtain normalization result { traffic }, and e,s (t)} tr :
where t=1, 2,..h, h represents the time length of the current period;represents { traffic } e,s (t) }.
Further, due to the time window T window Future traffic within length and d·t thereof window The previous D-step traffic within the length is related, so the normalization of future traffic of link e covered by slice s can be expressed as { traffic e,s (j)} tr The following steps are:
{traffic e,s (j)} tr ={traffic e,s (j),traffic e,s (j+T window ),...,traffic e,s (j+D·T window )} tr
where j=1, 2,.. window 。
Further, will { traffic } e,s (j)} tr The former step D of the (1) is used as an input value of the LSTM neural network, and the step D+1 is used as a comparison value of the minimum mean square error, so that the data is divided into T window For this, such data together form a batch of training data.
According to the training data, training the LSTM neural network to obtain a prediction model. The Loss function Loss of the LSTM neural network is as follows:
wherein o is j Representing the output value of the j-th pair of training data in the LSTM neural network.
Will transfer e,s (j) Takes the D-step data of the step (C) as the input of a prediction model, and outputs a predicted value traffic 'through the prediction model' e,s (j+(D+1)·T window ) The method comprises the steps of carrying out a first treatment on the surface of the Taking time window T window Predicted values within the length to obtain the predicted sequence { traffic' e,s (1+(D+1)·T window ),...,traffic′ e,s (T window +(D+1)·T window )}。
Further, the predicted sequence is subjected to inverse normalization processing to obtain predicted traffic in the next period.
Finally, the above steps are repeated to obtain the predicted traffic in the next period of all links in the network.
And step 33, calling a time slot resource allocation module, and allocating time slot resources in the sub-network according to the prediction result output by the LSTM prediction module and the link construction requirement information issued by the network layer.
In this embodiment, the link construction requirement information issued by the network layer is a feature of a service to be transmitted in a next period, that is, a future service flow, and includes: information such as source node, destination node, service priority, traffic volume, etc., and the format is generally: < source node, destination node, traffic priority, traffic size >.
Preferably, the specific timeslot resource allocation flow of the timeslot resource allocation module is as follows: 1) X output by general sense computing resource management module p And n p The local link flow characteristics of the corresponding node are sent to the network management node; the prediction result output by the LSTM prediction module and the link construction requirement information are input to a time slot resource allocation module, micro time slots required by service transmission between nodes in a frame are filled completely according to a time slot structure shown in fig. 4, and the number of the specific micro time slots contained in the frame is determined by specific use conditions.
And step 34, calling a frequency hopping pattern distribution module, and distributing frequency hopping patterns in the sub-network according to the spectrum interference results of all the common nodes in the whole network, which are summarized by the network management nodes.
In this embodiment, a specific hopping pattern allocation flow of the hopping pattern allocation module is as follows:
according to the irreducible polynomial x of stage 9 9 +x 4 +1 generates an m-sequence f_p and performs mod (f_num) operation on the m-sequence f_p; where mod represents the remainder, and f_num represents the total number of bins. Where p=1, 2, …, N.
And carrying out narrow-band judgment on frequency points used by N nodes at any time, and if |f_ (p+1) -f_p| <15, f_ (p+1) = [ f_ (p+1) +30] mod (500), obtaining frequency hopping patterns of N nodes in the sub-network.
In this embodiment, resource management may also be performed by calling the general computing resource management module:
a) Ordering the node service transmission demand from small to large, and dividing the busyness of the node into 4 grades according to the ordering size: the node of N/4 (containing N/4) before the sorting is the first gear, the node of N/4 to N/2 (containing N/2) is the second gear, the node of N/2 to 3N/4 (containing 3N/4) is the third gear, and the node of 3N/4 to N (containing N) is the fourth gear.
b) Initializing matrix G=adjacency matrix [ x ] p,q ]. Wherein x is p,q =1 means that node p communicates with node q, x p,q =0 indicates that node p is not in communication with node q; q+.p, q=1, 2, …, N.
c) Initializing all nodes to let x p =0,n p =0. Wherein x is p =1 means node i is selected as the sensing node, x p =0 means that the node is not selected; n is n p Indicating that the sense node i will be separated by n p The frame transmits the sensing data once.
d) The following operations are sequentially carried out on the nodes of the first gear, the second gear, the third gear and the fourth gear:
d1 For the same gear internal node, according to degree D i From the order of big to small, the node with the largest degree is preferentially selected.
d2 If the current node is a common node, selecting the current node as a sensing node to collect the burst traffic characteristics, namely updating the x of the current node i =1; and then updates matrix G: the current node is set to 0 in both the row and column.
d3 Judging whether all elements in G are 0, if so, ending; if not, the operations d 1) to d 2) are continued for other nodes of the same gear.
d4 If all nodes in the same gear are traversed, jumping to the next gear and performing the same operation until all elements in G are 0. If the selected sensing node is a node in the first or second gear, the sensing data transmission frequency is 2 frames transmitted once, namely, n of the node is updated p =2; if the selected sensing node is a node in the third gear or the fourth gear, the sensing data transmission frequency is 4 frames to be transmitted once, namely, the node n is updated p =4。
d5 Outputting all selected sensing nodes and the transmission frequency of sensing data of each selected sensing node, namely outputting x p And n p 。
Although the present invention has been described in terms of the preferred embodiments, it is not intended to be limited to the embodiments, and any person skilled in the art can make any possible variations and modifications to the technical solution of the present invention by using the methods and technical matters disclosed above without departing from the spirit and scope of the present invention, so any simple modifications, equivalent variations and modifications to the embodiments described above according to the technical matters of the present invention are within the scope of the technical matters of the present invention.
What is not described in detail in the present specification belongs to the known technology of those skilled in the art.
Claims (10)
1. The time-frequency resource dynamic allocation method based on task driving is characterized by comprising the following steps:
constructing a hierarchical clustering network; wherein the hierarchical clustering network is composed of M subnets, each subnet comprising N nodes: 1 network management node, K gateway nodes and N-1-K common nodes; each node has a unique identity; the data interaction is carried out among the sub-networks through the gateway node; the network management node is used for carrying out time-frequency resource allocation; m is more than or equal to 1, K is more than or equal to 1, and N is more than or equal to 3;
judging the current node type according to the node identity;
and when the current node is determined to be the network management node, calling a time-frequency resource dynamic allocation module to allocate the time-frequency resource.
2. The task-driven time-frequency resource dynamic allocation method according to claim 1, wherein the time-frequency resource dynamic allocation module comprises: the system comprises an LSTM prediction module, a time slot resource allocation module, a general computing resource management module and a frequency hopping pattern allocation module.
3. The task-driven time-frequency resource dynamic allocation method according to claim 2, wherein when determining that the current node is a network management node, invoking the time-frequency resource dynamic allocation module to perform time-frequency resource allocation comprises:
when the current node is determined to be a network management node, a general sense computing resource management module is called, and local link flow characteristics of other nodes are collected and used as historical service flow; wherein the historical traffic characteristics are characteristics of traffic transmitted in the past;
calling an LSTM prediction module, and predicting the local link flow characteristics in the next period according to the local link flow characteristics of other nodes collected by the general computing resource management module to obtain a prediction result;
a time slot resource allocation module is called, and time slot resource allocation in the sub-network is carried out according to the prediction result output by the LSTM prediction module and the link construction requirement information issued by the network layer; the link construction requirement information issued by the network layer is the characteristic of the service to be transmitted in the next period, namely future service flow;
and calling a frequency hopping pattern distribution module, and distributing the frequency hopping patterns in the sub-network according to the spectrum interference results of all the common nodes in the whole network, which are summarized by the network management nodes.
4. A task-driven time-frequency resource dynamic allocation method according to claim 3, wherein the historical traffic and the future traffic each comprise: source node, destination node, service priority and traffic size; the formats are as follows: < source node, destination node, traffic priority, traffic size >.
5. A task-driven time-frequency resource dynamic allocation method according to claim 3, wherein the flight process has a plurality of task phases, each task phase being regarded as a cycle; wherein, the task stage comprises: a navigation stage, a collision prevention stage, an enemy attack stage and a detection stage.
6. The task-driven time-frequency resource dynamic allocation method according to claim 3, wherein the LSTM prediction module, when predicting the local link traffic characteristics in the next period, includes:
the historical traffic of the link e covered by the t-time slice s is expressed as { traffic } e,s (t) }; for { traffic } e,s (t) } performing normalization processing to obtain normalization result { traffic }, and e,s (t)} tr :
where t=1, 2,..h, h represents the time length of the current period;represents { traffic } e,s An average value of (t);
then, time window T window The normalized result of future traffic within the length is expressed as { traffic } e,s (j)} tr The method comprises the following steps:
{traffic e,s (j)} tr ={traffic e,s (j),traffic e,s (j+T window ),...,traffic e,s (j+D·T window )} tr
where j=1, 2,.. window ;
Will { traffic } e,s (j)} tr Taking the previous step D of the (1) as an input value of the LSTM neural network, taking the step D+1 as a comparison value of the minimum mean square error, and dividing the data into T window Obtaining training data;
training the LSTM neural network according to the training data to obtain a prediction model;
will transfer e,s (j) Takes the D-step data of the step (C) as the input of a prediction model, and outputs a predicted value traffic 'through the prediction model' e,s (j+(D+1)·T window ) The method comprises the steps of carrying out a first treatment on the surface of the Taking time window T window Predicted values within the length to obtain the predicted sequence { traffic' e,s (1+(D+1)·T window ),...,traffic′ e,s (T window +(D+1)·T window )};
Performing inverse normalization processing on the predicted sequence to obtain predicted traffic in the next period;
repeating the steps to obtain the predicted traffic flow in the next period of all links in the network.
7. The task-driven time-frequency resource dynamic allocation method according to claim 6, wherein the Loss function Loss of the LSTM neural network is:
wherein o is j Representing the output value of the j-th pair of training data in the LSTM neural network.
8. A method for dynamically allocating time-frequency resources based on task driving as recited in claim 3, wherein the time slot resource allocation module, when allocating time slot resources in the subnet, comprises: and filling the micro time slots required by service transmission among nodes in a frame completely according to the prediction result output by the LSTM prediction module and the link construction requirement information.
9. The task-driven time-frequency resource dynamic allocation method according to claim 3, wherein the frequency hopping pattern allocation module, when performing intra-subnet frequency hopping pattern allocation, comprises:
according to the irreducible polynomial x of stage 9 9 +x 4 +1 generates an m-sequence f_p and performs mod (f_num) operation on the m-sequence f_p; where mod represents the remainder, and f_num represents the total number of bins;
carrying out narrow-band judgment on frequency points used by N nodes at any moment;
if |f_ (p+1) -f_p| <15, f_ (p+1) = [ f_ (p+1) +30] mod (500), obtaining the frequency hopping pattern of N nodes in the sub-network; where p=1, 2, …, N.
10. A task-driven time-frequency resource dynamic allocation method according to claim 3, further comprising: calling a general sense computing resource management module to manage resources:
ordering the node service transmission demand from small to large, and dividing the busyness of the node into 4 grades according to the ordering size: the node of N/4 before sequencing is a first gear, the node of N/4 to N/2 is a second gear, the node of N/2 to 3N/4 is a third gear, and the node of 3N/4 to N is a fourth gear;
initialization matrix g=adjacency matrix x p,q ]Wherein x is p,q =1 means that node p communicates with node q, x p,q =0 indicates that node p is not in communication with node q; q+.p, q=1, 2, …, N;
initializing all nodes to let x p =0,n p =0; wherein x is p =1 means node i is selected as the sensing node, x p =0 means that the node is not selected; n is n p Indicating that the sense node i will be separated by n p Transmitting primary sensing data by a frame;
the following operations are sequentially carried out on the nodes of the first gear, the second gear, the third gear and the fourth gear:
a) For the same gear internal node, according to the degree D i From big to small, the node with the largest degree is preferentially selected;
b) If the current node is a common node, selecting the current node as a sensing node to collect the burst traffic flow characteristics, namely updating the x of the current node i =1; and then updates matrix G: setting the row and column of the current node to 0;
c) Judging whether all elements in G are 0, if so, ending; if not, continuing to operate the a) to b) on other nodes in the same gear;
d) If all nodes in the same gear are traversed, jumping to the next gear, and performing the same operation until all elements in G are 0; if the selected sensing node is a node in the first or second gear, the sensing data transmission frequency is 2 frames transmitted once, namely, n of the node is updated p =2; if the selected sensing node is a node in the third gear or the fourth gear, the sensing data transmission frequency is 4 frames to be transmitted once, namely, the node n is updated p =4;
f) Outputting all selected sensing nodes and sensing nodesNode perceives the frequency of transmission of data, i.e. output x p And n p 。
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202311558633.XA CN117676596A (en) | 2023-11-21 | 2023-11-21 | Time-frequency resource dynamic allocation method based on task driving |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202311558633.XA CN117676596A (en) | 2023-11-21 | 2023-11-21 | Time-frequency resource dynamic allocation method based on task driving |
Publications (1)
Publication Number | Publication Date |
---|---|
CN117676596A true CN117676596A (en) | 2024-03-08 |
Family
ID=90072500
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202311558633.XA Pending CN117676596A (en) | 2023-11-21 | 2023-11-21 | Time-frequency resource dynamic allocation method based on task driving |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN117676596A (en) |
-
2023
- 2023-11-21 CN CN202311558633.XA patent/CN117676596A/en active Pending
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US7046639B2 (en) | System and method for ad hoc network access employing the distributed election of a shared transmission schedule | |
CN108235437B (en) | Wireless mobile self-organizing network channel time slot scheduling and distributing method | |
CN103298137B (en) | A kind of underwater sensor network media access control (MAC) method applicable based on state | |
CN111818652B (en) | Channel resource allocation method of multi-transceiver multi-channel wireless Ad Hoc network | |
CN112584347A (en) | UAV heterogeneous network multi-dimensional resource dynamic management method | |
CN114339660B (en) | Unmanned aerial vehicle cluster random access method | |
CN101651969A (en) | Implement method of power control cognition MAC protocol based on time division multiple access (TDMA) | |
CN111601398B (en) | Ad hoc network medium access control method based on reinforcement learning | |
Ahmed et al. | A QoS-aware MAC protocol for IEEE 802.11 ah-based Internet of Things | |
Bai et al. | An adaptive grouping scheme in ultra-dense IEEE 802.11 ax network using buffer state report based two-stage mechanism | |
CN111328052A (en) | Channel resource allocation method in high-density wireless network | |
CN107659940B (en) | Multi-priority multi-channel media access control method based on channel perception | |
CN111683384B (en) | Network optimization method for realizing dynamic weighting of communication link by artificial intelligence | |
CN117676596A (en) | Time-frequency resource dynamic allocation method based on task driving | |
Ranganathan et al. | Enhancing the selection of backoff interval using fuzzy logic over wireless ad hoc networks | |
CN109152079B (en) | Time slot scheduling and channel allocation method for TSCH (three-channel switched capacitor) network | |
CN112702793B (en) | Method for solving conflict-free node set of wireless mesh network | |
KR20060130496A (en) | Non-agile channelelization method for multi-channel mac | |
Gao et al. | Cluster-based interference-free MAC protocol with load aware in software defined VANET | |
CN102427592A (en) | Centralized link scheduling method for multi-channel wireless network | |
CN112839337B (en) | Method for coordinating interference between ultra-dense micro base stations based on divide-and-conquer recursion strategy | |
Mišić et al. | Cognitive MAC protocol with transmission tax: Dynamically adjusting sensing and data performance | |
JP3815668B2 (en) | Wireless channel assignment method in mesh wireless network | |
CN105933977A (en) | Service priority-based wireless multimedia sensor network resource scheduling algorithm | |
CN112929970B (en) | Self-organizing network time slot distribution method based on priority |
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 |