CN113938976A - Internet of things passive sensing routing algorithm for intelligent utility planning - Google Patents

Internet of things passive sensing routing algorithm for intelligent utility planning Download PDF

Info

Publication number
CN113938976A
CN113938976A CN202111221256.1A CN202111221256A CN113938976A CN 113938976 A CN113938976 A CN 113938976A CN 202111221256 A CN202111221256 A CN 202111221256A CN 113938976 A CN113938976 A CN 113938976A
Authority
CN
China
Prior art keywords
network
routing
node
nodes
internet
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.)
Granted
Application number
CN202111221256.1A
Other languages
Chinese (zh)
Other versions
CN113938976B (en
Inventor
万智萍
许志明
倪伟传
刘少江
邹嘉俊
王凤
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Guangzhou Xinhua College
Original Assignee
Guangzhou Xinhua College
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Guangzhou Xinhua College filed Critical Guangzhou Xinhua College
Priority to CN202111221256.1A priority Critical patent/CN113938976B/en
Publication of CN113938976A publication Critical patent/CN113938976A/en
Application granted granted Critical
Publication of CN113938976B publication Critical patent/CN113938976B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W40/00—Communication routing or communication path finding
    • H04W40/02—Communication route or path selection, e.g. power-based or shortest path routing
    • H04W40/04—Communication route or path selection, e.g. power-based or shortest path routing based on wireless node resources
    • H04W40/10—Communication route or path selection, e.g. power-based or shortest path routing based on wireless node resources based on available power or energy
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W24/00—Supervisory, monitoring or testing arrangements
    • H04W24/06—Testing, supervising or monitoring using simulated traffic
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W40/00—Communication routing or communication path finding
    • H04W40/02—Communication route or path selection, e.g. power-based or shortest path routing
    • H04W40/12—Communication route or path selection, e.g. power-based or shortest path routing based on transmission quality or channel quality
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D30/00—Reducing energy consumption in communication networks
    • Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks

Abstract

The invention provides an intelligent utility planning Internet of things passive perception routing algorithm, which comprises the following steps: constructing a network embodiment data set; calculating the utility value of each routing strategy according to the routing indexes; selecting a routing strategy with the maximum network instance utility value as a real label of the network instance; and testing the trained PCNN model by using the test data set to obtain a routing strategy with the maximum prediction probability, and comparing the routing strategy with the real label of the network instance. According to the routing quality requirement of the passive sensing network, the utility value is used for comprehensively analyzing the transmission success rate, the energy, the transmission delay and the waiting delay of the routing; a construction method of a network instance data set is provided to train a PCNN model, and a route with the highest utility value is selected as an optimal routing strategy of the current network; saving a large amount of calculation, saving energy consumption and reducing time delay.

Description

Internet of things passive sensing routing algorithm for intelligent utility planning
Technical Field
The invention relates to the technical field of passive sensing networks, in particular to an internet of things passive sensing routing algorithm for intelligent utility planning.
Background
The passive sensing network is a network formed by passive sensing nodes, and the nodes are not equipped with or mainly depend on self power supply equipment for power supply, but support calculation, sensing, communication and networking by acquiring energy from the environment. The passive nodes can maintain the operation of the network by capturing the energy of the surrounding environment, so that the passive nodes can adapt to a plurality of application scenes with limited energy supply, and belong to a network form with a very promising application prospect in the Internet of things. Although the passive sensing network can capture the energy of the surrounding environment, it does not mean that it can stably obtain energy supply for a long time, for example, for a passive sensing node relying on optical energy, the passive sensing node can suffer from insufficient energy when facing a situation that an optical signal is weak, and therefore, when selecting a network route, the transmission energy loss of the network still needs to be focused. In addition, the passive sensing node is often used for being deployed in a complex environment to perform a monitoring task, so that the transmission delay is higher under the influence of the environment and the terrain, and therefore, the problem of the routing propagation delay of the passive sensing network is also a problem that needs to be focused when a routing protocol is researched.
The existing passive sensing routing mainly comprises an EoR protocol and an opportunistic routing protocol (EDOR-Cos), wherein the EoR protocol is improved on the basis of the traditional low duty ratio opportunistic routing protocol, so that the Ethernet passive sensing routing is suitable for the characteristics of a passive sensing network and obtains a better delay effect; the protocol estimates an expected energy consumption value of a node by analyzing a node communication process, so that the node selects a neighbor node with lower energy consumption as a forwarding candidate, and when the forwarding node is finally determined, the protocol makes a decision by combining duty ratio information of a neighbor node of a next hop of the candidate node, so that a sending node selects the candidate node capable of forwarding data more quickly to reduce delay, and balance of energy consumption and delay performance is realized.
However, the process of acquiring energy by the passive sensing node deployed in the field terrain is unstable and is often influenced by factors such as illumination, and the problem of waiting delay of the passive sensing node when the residual energy is insufficient is not considered in the EoR protocol and the EDOR-Cos protocol, so that the delay of the whole routing algorithm is greatly influenced as long as the residual energy of the routing node is insufficient. Therefore, the design of the passive sensing routing algorithm needs to consider not only transmission energy consumption and transmission delay, but also waiting delay caused by insufficient node residual energy.
Disclosure of Invention
In order to solve the problems in the prior art, the invention provides an intelligent utility planning internet of things passive sensing routing algorithm, which comprehensively analyzes the problems of transmission success rate, energy, transmission delay, waiting delay and the like of a route by using a utility value according to the routing quality requirement of a passive sensing network.
The technical scheme of the invention is as follows:
an internet of things passive awareness routing algorithm for intelligent utility planning, comprising:
constructing a network embodiment data set;
calculating the utility value of each routing strategy according to the routing indexes;
selecting a routing strategy with the maximum network instance utility value as a real label of the network instance;
and testing the trained PCNN model by using the test data set to obtain a routing strategy with the maximum prediction probability, and comparing the routing strategy with the real label of the network instance.
As a further technical solution of the present invention, the utility value of each routing policy is calculated according to the routing index; the method specifically comprises the following steps:
carrying out comprehensive analysis on the routing quality of the passive sensing network by adopting a utility value, and calculating the routing utility value according to weight factors corresponding to the link transmission success rate, transmission energy consumption, end-to-end transmission delay and data packet waiting delay, wherein the routing utility value is as follows:
Figure BDA0003312708050000031
wherein, PijA link transmission success rate; eijTo represent the total transmission loss of the link; delayijEnd-to-end transmission delay; t is tijWaiting for a time delay; w is a1、w2、wa、w4Weight factors, w, of successful transmission rate, transmission energy consumption, transmission delay, and latency of the link1＞0，w2＞0，wa＞0，w4> 0 and w1+w2+w3+w4＝1。
The further technical scheme of the invention is that the weighting factors of the successful transmission rate, the transmission energy consumption, the transmission delay and the waiting time delay of the link are the same, namely w1＝w2＝wa＝w4＝0.25。
The further technical proposal of the invention is that the data set of the embodiment of the network is constructed; the method specifically comprises the following steps:
constructing a network instance data set for the nodes of the Internet of things;
an OMNET + + simulator is used to obtain a simulated network instance dataset.
As a further technical scheme of the invention, a network instance data set is constructed for the nodes of the internet of things; the method specifically comprises the following steps:
in a passive sensing network formed by nodes of the Internet of things, a graph G (V, E) is given, and V represents a set (V) of all nodes1，v2，...，vn) E represents the set of all links in the network; qijFor representing the link utility value between the i and j nodes; defining NSiRepresenting a set of neighbor nodes within an effective propagation radius R of a signal at node i; the link details of the network are recorded using a matrix S, denoted
Figure BDA0003312708050000032
If s isij1, indicates that a link i to j exists, otherwise sijDefault s is 0ii1 is ═ 1; gaugeA straight line between a source node and a target node is taken as an abscissa, the direction of the abscissa towards the target node is defined as a positive direction of signal propagation, and the direction towards the source node is defined as a negative direction of the signal propagation; each node is provided with a non-repeated sequence number, the sequence number range is (1.. multidot.n), when a propagation route is established between a source node and a destination node, the propagation route can only be specified to be propagated in a positive direction, and the signals can be successfully propagated from the source node to the destination node.
The further technical scheme of the invention is that an OMNET + + simulator is adopted to obtain a simulated network instance data set; the method specifically comprises the following steps: the method comprises the steps of setting a simulated network area by using an OMNET + + simulator, setting the propagation radius of nodes, and constructing a network example by using nodes which are randomly generated and randomly distributed in the network area, wherein the network example comprises a network example data set for training, a network example data set for verification and a network example data set for testing.
As a further technical solution of the present invention, the testing the trained PCNN model with the test data set to obtain a routing strategy with a maximum prediction probability, and comparing the routing strategy with a real label of a network instance specifically includes:
the PCNN deep learning model structure, the cross entropy loss function and the Adam function are used for training a model, a routing strategy with the maximum prediction probability is obtained under the condition that an input sample matrix reflecting the number and distribution of nodes is given, and the routing strategy is compared with a routing strategy with the highest utility value obtained by directly using utility calculation.
The invention has the beneficial effects that:
according to the routing quality requirement of the passive sensing network, the utility value is used for comprehensively analyzing the problems of transmission success rate, energy, transmission delay, waiting delay and the like of the routing, a deep learning model PCNN is provided, and a construction method of a network instance data set is provided for training the PCNN model, so that the PCNN can intelligently plan the routing for the passive sensing network only by giving the node number and the node distribution condition of the passive sensing network, and the routing with the highest utility value is selected as the optimal routing strategy of the current network. The trained PCNN can rapidly output a route prediction result, and the utility value of the route does not need to be calculated in the future use process, so that a large amount of operations are saved, and the route strategy with the highest utility value can be obtained only by giving an input sample matrix of a network instance; because the routing strategy output by the PCNN model is planned based on the maximum utility value, better effects can be obtained in the aspects of saving energy consumption and reducing time delay.
Drawings
Fig. 1 is a flow chart of an internet of things passive sensing routing algorithm for intelligent utility planning according to the present invention;
FIG. 2 is a schematic diagram of the positive and negative directions of signal propagation according to the present invention;
fig. 3 is a schematic diagram of a propagation process of a 10-node passive network according to the present invention;
FIG. 4 is a schematic diagram of a plurality of propagation routes from a source node to a destination node according to the present invention;
FIG. 5 is a diagram of a PCNN model framework according to the present invention;
FIG. 6 is a graph of test accuracy proposed by the present invention;
FIG. 7 is a graph of total energy consumption for routing results presented by the present invention;
fig. 8 is a total routing delay diagram according to the present invention.
Detailed Description
The conception, the specific structure, and the technical effects produced by the present invention will be clearly and completely described below in conjunction with the embodiments and the accompanying drawings to fully understand the objects, the features, and the effects of the present invention. It is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all embodiments, and those skilled in the art can obtain other embodiments without inventive effort based on the embodiments of the present invention, and all embodiments are within the protection scope of the present invention.
Referring to fig. 1, a diagram of an internet of things passive sensing routing algorithm for intelligent utility planning is provided in the present invention;
as shown in fig. 1, an internet of things passive awareness routing algorithm for intelligent utility planning includes:
step 100, constructing a network embodiment data set;
step 200, calculating the utility value of each routing strategy according to the routing index;
step 300, selecting a routing strategy with the maximum network instance utility value as a real label of the network instance;
step 400, the trained PCNN model is tested by using a test data set, and a routing strategy with the maximum prediction probability is obtained and compared with a real label of a network instance.
According to the routing quality requirement of the passive sensing Network, utility values are used for comprehensively analyzing the transmission success rate, energy, transmission delay and waiting delay of the routing, a deep learning model PCNN (segmented Convolutional Neural Network) is provided, and a construction method of a Network instance data set is provided for training the PCNN model, so that the node number and the node distribution condition of the passive sensing Network are only required to be given; the PCNN can intelligently perform route planning for the passive sensing network, and a route with the highest utility value is selected as an optimal route strategy of the current network. The trained PCNN can rapidly output a route prediction result, and the utility value of the route does not need to be calculated in the use process, so that a large amount of operations are saved, and the routing strategy with the highest utility value can be obtained only by giving an input sample matrix of a network instance. Because the routing strategy output by the PCNN model is planned based on the maximum utility value, better effects can be obtained in the aspects of saving energy consumption and reducing time delay.
In step 200, calculating utility values of each routing strategy according to the routing indexes; the method specifically comprises the following steps:
carrying out comprehensive analysis on the routing quality of the passive sensing network by adopting a utility value, and calculating the routing utility value according to weight factors corresponding to the link transmission success rate, transmission energy consumption, end-to-end transmission delay and data packet waiting delay, wherein the routing utility value is as follows:
Figure BDA0003312708050000061
wherein, PijA link transmission success rate;*Eijto represent the total transmission loss of the link; delayijEnd-to-end transmission delay; t is tijWaiting for a time delay; w is a1、w2、w3、w4Weight factors, w, of successful transmission rate, transmission energy consumption, transmission delay, and latency of the link1＞0，w2＞0，w3＞0，w4> 0 and w1+w2+w3+w4＝1。
Wherein, w1、w2、w3、w4The specific weight setting of (2) can be set according to the performance preference of the network, for example, the transmission energy consumption of the network needs to be increased, and can be 1 > w2Set w in the range of > 0.52Then will be
Figure BDA0003312708050000071
Is assigned to w1、w3、w4. If there is no performance preference, setting four standards of link successful transmission rate, transmission energy consumption, transmission delay and waiting time delay are equally important, that is, the weight factors of link successful transmission rate, transmission energy consumption, transmission delay and waiting time delay are the same, then w1＝w2＝w3＝w4＝0.25。
The link transmission success rate is specifically calculated as follows:
use (node)i，nodej) Representing the ith Internet of things node in the networkiAnd j' th node of internet of thingsiCommunication link between them, i.e. a link (node)i，nodej) The bit error rate of (a) is:
Figure BDA0003312708050000072
Figure BDA0003312708050000073
wherein, Ploss＝Pl+10γlog10(di，j)+Pt。PlDenotes the power loss per unit distance, PtRepresenting the received power threshold of the device, the received power threshold of all devices being the same by default. di，jRepresenting the node from the ith Internet of things nodeiAnd j' th node of internet of thingsiThe distance between, gamma, represents the power attenuation coefficient.
In the network, it is assumed that the dynamic allocation strategy is adopted for sending data packets, and if A is adoptedi，jRepresentation (node)i，nodej) Middle nodeiThe number of the adopted fragments is the nodeiBy (node)i，nodej) The link successfully transmits a data packet to the nodejThe success probability of (c) is:
Figure BDA0003312708050000074
wherein Kv6Indicating the length of the data packet.
For a passive sensing network, the nodes of the internet of things maintain the operation of the network by capturing energy from the environment, for example, the nodes can sense and capture energy from the surrounding environment such as sunlight, temperature, wind power, radio frequency signals and the like, so that the operation of the internet of things equipment is supported, and the problem of energy limitation of the internet of things is solved. However, the node cannot stably obtain energy from the periphery for a long time due to the comprehensive influence of the complex environment, so that the node is an important standard for evaluating the routing performance of the node for the passive sensing network, which reduces the transmission loss as much as possible and saves the energy.
For on-transmission links (nodes)i，nodej) Energy loss of (1), we use EijTo represent the unit transmission loss of the link, expressed as:
Eij＝α-1e0+β-1e1Di 2 (6)
where α represents a transmission loss coefficient, e0Representing the energy consumption of the node transmitting or receiving 1bit on the circuit, e1Represents the communication radius D at the nodeiAnd the energy consumption of the amplifier for transmitting 1bit data is reduced, and beta represents the loss coefficient of the amplifier.
Then an s-length coded packet, node, is transmittediThe transmission energy consumption is as follows:
Eij t(s)＝sEij (7)
nodejThe receiving energy consumption is as follows:
Eij r(s)＝sα-1e0 (8)
considering that if the data packet adopts the dynamic allocation strategy, the A needs to be correctly receivedi，jOnly when one coded packet is decoded can the node be enabledjA complete data packet is successfully received. Therefore, we consider here the transmission energy consumption in case of successful reception of a data packet. According to nodeiBy (node)i，nodej) The link successfully transmits a data packet to the nodejHas a success probability of pijThen a data packet is sent from nodeiSend to nodejThe total transmission energy consumption for successful reception is:
Figure BDA0003312708050000081
for a passive sensing network used in emergency scenes such as a disaster monitoring system or a battlefield environment monitoring system, the requirement for link transmission delay is very high, so in order to construct a better node route, the problem of transmission delay needs to be considered.
NodeiTransmitting a signal of length li，jData packet to node ofjThe transmission delay is:
Figure BDA0003312708050000082
tpdisplay sectionThe point is the time required to contend for the channel and encode the data before transmitting the data packet. t is tvWhich indicates the transmission rate of a unit of data,
Figure BDA0003312708050000091
denotes the transmission duration, and epsilon denotes the loss factor per propagation distance.
Considering the need to correctly receive Ai，jOnly when an encoded packet is decoded and reassembled into a complete data packet, according to (node)i，nodej) Probability p of successful transmission of data packets of a linkijThen a data packet is sent from nodeiSend to nodejThe delay for successful reception is:
Figure BDA0003312708050000092
since the passive node may not be able to completely transmit the IPv6 packet due to insufficient energy, it needs to wait for capturing enough energy from the surrounding environment to continue to complete the transmission of the IPv6 packet, and therefore, it is also necessary to consider the waiting delay when the remaining energy of the node is insufficient for the transmission process of the data packet.
Let us assume a nodeiThe residual energy of
Figure BDA0003312708050000093
NodeiIt is required to successfully send k IPv6 packets to the nodejIf the remaining energy is insufficient, the nodeiThe amount of energy to be captured is
Figure BDA0003312708050000094
Figure BDA0003312708050000095
The node can obtain the electric energy from the surrounding environment within the duration t as follows:
Ec(t)＝δηt (13)
δ represents the charge efficiency coefficient of the node capacitor, and η represents the average energy capture rate of the node.
According to
Figure BDA0003312708050000097
And Ec(t) when nodeiThe residual energy of
Figure BDA0003312708050000098
And needs to successfully send k IPv6 packets to the nodejTime, required waiting delay tijComprises the following steps:
Figure BDA0003312708050000096
in step 100, a network embodiment data set is constructed; the method specifically comprises the following steps:
constructing a network instance data set for the nodes of the Internet of things;
an OMNET + + simulator is used to obtain a simulated network instance dataset.
The method comprises the following steps that a network instance data set is constructed for nodes of the Internet of things; the method specifically comprises the following steps: in a passive sensing network formed by nodes of the Internet of things, a graph G (V, E) is given, and V represents a set (V) of all nodes1，v2，...，vn) E represents the set of all links in the network; qijFor representing the link utility value between the i and j nodes; defining NSiRepresenting a set of neighbor nodes within the effective propagation radius R of the signal at node i (all nodes are assumed to have the same propagation radius by default). The link details of the network are recorded using a matrix S, denoted
Figure BDA0003312708050000101
If s isij1, indicates that a link i to j exists, otherwise sijDefault s is 0ii1 (all nodes are connected to themselves). A straight line between a source node and a target node is defined as an abscissa, and a direction of the abscissa toward the target node is defined as a positive direction of signal propagation toward the source nodeThe direction of the node is defined as the negative direction of signal propagation, as shown in fig. 2. Assuming that each node has a non-repeating sequence number with a sequence number range of (1.. multidot.n), when a propagation route is constructed between a source node and a destination node, the propagation route can only be propagated in a positive direction (the propagation in a negative direction can form a circular link), and the signals can be successfully propagated from the source node to the destination node.
In the embodiment of the invention, an OMNET + + simulator is adopted to obtain a simulated network instance data set; the method specifically comprises the following steps: the method comprises the steps of setting a simulated network area by using an OMNET + + simulator, setting the propagation radius of nodes, and constructing a network example by using nodes which are randomly generated and randomly distributed in the network area, wherein the network example comprises a network example data set for training, a network example data set for verification and a network example data set for testing.
For a network instance data set for PCNN network training, an OMNET + + simulator is adopted to obtain a simulated network instance data set, the range of a simulated network area is set to be 1000m x 1000m by using the OMNET + + simulator, the propagation radius of nodes is set to be 50m, the number of nodes which can be randomly generated by each network instance is limited to 2000 (node sequence numbers are from 1 to 2000), network instances are constructed by using randomly generated nodes and a mode that the nodes are randomly distributed in the network area, 10000 network instance data sets for training, 5000 network instance data sets for verification and 2000 network instance data sets for testing are constructed.
In step 300, selecting the routing policy with the maximum network instance utility value as the real label of the network instance; a method of utility value is used to determine the true label for each network instance. Taking an example of a network with 10 passive nodes as an example in fig. 3, a method for determining a true tag of the network example is described:
the linear distance between the source node (node 1) and the destination node (node 10) is used to form the abscissa, and the initial matrix s is a zero matrix. From node 1, the signal is propagated in the forward direction, node 1 and neighbor node set NS in effective propagation radius R1The nodes therein are all connected, as shown in the following figure, node 1 is connectedFollowing node 2 and node 3, two links 1 → 2, 1 → 3 are formed, respectively.
Thus s12＝1，s13Recording is performed using the S matrix as 1:
Figure BDA0003312708050000111
then, the node 2 and the node 3 propagate signals in the forward direction, and the node 2 is connected with a neighbor node set NS2Node 4 and node 3 are also connected with neighbor node set NS3The node 4 at the inner side forms two links 1 → 2 → 4, 1 → 3 → 4, respectively, as shown in fig. 4. Thus s12，s24＝1，s13，s34When 1, the S matrix is updated continuously:
Figure BDA0003312708050000112
then, the node 4 propagates a signal in the forward direction, and is connected with a neighbor node set NS4Node 5 and node 6, four links are formed, s12，s24，s45＝1、s12，s24，s46＝1、s13，s34，s45＝1、s13，s34，s 461. And continuously updating the S matrix:
Figure BDA0003312708050000121
the signal then continues to propagate in the forward direction until the signal reaches the destination node (node 10), as shown in fig. 4, where there are multiple propagation routes.
And updates the S matrix as:
Figure BDA0003312708050000122
from the S matrix recording the link details, it can be seen that8 propagation routes are arranged from the source node 1 to the destination node 10, the utility of each propagation route is calculated and compared to obtain the propagation route with the highest utility, and the route is updated to a utility matrix SmIn (1). Assume that in the propagation route of fig. 4, the route with the highest utility value is:
1 → 2 → 4 → 5 → 7 → 9 → 10, then the utility matrix SmComprises the following steps:
Figure BDA0003312708050000131
will smAs a real label for the current network example (10-node passive network).
In step 400, the trained PCNN model is tested using a test data set to obtain a routing strategy with a maximum prediction probability, which is compared with a real label of a network instance, and specifically includes:
the PCNN deep learning model structure, the cross entropy loss function and the Adam function are used for training a model, a routing strategy with the maximum prediction probability is obtained under the condition that an input sample matrix reflecting the number and distribution of nodes is given, and the routing strategy is compared with a routing strategy with the highest utility value obtained by directly using utility calculation.
The Convolutional Neural Networks (CNN) have excellent capacity of extracting space and abstract features in the field of computer vision, a PCNN model is designed based on the CNN and can extract input features of a current network, the PCNN model consists of a Convolutional layer, a pooling layer, a feature fusion layer and a logistic regression layer, the output probability of each route in an output layer can be obtained by training the PCNN model, the larger the probability is, the closer the probability is to a real label (the route determined by the optimal utility value), and the route with the maximum output probability is selected as a routing strategy of the current network.
The framework of PCNN is shown in fig. 5, where K × 1 represents the length × width × channel of the input sample, and K represents the maximum number of nodes of the network instance (i.e., input sample), i.e., K ═ 2000. The matrix structure of the sample is explained:
Figure BDA0003312708050000141
in a matrix of the input samples, e.g. aij1 denotes that j node is within the propagation radius of i node, default aii1, the node is in the propagation radius range of the node, and the matrix structure of the input sample can reflect the node number and the node distribution.
The model structure was analyzed as follows:
(1) for the first convolution layer, 3 x 2 is defined as the convolution kernel of one length x width kernel, filling 2 layers, with a step size of 1.
(2) And the pooling layer adopts maximum pooling and is filled.
(3) For the second convolution layer, 2 layers were filled with a convolution kernel of 3 x 8, step size 1.
(4) And the pooling layer adopts maximum pooling and is filled.
(5) For the third convolution layer, 2 layers were filled with a convolution kernel of 3 x 16, step size 1.
(6) And the pooling layer adopts maximum pooling and is filled.
(7) A full connection layer for mapping the characteristic diagram into vectors and outputting the vectors, wherein the elements in the vectors are each possible routing strategy of the network example (each routing strategy is represented as a link detail matrix S)
(8) And outputting the absolute value of each link detail matrix S, and converting the absolute value into probability output through a softmax function. The softmax function is commonly used for the multi-class structure of the deep neural network, so that all output results are normalized.
Figure BDA0003312708050000142
eiRepresenting the ith output result, and having m outputs in total.
Network loss function: the loss function we use is "cross entropy loss". The cross-entropy loss is very effective for estimating the loss of the multi-classification method, and the adopted cross-entropy loss function is in the form of:
Figure BDA0003312708050000151
xidenotes a specific example, p (x)i) Representing genuine labels, when the label of the instance belongs to the genuine category, p (x)i) When not in the true category, p (x)i)＝0。q(xi) And showing the prediction result of the example processed by the model.
The network optimization method comprises the following steps: in this deep neural network model, we use the Adam function of the pytorech framework as the network optimizer. The learning rate adopts a default learning rate of 0.001 which is "Adam Optimizer".
In the embodiment of the invention, a PCNN deep learning model structure training model is used, a training set and a verification set are used for training the model, and a test set is used for testing the performance of the PCNN; through training, the PCNN model automatically outputs the prediction probability of each routing strategy in the network instance according to the number of nodes of the network instance and the node distribution, and the routing strategy with the maximum prediction probability is selected as the optimal route of the current network instance.
The test accuracy of the PCNN model is shown below when the PCNN model is trained for 1000 times, and the PCNN model is trained and tested by using a Pythroch frame and realized by using a python language.
Testing the trained PCNN model by using a test data set, obtaining a routing strategy with the maximum prediction probability under the condition of giving an input sample matrix reflecting the number and distribution of nodes (not giving a real label), comparing the routing strategy with a routing strategy with the highest utility value obtained by directly using utility calculation, and if the routing strategy is consistent with the routing strategy, enabling N to be the sametrue＝Ntrue+1 (initial Ntrue0), the number of test samples is represented by N, the test accuracy is represented by
Figure BDA0003312708050000152
And (4) determining.
In the test process, the test results of the PCNN model, as shown in fig. 6, are about 97% in accuracy, where N is 500, 1000, 1500, and 2000. The result of the output routing strategy can be very close to the routing strategy with the maximum utility value after the PCNN model is trained.
For the performance of the invention in terms of routing energy consumption and routing delay of the passive sensing network, a trained PCNN model (the training times are 1000 times) is used for comparing the network performance with other passive sensing routing algorithms. The comparison algorithm is an EoR protocol and an EDOR-Cos protocol. The experimental environments of all algorithms are kept consistent, and an OMNET + + simulator is used for constructing a network simulation environment and performing simulation tests.
And comparing routing energy consumption: the simulated network area is 1000m by 1000m, the propagation radius of each node is 50m, simulation is performed under the condition that the number of network nodes is V500, 1000, 1500 and 2000 respectively, the nodes are randomly distributed in the network area, a source node and a destination node are randomly selected by a system, routing energy consumption is represented as total routing node energy consumption required for successfully sending a data packet from the source node to the destination node, and the simulation result is an average value after 100 times of simulation.
As can be seen from the total energy consumption result of the route in fig. 7, as the number of network nodes increases, the energy consumption of the route increases. This is because the source node and the destination node are systematically and randomly selected, and when the number of network nodes increases, the routing path from the source node to the destination node may be longer, that is, the more nodes that are needed to be used as relays, the greater the energy consumption. From a comparison, the PCNN model has lower network routing energy consumption than the EoR protocol and the EDOR-Cos protocol.
Comparing the transmission time delay of the route: the simulated network area is 1000m by 1000m, the propagation radius of each node is 50m, simulation is performed under the condition that the number of network nodes is V500, 1000, 1500 and 2000 respectively, the nodes are randomly distributed in the network area, a source node and a destination node are randomly selected by a system, routing delay is represented by total time required for successfully sending a data packet from the source node to the destination node, and the simulation result is an average value after 100 times of simulation.
As can be seen from the total delay result of the route in fig. 8, as the number of network nodes increases, the total transmission delay of the route also increases. This is because the source node and the destination node are systematically and randomly selected, and when the number of network nodes increases, the routing path from the source node to the destination node may be longer, and the time required for transmitting the packet from the source node to the destination node increases. From a comparison, the PCNN model has a lower total delay of routing transmission than the EoR protocol and the EDOR-Cos protocol. The utility value analysis method adopted by the invention not only focuses on the end-to-end transmission time of the link, but also focuses on the residual energy of the nodes, so that the waiting time in the transmission process can be prevented from being prolonged due to the insufficient residual energy of the passive routing nodes, and the total transmission time is prevented from being prolonged. And the EoR protocol and the EDOR-Cos protocol do not concern the latency problem that may be caused by insufficient energy remaining in the routing node.
The present invention has been described in detail, but the present invention is not limited to the above embodiments, and various changes can be made without departing from the gist of the present invention within the knowledge of those skilled in the art. Many other changes and modifications can be made without departing from the spirit and scope of the invention. It is to be understood that the invention is not to be limited to the specific embodiments, but only by the scope of the appended claims.

Claims (7)

1. An internet of things passive perception routing algorithm for intelligent utility planning, comprising:
constructing a network embodiment data set;
calculating the utility value of each routing strategy according to the routing indexes;
selecting a routing strategy with the maximum network instance utility value as a real label of the network instance;
and testing the trained PCNN model by using the test data set to obtain a routing strategy with the maximum prediction probability, and comparing the routing strategy with the real label of the network instance.
2. The passive-aware routing algorithm for internet of things for intelligent utility planning according to claim 1, wherein the utility value of each routing strategy is calculated according to a routing index; the method specifically comprises the following steps:
carrying out comprehensive analysis on the routing quality of the passive sensing network by adopting a utility value, and calculating the routing utility value according to weight factors corresponding to the link transmission success rate, transmission energy consumption, end-to-end transmission delay and data packet waiting delay, wherein the routing utility value is as follows:
Figure FDA0003312708040000011
wherein, PijA link transmission success rate; eijTo represent the total transmission loss of the link; delayijEnd-to-end transmission delay; t is tijWaiting for a time delay; w is a1、w2、w3、w4Weight factors, w, of successful transmission rate, transmission energy consumption, transmission delay, and latency of the link1＞0，w2＞0，w3＞0，w4> 0 and w1+w2+w3+w4＝1。
3. The passive-aware routing algorithm for internet of things for intelligent utility planning of claim 2, wherein the weighting factors of successful transmission rate, transmission energy consumption, transmission delay and waiting delay of the link are the same, i.e. w1＝w2＝w3＝w4＝0.25。
4. The passive-aware routing algorithm for internet of things for intelligent utility planning according to claim 1, wherein the construction network embodiment dataset; the method specifically comprises the following steps:
constructing a network instance data set for the nodes of the Internet of things;
an OMNET + + simulator is used to obtain a simulated network instance dataset.
5. The passive-aware routing algorithm for internet of things for intelligent utility planning of claim 4, wherein the network instance dataset is constructed for the nodes of internet of things; the method specifically comprises the following steps:
in a passive sensing network formed by nodes of the Internet of things, a graph G (V, E) is given, and V represents a set (V) of all nodes1，υ2，...，vn) E represents the set of all links in the network; qijFor representing the link utility value between the i and j nodes; defining NSiRepresenting a set of neighbor nodes within an effective propagation radius R of a signal at node i; the link details of the network are recorded using a matrix S, denoted
Figure FDA0003312708040000021
If s isij1, indicates that a link i to j exists, otherwise sijDefault s is 0ii1 is ═ 1; a straight line between a source node and a target node is taken as an abscissa, the direction of the abscissa towards the target node is defined as a positive direction of signal propagation, and the direction towards the source node is defined as a negative direction of signal propagation; each node is provided with a non-repeated sequence number, the sequence number range is (1.. multidot.n), when a propagation route is established between a source node and a destination node, the propagation route can only be specified to be propagated in a positive direction, and the signals can be successfully propagated from the source node to the destination node.
6. The passive-aware routing algorithm for internet of things for intelligent utility planning according to claim 4, wherein an OMNET + + simulator is used to obtain a simulated network instance dataset; the method specifically comprises the following steps: the method comprises the steps of setting a simulated network area by using an OMNET + + simulator, setting the propagation radius of nodes, and constructing a network example by using nodes which are randomly generated and randomly distributed in the network area, wherein the network example comprises a network example data set for training, a network example data set for verification and a network example data set for testing.
7. The internet of things passive perception routing algorithm for intelligent utility planning according to claim 1, wherein the testing the trained PCNN model using the test data set to obtain the routing strategy with the highest prediction probability, and comparing the routing strategy with the real label of the network instance specifically includes:
the PCNN deep learning model structure, the cross entropy loss function and the Adam function are used for training a model, a routing strategy with the maximum prediction probability is obtained under the condition that an input sample matrix reflecting the number and distribution of nodes is given, and the routing strategy is compared with a routing strategy with the highest utility value obtained by directly using utility calculation.
CN202111221256.1A 2021-10-20 2021-10-20 Intelligent utility planning passive perception routing algorithm of Internet of things Active CN113938976B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111221256.1A CN113938976B (en) 2021-10-20 2021-10-20 Intelligent utility planning passive perception routing algorithm of Internet of things

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111221256.1A CN113938976B (en) 2021-10-20 2021-10-20 Intelligent utility planning passive perception routing algorithm of Internet of things

Publications (2)

Publication Number Publication Date
CN113938976A true CN113938976A (en) 2022-01-14
CN113938976B CN113938976B (en) 2023-11-07

Family

ID=79280804

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111221256.1A Active CN113938976B (en) 2021-10-20 2021-10-20 Intelligent utility planning passive perception routing algorithm of Internet of things

Country Status (1)

Country Link
CN (1) CN113938976B (en)

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2000045584A1 (en) * 1999-01-28 2000-08-03 Bios Group Lp A method and system for routing control in communication networks and for system control
US20060046658A1 (en) * 2002-09-05 2006-03-02 Cruz Rene L Scheduling methods for wireless networks
US20080195360A1 (en) * 2006-07-10 2008-08-14 Cho-Yu Jason Chiang Automated policy generation for mobile ad hoc networks
CN103763193A (en) * 2014-02-21 2014-04-30 重庆邮电大学 Multi-replication routing method for selecting eruption range in vehicular vdhoc networks
CN112821940A (en) * 2021-01-15 2021-05-18 重庆邮电大学 Satellite network dynamic routing method based on inter-satellite link attribute

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2000045584A1 (en) * 1999-01-28 2000-08-03 Bios Group Lp A method and system for routing control in communication networks and for system control
US20060046658A1 (en) * 2002-09-05 2006-03-02 Cruz Rene L Scheduling methods for wireless networks
US20080195360A1 (en) * 2006-07-10 2008-08-14 Cho-Yu Jason Chiang Automated policy generation for mobile ad hoc networks
CN103763193A (en) * 2014-02-21 2014-04-30 重庆邮电大学 Multi-replication routing method for selecting eruption range in vehicular vdhoc networks
CN112821940A (en) * 2021-01-15 2021-05-18 重庆邮电大学 Satellite network dynamic routing method based on inter-satellite link attribute

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
JIUXIANG GU；ZHENHUA WANG: "Recent advances in convolutional neural networks", 《PATTERN RECOGNITION》 *
万智萍: "基于协同路由架构的认知无线网络频谱及功率分配算法", 《南京邮电大学学报(自然科学版)》 *
邱旭: "基于轨迹预测的车载网路由协议研究", 《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》 *

Also Published As

Publication number Publication date
CN113938976B (en) 2023-11-07

Similar Documents

Publication Publication Date Title
CN110263280B (en) Multi-view-based dynamic link prediction depth model and application
Al-Qurabat et al. Two tier data reduction technique for reducing data transmission in IoT sensors
CN109831386B (en) Optimal path selection algorithm based on machine learning under SDN
KR102143593B1 (en) Method for detecting anomaly of Internet of Things device based on autoencoder and system thereof
Al‐Qurabat et al. Data gathering and aggregation with selective transmission technique to optimize the lifetime of Internet of Things networks
CN111935747B (en) Method for predicting link quality of wireless sensor network by adopting GRU (generalized regression Unit)
CN103428704B (en) A kind of frequency spectrum sensing method and device
Lee et al. Performance analysis of local exit for distributed deep neural networks over cloud and edge computing
CN113469425B (en) Deep traffic jam prediction method
Al-Qurabat et al. Distributed data aggregation protocol for improving lifetime of wireless sensor networks
Boucetta et al. QoS in IoT networks based on link quality prediction
Jain et al. Data-prediction model based on stepwise data regression method in wireless sensor network
Affane et al. Energy enhancement of routing protocol with hidden Markov model in wireless sensor networks
Sridhar et al. A machine learning-based intelligence approach for multiple-input/multiple-output routing in wireless sensor networks
CN113938976B (en) Intelligent utility planning passive perception routing algorithm of Internet of things
KR102006292B1 (en) Apparatus and method for transporting multimedia using FEC, and DNN model selecting apparatus for determining FEC parameter
CN116155805A (en) Distributed intelligent routing method, system, electronic equipment and storage medium
CN115622603A (en) Age minimization optimization method for auxiliary transmission information
Chen et al. Tasks-oriented joint resource allocation scheme for the Internet of vehicles with sensing, communication and computing integration
CN108769944B (en) MP-MR-MC wireless sensor network data collection method for bridge structure health monitoring
CN108668265B (en) Method for predicting meeting probability among mobile users based on cyclic neural network
Abdoulaye et al. Embedded Artificial Neural Network for Data Prediction in Energy Efficient Wireless Sensors Networks
CN108990128B (en) Route design method based on mobile perception in mobile network
Zhang et al. Selecting the best routing traffic for packets in LAN via machine learning to achieve the best strategy
CN117692026B (en) Link sensing method and device for power line communication

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
GR01 Patent grant
GR01 Patent grant