A Smart Non-uniform Clustering Method for Underwater Acoustic Cooperative Networks Based on Data Hierarchy
By introducing a data hierarchical and dynamic coding cooperative transmission mechanism into the underwater acoustic network, and combining it with the particle swarm optimization algorithm for uneven clustering, the problems of communication reliability and real-time data transmission in the underwater acoustic network are solved, achieving efficient data transmission and extended network lifetime.
Patent Information
- Application Number
- CN202311218196.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-09-20
- Publication Date
- 2026-11-17
- Estimated Expiration
- 2043-09-20
AI Technical Summary
Existing underwater acoustic network clustering methods fail to effectively combine cooperative communication technology and data hierarchical information, resulting in insufficient communication reliability and real-time data transmission, which cannot meet the needs of underwater data collection tasks.
A data-hierarchical underwater acoustic cooperative network intelligent non-uniform clustering method is adopted. The cluster head node is selected by particle swarm optimization algorithm, combined with dynamic coding cooperative transmission mechanism, and the concept of data hierarchy is introduced to ensure the real-time priority transmission of high-importance data and balance network energy consumption.
It improves the communication reliability and throughput of the underwater acoustic network, extends the network lifespan, and meets the real-time and high-efficiency requirements of underwater data collection tasks.
Smart Images

Figure CN117221981B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to underwater communication, and more particularly to an intelligent non-uniform clustering method for underwater acoustic cooperative networks based on data hierarchical classification. Background Technology
[0002] Underwater acoustic networks are network systems used for underwater data collection, playing a crucial role in various fields such as underwater exploration, military surveillance, data acquisition, and underwater disaster prevention. Given the characteristics of underwater acoustic networks, including low bandwidth, long latency, high energy consumption, and limited energy supply, designing more efficient and reliable underwater routing protocols to improve network throughput, reduce end-to-end transmission latency, and enhance communication reliability is a key technical challenge in the development of underwater acoustic communication networking technology. In underwater acoustic network data collection tasks, some critical data needs to be transmitted and processed promptly to ensure network real-time performance. Therefore, prioritizing the transmission of highly important data and improving the efficiency and effectiveness of underwater acoustic networks has become a crucial direction for current development.
[0003] Cooperative communication technology can improve the communication reliability of underwater acoustic networks. Y. Chen et al. proposed an OFDM-Modulated Dynamic Coded Cooperation (DCC) scheme suitable for underwater acoustic channels (Y. Chen, et al. OFDM-Modulated Dynamic Coded Cooperation in Underwater Acoustic Channels[J]. IEEE Journal of Oceanic Engineering, 2015,40(1):159-168.), which realizes the effective reuse of multi-dimensional resources such as frequency, time, and space, supports higher-speed and higher-bandwidth underwater acoustic data transmission, and effectively improves the coverage performance of underwater acoustic networks.
[0004] Clustering techniques can effectively balance the energy consumption of underwater acoustic networks, improving network lifetime and scalability. Heinzelman et al. proposed a low-energy adaptive hierarchical clustering (LEACH) scheme (WB Heinzelman, AP Chandrakasan, H. Balakrishnan. An Application-specific Protocol Architecture for Wireless Microsensor Networks[J]. IEEE Transactions on Wireless Communications, 2002,1(4):660-670.), which distributes the energy load of the entire network to each node by periodically rotating the cluster head with equal probability, thereby reducing network communication energy consumption and extending the overall network lifetime. However, the randomness of cluster head selection can easily lead to uneven energy consumption throughout the network, causing some nodes to die prematurely and reducing network performance. To provide a good solution and balance network energy consumption, researchers have proposed various clustering methods for underwater acoustic networks.
[0005] Krishnaswamy et al. proposed a clustering method based on fuzzy clustering and particle swarm optimization (V.Krishnaswamy, SS Manvi. Fuzzy and PSO Based Clustering Scheme in Underwater Acoustic Sensor Networks Using Energy and Distance Parameters[J].Wireless Pers. Commun., 2019,108(3):1529-1546.). This method considers the transmission distance of nodes within a cluster, the distance between the cluster head and the sink, and energy consumption. It uses fuzzy clustering to cluster nodes and then selects the cluster head using particle swarm optimization, which effectively reduces network energy consumption. R. Hou et al. proposed an Unequal Clustering Method Based on Particle Swarm Optimization in Underwater Acoustic Sensor Networks (R. Hou, et al. An Unequal Clustering Method Based on Particle Swarm Optimization in Underwater Acoustic Sensor Networks[J]. IEEE Internet of Things Journal, 2022,9(24):25027-25036.). This method uses the particle swarm optimization algorithm to select cluster heads and combines the idea of unequal clustering. By controlling the size of each cluster, the energy consumption of the cluster head nodes is balanced, effectively balancing and reducing the energy consumption of the entire network and extending the network lifetime.
[0006] In summary, current research on clustering methods for underwater acoustic networks mainly focuses on balancing network energy consumption and extending network lifetime, but there is a lack of research combining cooperative communication technology and considering data hierarchical information for clustering. Further research is needed to improve communication reliability and data transmission real-time performance to meet the requirements of underwater data collection tasks. Summary of the Invention
[0007] The purpose of this invention is to provide an intelligent non-uniform clustering method for underwater acoustic cooperative networks based on data hierarchy. This method comprehensively considers node energy, transmission energy consumption, and available cooperative nodes, and uses a particle swarm optimization algorithm for intelligent non-uniform clustering to balance the energy consumption of cluster head nodes, alleviate the "energy hotspot" problem, and improve network lifetime. Combined with a Dynamic Coding Cooperative (DCC) transmission mechanism, it improves data transmission reliability and network throughput. Furthermore, by introducing the concept of data hierarchy, different clustering strategies are selected for different data levels to ensure that high-importance data can be transmitted in real-time and quickly with priority. This results in a clustering method that balances network energy consumption, improves communication reliability, network throughput, and the overall network lifetime.
[0008] This invention includes the following steps:
[0009] 1) Assuming in m× m× In an underwater environment within a range of m, the underwater acoustic nodes constituting the underwater acoustic network are randomly distributed, denoted as node 1, node 2, node 3, ..., node k. All underwater acoustic nodes have the same initial energy and can control their own transmission power. A sink node is deployed on the water surface. This node is equipped with both an acoustic modem and a radio modem. It can communicate with the underwater acoustic nodes through an acoustic link and also transmit information to the shore-based control center through radio communication.
[0010] Among them, the Sink node has strong storage and computing capabilities, and can store and process the location and energy information of all underwater acoustic nodes. It is also responsible for calculating the clustering of all underwater acoustic nodes, and can be powered by solar energy without considering energy consumption issues.
[0011] 2) The Sink node selects the cluster head node (CH) using the particle swarm optimization algorithm. The steps are as follows:
[0012] 2.1) Particle Encoding: In the particle swarm optimization algorithm, a particle PS is a sequence of numbers consisting of "0" and "1". Each bit in the sequence corresponds to a surviving underwater acoustic node. A "1" at the corresponding position indicates that the underwater acoustic node is a cluster head node CH, and a "0" indicates that the underwater acoustic node is a non-cluster head node nCH. Each particle represents a possible cluster head selection scheme.
[0013] 2.2) Particle Initialization: During data transmission, the energy consumption of CH is much greater than that of nCH. Therefore, it is necessary to select underwater acoustic nodes with higher remaining energy as CHs. To protect underwater acoustic nodes with lower remaining energy and ensure that enough underwater acoustic nodes participate in cluster head selection, the remaining energy of all underwater acoustic nodes is first sorted in descending order, and nodes with higher remaining energy are selected. % of the underwater acoustic nodes are selected as candidate nodes, according to the proportion P CH Randomly generate cluster head nodes to create a particle population. PS i This represents one possible cluster head selection scheme;
[0014] To ensure a more uniform distribution of CHs when generating the particle population, interval thresholds are set between CHs. :
[0015] (1)
[0016] in, To determine the number of surviving underwater acoustic nodes, the distance between CHs must be greater than a certain threshold during particle initialization and particle update. conditions;
[0017] 2.3) Fitness Function: The fitness function is used to judge the performance of cluster head selection results, and is expressed as follows:
[0018] (2)
[0019] in, , and Let represent the optimization objective functions related to residual energy, transmission energy consumption, and cooperative gain, respectively. , and Let the weights be factors, satisfying the following relationship:
[0020] (3)
[0021] To ensure that the remaining energy of the selected CH is as high as possible, optimize the objective function. Defined as:
[0022] (4)
[0023] in, The average remaining energy of all underwater acoustic nodes. The average remaining energy of particles labeled CH is the average remaining energy of CH. The higher the average remaining energy of CH, the better the optimization objective function. The smaller it is;
[0024] Optimize objective function The performance evaluation of cluster head selection results, which incorporates single-round transmission energy consumption, is defined as follows:
[0025] (5)
[0026] in, The total energy consumption for a single round of transport according to the CH and nCH labeled by the particles;
[0027] Available cooperative nodes are defined as underwater acoustic nodes located between two CHs or between a CH and a Sink node on the transmission path, where there is sufficient remaining energy. In the fitness function formula (2), cooperative communication is taken into account for performance considerations, thus defining the optimization objective function. for:
[0028] (6)
[0029] in, For the number of CHs, RN i For node CH during inter-cluster transmission i The number of times N acts as a relay nodeRi For node CH i The number of available cooperative nodes on its next-hop path; the number of times it acts as a relay during data collection (RN). i More CH i The larger the amount of data transmitted, the more available cooperative nodes are needed to ensure communication reliability and prevent the cooperative nodes from consuming too much energy.
[0030] because , and The values differ significantly, so the Sigmoid function is needed first. Map it to a value between 0 and 1, and then use a weighting factor. , and Weighted calculation; the smaller the value of the particle fitness function, the higher the average remaining energy of the CH in the clustering result of the particle, the less energy consumption per round of transmission, and the more reasonable allocation of cooperative nodes.
[0031] 2.4) After particle initialization, calculate and compare the fitness values of all particles to obtain the historical best solution set for all particles. And the globally optimal particle GB, where HB i It is particle PS i The corresponding historical best;
[0032] 2.5) Particles are updated by tracking the current historical optimum and the global optimum: For particle PS i In the In the round of iteration:
[0033] (6)
[0034] in, Indicates the first Particle PS after the end of the round iteration i ,function Indicated by Probabilistic mutation of particles:
[0035] (7)
[0036] in, The mutation operator represents the mutation of particle PS. i Perform a mutation operation, causing one CH node in the candidate node to mutate into nCH, and one nCH node to mutate into CH; probability. With the current particle The degree of difference between the global optimal particle GB and related:
[0037] (8)
[0038]
[0039] in, and Defined respectively The maximum and minimum values, the operator Used to calculate particle PS i The number of bits in different encoding bits of the globally optimal particle GB. The dimension of the particle is the number of surviving underwater acoustic nodes.
[0040] In formula (6), the function and They represent respectively with and Probability of particles PS i Perform cross operations:
[0041] (9)
[0042]
[0043] (10)
[0044]
[0045] Among them, the cross operator Indicates the relationship between particles PS i And the best historical particle HB i Perform cross operations, proportional to P H Extract particle fragments and cross them. and These represent the current iteration number and the maximum iteration number, respectively.
[0046] 2.6) In step 2.5), after each mutation or crossover operation, the particle needs to be checked to see if it meets the interval threshold condition of formula (1). If it does not meet the condition, the mutation or crossover operation is repeated until the condition is met. At the same time, the fitness value of the updated particle is calculated according to formula (2). If the particle fitness value is larger, it means that the cluster head selection result corresponding to the particle is worse. In order to ensure particle diversity, the probability is still used. Accept the particle as a new particle:
[0047] (11)
[0048] If a particle's fitness value is smaller, the particle is directly accepted as a new particle. After all particles have completed one round of updates, the historical best solution set HB and the globally best particle GB are updated, and the iteration continues.
[0049] 2.7) After reaching the maximum number of iterations, the globally optimal particle GB is used as the result of cluster head selection, and the Sink node broadcasts the result to all CHs;
[0050] 3) After selecting the cluster head node, it is necessary to determine the inter-cluster multi-hop transmission path, as follows:
[0051] 3.1) Arrange the CHs in ascending order of depth. Starting with the shallowest CH, calculate the next-hop relay for each CH. The next hop of the shallowest CH must be a Sink node. For deeper CHs... i It can select a CH or Sink node with a shallower depth than itself as the next-hop relay and calculate the transmission cost. :
[0052] (12)
[0053] in, The total energy consumption along the transmission path is calculated by CH. i The energy consumption generated by transmitting a 1-bit data packet to the Sink node along the selected path is obtained. After selecting this path, CH i The number of jumps, The weight values are those ranging from (0,1); CH i Choose to The smallest CH or Sink node is used as the next-hop relay;
[0054] 3.2) Once the deepest CH completes the next hop selection, the inter-cluster multi-hop transmission path is constructed. The number of times each CH acts as a relay node (RN) is recorded. At the same time, the Sink node also needs to determine all available cooperative nodes on the next hop path of each CH based on the current node's location information, and broadcast the next hop node and available cooperative node information to the corresponding CH.
[0055] 3.3) After receiving the broadcast information from the Sink node, the CH records the information of the next-hop node and available cooperating nodes in its local storage space;
[0056] 4) Initial cluster range for each CH As determined by equation (1), in order to balance the energy consumption of each CH and extend the network lifetime, it is necessary to change the cluster range according to the actual situation and define the cluster range of CH. :
[0057] (13)
[0058] in, and Representing nodes CH respectively iThe remaining energy and its distance from the Sink node, This is an approximation of the average energy consumption of CH after one round of data transmission. This represents the maximum distance between nodes in the network and the sink node. , and All values are weights within the range (0,1); additionally, the Sink node can also serve as a cluster head node, with a fixed cluster range. ;
[0059] 5) Data Importance It can be divided into The levels represent unimportant, general, and important data, respectively. For each nCH node, the Sink node calculates the clustering cost according to the different data levels. :
[0060] (14)
[0061] in, Indicates nCH to CH i distance, For CH i The number of hops is used to select the cluster head node for each nCH node based on the minimum clustering cost under different data grading conditions: for unimportant data and general data, the cluster range selected by nCH is determined. Larger and closer CHs serve as cluster head nodes, forming a non-uniform clustering structure. The clustering result is called non-uniform clustering, and the corresponding CHs are denoted as CH. unequal For critical data, nCH selects cluster head nodes primarily based on their distance from the CH and the number of hops in the CH, ensuring the shortest transmission path. The clustering result is called shortest path clustering, and the corresponding CH is denoted as CH. shortest The Sink node broadcasts the clustering results, and the nCH node stores the information of the CH node corresponding to the two clustering results in its local storage space.
[0062] 6) During data transmission, due to the short distance between CH and nCH within the same cluster, it can be assumed that CH can successfully receive and decode all signals from nCH without the help of cooperating nodes; when an nCH node collects data, it first determines the data classification of the data. ,like Then select CH unequal Transmit; if Then select CH shortest Transmit;
[0063] 7) Inter-cluster transmission uses a multi-hop method. Due to the long transmission path and large amount of data transmitted between CHs, a Dynamic Coding Coordination (DCC) transmission mechanism is needed to improve transmission reliability. The specific steps are as follows:
[0064] 7.1) Assuming that data transmission between adjacent hops will not affect each other, let the i-th hop node be CH. i Let its next hop be CH. i+1 CH i Calculate the cooperative communication cost of all available cooperative nodes within its local storage space. :
[0065] (15)
[0066] in, The energy consumption of cooperative communication when node i is selected as the cooperative node. CH represents the remaining energy of node i. i The node with the lowest cost is selected as the cooperative node for cooperative communication. The cooperative node is denoted as R. i+1 ;
[0067] 7.2) During the i-th hop transmission process, i.e., "CH i -R i+1 -CH i+1 Assuming the underwater acoustic communication physical layer uses OFDM modulation technology, let CH i Divide the data packet into N bl Blocks, each block modulated on an OFDM symbol; these N bl Each OFDM symbol block will be CH i Send, received by collaborating node R i+1 and the next hop CH i+1 take over;
[0068] 7.3) Collaborative Node R i+1 Initially in the listening phase, R... i+1 They will all attempt to decode the entire data packet using accumulated OFDM blocks; in CH i If the transmitted data packet is successfully decoded before the transmission ends, R i+1 It will switch to the collaborative communication phase;
[0069] 7.4) Define N li For R i+1 The number of OFDM blocks required to successfully decode the data packet, where Δ is an integer required for OFDM symbol block synchronization design; for the previous A symbol block, R i+1 CH does not participate in transmission as a cooperating node. i+1 Only CH received iTransmitted OFDM blocks; for the later A symbol block, employing a repetitive redundancy (RR) DCC cooperative scheme, with cooperative node R... i+1 Generate and transmit with CH i The same OFDM block, CH i+1 Can receive from CH i and R i+1 The superimposed signal; in two different stages, for node CH i+1 The input-output relationship of the received l-th OFDM symbol block can be expressed as:
[0070] (16)
[0071] in, It is CH i+1 The symbol block vector received on K subcarriers, It is the symbol block vector transmitted on K subcarriers. CH i With CH i+1 The channel matrix between, R represents i+1 With CH i+1 The channel matrix between them Represents the environmental noise vector;
[0072] 7.5) CH i+1 After successfully receiving the data packet, it continues to transmit to the next-hop node specified in the local storage space until the data packet reaches the Sink node, at which point the transmission is complete.
[0073] This invention can effectively balance the node load in an underwater acoustic network, thereby reducing network energy consumption and extending the overall network lifespan and lifetime. Considering the need for priority transmission of important data in underwater data collection tasks, this invention introduces the concept of data hierarchy and adopts different clustering strategies for different data levels to ensure that high-importance data is transmitted in real time and quickly. At the same time, combined with the DCC transmission mechanism, the aim is to obtain a clustering method that can balance network energy consumption, improve communication reliability, network throughput and the overall network lifetime.
[0074] The present invention has the following outstanding advantages:
[0075] 1) Introduce the concept of data hierarchy, classify the importance of data, and select different clustering strategies for different data importance to ensure that important data can be quickly transmitted to the Sink node through the shortest path to meet the needs of underwater data collection tasks.
[0076] 2) An underwater acoustic cooperative communication mechanism is introduced. In the process of selecting the cluster head in the particle swarm optimization algorithm, the available cooperative nodes are taken as one of the optimization objectives to ensure the rationality of the allocation of cooperative nodes, and a practical and feasible method for clustering underwater acoustic cooperative networks is provided.
[0077] 3) Combining DCC transmission mechanism in the underwater acoustic network data transmission process improves network communication reliability and network throughput. Attached Figure Description
[0078] Figure 1 This is a scenario diagram of the intelligent non-uniform clustering method for underwater acoustic cooperative networks based on data hierarchical classification, as presented in this invention.
[0079] Figure 2 This is a flowchart of the particle swarm head selection algorithm for the intelligent non-uniform clustering method based on data-hierarchical underwater acoustic cooperative networks of the present invention.
[0080] Figure 3 This is a flowchart of the particle update process for the particle swarm head selection algorithm in the intelligent non-uniform clustering method of underwater acoustic cooperative networks based on data hierarchical classification, as presented in this invention.
[0081] Figure 4 This is a schematic diagram of the dynamic coding cooperation (DCC) transmission mechanism of the intelligent non-uniform clustering method for underwater acoustic cooperative networks based on data hierarchical data, as presented in this invention.
[0082] Figure 5 This is a comparison chart of the important data transmission delays of the intelligent non-uniform clustering method for underwater acoustic cooperative networks based on data hierarchy and the comparative method (UCPSO method) of this invention.
[0083] Figure 6 The present invention presents a comparison chart of the node average energy consumption of the intelligent non-uniform clustering method for underwater acoustic cooperative networks based on data hierarchical classification and the comparative method (UCPSO method).
[0084] Figure 7 This is a comparison chart of network throughput between the data-hierarchical underwater acoustic cooperative network intelligent non-uniform clustering method of this invention and a comparative method (UCPSO method). Detailed Implementation
[0085] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments.
[0086] like Figure 1 As shown, the embodiments of the present invention include the following steps:
[0087] 1) such as Figure 1Assuming an underwater environment with a range of 1000m×1000m×1000m, the underwater acoustic nodes constituting the underwater acoustic network are randomly distributed, denoted as node 1, node 2, node 3, ..., node 20. All underwater acoustic nodes have the same initial energy and can control their own transmission power. A sink node is deployed on the water surface. This node is equipped with both an acoustic modem and a radio modem. It can communicate with the underwater acoustic nodes through acoustic links and also transmit information to the shore-based control center through radio communication.
[0088] Among them, the Sink node has strong storage and computing capabilities, and can store and process the location and energy information of all underwater acoustic nodes. It is also responsible for calculating the clustering of all underwater acoustic nodes, and can be powered by solar energy without considering energy consumption issues.
[0089] 2) such as Figure 2 The Sink node selects the cluster head node (CH) using the particle swarm optimization algorithm, and the steps are as follows:
[0090] 2.1) Particle Encoding: In the particle swarm optimization algorithm, a particle PS is a sequence of numbers consisting of "0" and "1". Each bit in the sequence corresponds to a surviving underwater acoustic node. A "1" at the corresponding position indicates that the underwater acoustic node is a cluster head node CH, and a "0" indicates that the underwater acoustic node is a non-cluster head node nCH. Each particle represents a possible cluster head selection scheme.
[0091] 2.2) Particle Initialization: During data transmission, the energy consumption of CH is much greater than that of nCH. Therefore, it is necessary to select underwater acoustic nodes with higher remaining energy as CHs. To protect underwater acoustic nodes with lower remaining energy and ensure that a sufficient number of underwater acoustic nodes participate in cluster head selection, all underwater acoustic nodes are first sorted in descending order of remaining energy. The underwater acoustic nodes with the highest remaining energy (top 60%) are selected as candidate nodes, and then selected according to the proportion P. CH = 0.1 Randomly generate cluster head nodes to create a particle population PS i This represents one possible cluster head selection scheme; when generating the particle population, to ensure a more uniform distribution of CHs, interval thresholds are set between CHs. :
[0092] (1)
[0093] in, To determine the number of surviving underwater acoustic nodes, the distance between CHs must be greater than a certain threshold during particle initialization and particle update. The conditions. In this embodiment, the number of surviving underwater acoustic nodes is 20, therefore the interval threshold. ;
[0094] 2.3) Fitness Function: The fitness function is used to judge the performance of cluster head selection results, and is expressed as follows:
[0095] (2)
[0096] in, and Let represent the optimization objective functions related to residual energy, transmission energy consumption, and cooperative gain, respectively. , and Let the weights be factors, satisfying the following relationship:
[0097] (3)
[0098] To ensure that the remaining energy of the selected CH is as high as possible, optimize the objective function. Defined as:
[0099] (4)
[0100] in, The average remaining energy of all underwater acoustic nodes. The average remaining energy of particles labeled CH is the average remaining energy of CH. The higher the average remaining energy of CH, the better the optimization objective function. The smaller it is;
[0101] Optimize objective function The performance evaluation of cluster head selection results, which incorporates single-round transmission energy consumption, is defined as follows:
[0102] (5)
[0103] in, The total energy consumption for a single round of transport according to the CH and nCH labeled by the particles;
[0104] Available cooperative nodes are defined as underwater acoustic nodes located between two CHs or between a CH and a Sink node on the transmission path, where there is sufficient remaining energy. In the fitness function formula (2), cooperative communication is taken into account for performance, thus defining the optimization objective function. for:
[0105] (6)
[0106] in, For the number of CHs, RN i For node CH during inter-cluster transmission i The number of times N acts as a relay node Ri For node CH iThe number of available cooperative nodes on the next hop path. The number of times RN acts as a relay during data collection. i More CH i The larger the amount of data transmitted, the more available cooperative nodes are needed to ensure communication reliability and prevent the cooperative nodes from consuming too much energy.
[0107] because , and The values differ significantly, so the Sigmoid function is needed first. Map it to a value between 0 and 1, and then use a weighting factor. , and Weighted calculation. The smaller the value of the particle fitness function, the higher the average remaining energy of the CH in the clustering result corresponding to the particle, the less energy consumption per round of transmission, and the more reasonable allocation of cooperative nodes;
[0108] 2.4) After particle initialization, calculate and compare the fitness values of all particles to obtain the historical best solution set for all particles. And the globally optimal particle GB, where HB i It is particle PS i The corresponding historical best;
[0109] 2.5) As Figure 3 The particles are updated by tracking the current historical optimum and the global optimum: a maximum number of iterations is set. For particles PS i In the In the round of iteration:
[0110] (6)
[0111] in, Indicates the first Particle PS after the end of the round iteration i ,function Indicated by Probabilistic mutation of particles:
[0112] (7)
[0113] in, The mutation operator represents the mutation of particle PS. i Perform a mutation operation, causing one CH node in the candidate node to mutate to nCH, and one nCH node to mutate to CH. (Probability) With the current particle The degree of difference between the global optimal particle GB and related:
[0114] (8)
[0115]
[0116] in, and Defined respectively The maximum and minimum values are set to 0.6 and 0.1 respectively, and the operator... Used to calculate particle PS i The number of bits in different encoding bits of the globally optimal particle GB. The dimension of the particle, i.e., the number of surviving underwater acoustic nodes, is 20;
[0117] In formula (6), the function and They represent respectively with and Probability of particles PS i Perform cross operations:
[0118] (9)
[0119]
[0120] (10)
[0121]
[0122] Among them, the cross operator Indicates the relationship between particles PS i And the best historical particle HB i Perform cross operations, proportional to P H = 0.1 Extract particle fragments for crossover. and These represent the current iteration number and the maximum iteration number, respectively.
[0123] 2.6) In step 2.5), after each mutation or crossover operation, the particle needs to be checked to see if it meets the interval threshold condition of formula (1). If it does not meet the condition, the mutation or crossover operation is repeated until the condition is met. At the same time, the fitness value of the updated particle is calculated according to formula (2). If the particle fitness value is larger, it means that the cluster head selection result corresponding to the particle is worse. In order to ensure particle diversity, the probability is still used. Accept the particle as a new particle:
[0124] (11)
[0125] If a particle's fitness value is smaller, it is directly accepted as a new particle. After all particles have completed one round of updates, the historical best solution set HB and the globally best particle GB are updated, and the iteration continues.
[0126] 2.7) After reaching the maximum number of iterations, the globally optimal particle GB is used as the result of cluster head selection. Assume that in... Figure 1 In the process, after the particle swarm optimization algorithm completes its iterations, nodes 4, 10, 13, and 16 are selected as CHs, and the Sink node will broadcast the results to these nodes.
[0127] 3) After selecting the cluster head node, it is necessary to determine the inter-cluster multi-hop transmission path, as follows:
[0128] 3.1) Arrange the CHs in ascending order of depth. In this example, the order is node 10, node 4, node 13, and node 16. Starting with the shallowest node 10, calculate the next-hop relay for each CH. The next hop for node 10 must be a sink node. For deeper CHs... i (That is, nodes 4, 13, and 16) can select a shallower CH or Sink node as the next-hop relay and calculate the transmission cost. :
[0129] (12)
[0130] in, The total energy consumption along the transmission path is calculated by CH. i The energy consumption generated by transmitting a 1-bit data packet to the Sink node along the selected path is obtained. After selecting this path, CH i The number of jumps, The weight values are those ranging from (0,1). CH i Choose to The smallest CH or Sink node is used as the next-hop relay. Taking node 13 as an example, the transmission cost is calculated using formula (12). Among the shallower nodes 4, 10, and Sink, node 10 corresponds to... The smallest value is selected, therefore node 10 is chosen as its next-hop relay.
[0131] 3.2) After the deepest node 16 completes the next hop selection, the inter-cluster multi-hop transmission path is completed. The number of times each CH acts as a relay node is recorded (RN). At the same time, the Sink node also needs to determine all available cooperative nodes on the next hop path of each CH based on the current node's location information, and broadcast the next hop node and available cooperative node information to the corresponding CH.
[0132] 3.3) After receiving the broadcast information from the Sink node, the CH records the information of the next-hop node and available cooperating nodes in its local storage space;
[0133] 4) Initial cluster range for each CH As determined by equation (1), in order to balance the energy consumption of each CH and extend the network lifetime, it is necessary to change the cluster range according to the actual situation and define the cluster range of CH. :
[0134] (13)
[0135] in, and Representing nodes CH respectively i The remaining energy and its distance from the Sink node, This is an approximation of the average energy consumption of CH after one round of data transmission. This represents the maximum distance between nodes in the network and the sink node. , and All values are weights within the range (0,1). Furthermore, a Sink node can also serve as a cluster head node, with a fixed cluster range. ;
[0136] 5) Data Importance It can be divided into The levels represent unimportant, general, and important data, respectively. For each nCH node, the Sink node calculates the clustering cost according to the different data levels. :
[0137] (14)
[0138] in, Indicates nCH to CH i distance, For CH i The number of hops is used to select the cluster head node for each nCH node based on the minimum clustering cost under different data grading conditions: for unimportant data and general data, the cluster range selected by nCH is determined. Larger and closer CHs serve as cluster head nodes, forming a non-uniform clustering structure. The clustering result is called non-uniform clustering, and the corresponding CHs are denoted as CH. unequal For critical data, nCH selects cluster head nodes primarily based on their distance from the CH and the number of hops in the CH, ensuring the shortest transmission path. The clustering result is called shortest path clustering, and the corresponding CH is denoted as CH. shortest The Sink node broadcasts the clustering results, and the nCH node stores the information of the CH node corresponding to the two clustering results in its local storage space.
[0139] 6) During data transmission, due to the short distance between the CH and nCH within the same cluster, it can be assumed that the CH can successfully receive and decode all signals from the nCH without the assistance of cooperating nodes. When an nCH node collects data, it first determines the data classification of the data. ,like Then select CH unequal Transmit; if Then select CH shortest Transmit;
[0140] 7) Inter-cluster transmission uses a multi-hop method. Due to the long transmission path and large data volume between CHs, a Dynamic Coding Coordination (DCC) transmission mechanism is needed to improve transmission reliability, such as... Figure 4 The specific steps are as follows:
[0141] 7.1) Assuming that data transmission between adjacent hops will not affect each other, let the i-th hop node be CH. i Let its next hop be CH. i+1 CH i Calculate the cooperative communication cost of all available cooperative nodes within its local storage space. :
[0142] (15)
[0143] in, The energy consumption of cooperative communication when node i is selected as the cooperative node. CH represents the remaining energy of node i. i The node with the lowest cost is selected as the cooperative node for cooperative communication. The cooperative node is denoted as R. i+1 ;
[0144] 7.2) Taking the transmission process from node 13 to node 10 as an example, node 12 is selected as the cooperating node using formula (15): Assuming that the underwater acoustic communication physical layer adopts OFDM modulation technology, node 13 divides the data packet into N bl Each block is modulated on an OFDM symbol. This N bl Each OFDM symbol block will be sent by node 13 and received by cooperating node 12 and next-hop node 10;
[0145] 7.3) Collaborating node 12 is initially in the listening phase. Each time it receives a new OFDM symbol block, node 12 attempts to decode the entire data packet using the accumulated OFDM blocks. In CH i If the transmitted data packet is successfully decoded before the transmission ends, node 12 will switch to the cooperative communication phase.
[0146] 7.4) Define Nli The number of OFDM blocks required for node 12 to successfully decode data packets, where Δ is an integer required for OFDM symbol block synchronization design. For the previous In the case of a symbol block, node 12 does not participate in the transmission as a cooperating node, and node 10 only receives the OFDM block transmitted by node 13; for the subsequent... For each OFDM symbol block, a repetitive redundancy (RR) DCC cooperative scheme is used. Cooperative node 12 generates and transmits the same OFDM block as node 13. Node 10 can receive the superimposed signals from nodes 13 and 12. In two different stages, for node 10 to receive the l-th OFDM symbol block, its input-output relationship can be expressed as:
[0147] (16)
[0148] in, It is the symbol block vector received by node 10 on K subcarriers. It is the symbol block vector transmitted on K subcarriers. This represents the channel matrix between node 13 and node 10. This represents the channel matrix between cooperating nodes 12 and 10. Represents the environmental noise vector;
[0149] 7.5) After node 10 successfully receives the data packet, it continues to transmit it to the next-hop node specified in the local storage space until the data packet reaches the Sink node, at which point the transmission is complete.
[0150] To verify the advantages of the data-hierarchical underwater acoustic cooperative network intelligent non-uniform clustering method described in this invention compared with the comparative method (i.e., the particle swarm optimization-based underwater acoustic sensor network non-uniform clustering (UCPSO) method (R. Hou, et al. AnUnequal Clustering Method Based on Particle Swarm Optimization in UnderwaterAcoustic Sensor Networks[J]. IEEE Internet of Things Journal, 2022,9(24):25027-25036.)) in terms of important data transmission latency, network energy consumption, and network throughput, the method was verified by computer simulation below:
[0151] Arrange 30-90 underwater acoustic nodes within a 1000m×1000m×1000m area, and place a sink node at (500m, 500m, 0). The operating frequency of each underwater acoustic node is... Transmission rate Assuming a single OFDM symbol block size is 108 bits, each data packet contains 8 OFDM information symbol blocks and 6 OFDM check symbol blocks.
[0152] Without considering retransmissions (i.e., setting the transmission success rate between CHs to 100%), the latency of important data transmissions in the network varies with the number of nodes, for example... Figure 5 As shown in the figure. With the maximum number of retransmissions set to 1, the average node energy consumption and network throughput are compared as follows, varying with the number of nodes: Figure 6 and Figure 7 As shown in the figure. The simulation results are shown in Table 1.
[0153] Table 1
[0154]
[0155] As shown in Table 1, the method of this invention can effectively shorten the transmission latency of important data. When the number of nodes is 30, 50, 70, and 90, the transmission latency of important data is shortened by 13.37%, 21.58%, 13.30%, and 14.56% compared to the comparative method, respectively. Simultaneously, the method of this invention can significantly reduce node energy consumption and improve network throughput: with the number of nodes, the average node energy consumption decreases by 34.44%, 33.20%, 16.70%, and 14.45%, respectively; the network throughput increases by more than 20%, specifically 35.45%, 63.90%, 26.39%, and 22.59%, respectively. Therefore, this invention has significant advantages in terms of transmission latency of important data, network energy consumption, and network throughput.
[0156] This invention comprehensively considers node energy, transmission energy consumption, and available cooperative nodes. It employs a particle swarm optimization algorithm for uneven clustering, which balances the energy consumption of cluster head nodes, alleviates the "energy hotspot" problem, and improves network lifetime. The introduction of a Dynamic Coding Cooperative (DCC) transmission mechanism effectively enhances data transmission reliability and network throughput. Furthermore, the invention introduces a data hierarchy concept, selecting different clustering strategies for different data levels to ensure real-time transmission of high-importance data. This invention effectively balances network energy consumption, improves communication reliability, network throughput, and overall network lifetime.
[0157] The above embodiments are merely preferred embodiments of the present invention and should not be considered as limiting the scope of the present invention. All equivalent variations and improvements made within the scope of the present invention should still fall within the patent coverage of the present invention.
Claims
1. A method for intelligent non-uniform clustering of underwater acoustic cooperative networks based on data hierarchical classification, characterized in that... Includes the following steps: 1) Assuming in m× m× In an underwater environment within a range of m, the underwater acoustic nodes constituting the underwater acoustic network are randomly distributed, denoted as node 1, node 2, node 3, ..., node k. All underwater acoustic nodes have the same initial energy and can control their own transmission power. A sink node is deployed on the water surface. This node is equipped with both an acoustic modem and a radio modem, and can communicate with the underwater acoustic nodes through an acoustic link and also transmit information to the shore-based control center through radio communication. Among them, the Sink node has strong storage and computing capabilities, and can store and process the location and energy information of all underwater acoustic nodes. It is also responsible for calculating the clustering of all underwater acoustic nodes, and is powered by solar energy, so energy consumption is not a concern. 2) The Sink node selects the cluster head node (CH) using the particle swarm optimization algorithm. The steps are as follows: 2.1) Particle Encoding: In the particle swarm optimization algorithm, a particle PS is a sequence of numbers consisting of "0" and "1". Each bit in the sequence corresponds to a surviving underwater acoustic node. A "1" at the corresponding position indicates that the underwater acoustic node is a cluster head node CH, and a "0" indicates that the underwater acoustic node is a non-cluster head node nCH. Each particle represents a possible cluster head selection scheme. 2.2) Particle Initialization: During data transmission, if the energy consumption of CH is greater than nCH, the underwater acoustic node with higher remaining energy is selected as CH; all underwater acoustic nodes are sorted in descending order of remaining energy, and the node with the highest remaining energy is selected. % of the underwater acoustic nodes are selected as candidate nodes, according to the proportion P CH Randomly generate cluster head nodes to create a particle population. Among them, PS i This represents one possible cluster head selection scheme; When generating the particle population, set interval thresholds between CHs. : (1) in, To determine the number of surviving underwater acoustic nodes, the distance between CHs must be greater than a threshold value during particle initialization and particle update. conditions; 2.3) Fitness Function: The fitness function is used to judge the performance of cluster head selection results, and is expressed as follows: (2) in, , and Let represent the optimization objective functions related to residual energy, transmission energy consumption, and cooperative gain, respectively. , and Let be the weighting factors, satisfying the following relationship: (3) Optimize objective function Defined as: (4) in, The average remaining energy of all underwater acoustic nodes. The average remaining energy of particles labeled CH is the average remaining energy of CH. The higher the average remaining energy of CH, the better the optimization objective function. The smaller it is; Optimize objective function The performance evaluation of cluster head selection results, which incorporates single-round transmission energy consumption, is defined as follows: (5) in, The total energy consumption for a single round of transport according to the CH and nCH labeled by the particles; Available cooperative nodes are defined as underwater acoustic nodes located between two CHs or between a CH and a Sink node on the transmission path, where there is sufficient remaining energy. In the fitness function formula (2), cooperative communication is taken into account for performance, and the optimization objective function is defined. for: (6) in, For the number of CHs, RN i For node CH during inter-cluster transmission i The number of times N acts as a relay node Ri For node CH i The number of available cooperative nodes on its next-hop path; the number of times it acts as a relay during data collection (RN). i More CH i The larger the amount of data transmitted; because , and The values differ significantly, which can be addressed using the Sigmoid function. Map it to a value between 0 and 1, and then use a weighting factor. , and Weighted operations; 2.4) After particle initialization, calculate and compare the fitness values of all particles to obtain the historical best solution set for all particles. And the globally optimal particle GB, where HB i It is particle PS i The corresponding historical best; 2.5) Particles are updated by tracking the current historical optimum and the global optimum: For particle PS i In the In the round of iteration: (6) in, Indicates the first Particle PS after the end of the round of iteration i ,function Indicates Probabilistic mutation of particles: (7) in, The mutation operator represents the mutation of particle PS. i Perform a mutation operation, causing one CH node in the candidate node to mutate into nCH, and one nCH node to mutate into CH; probability. With the current particle The degree of difference between the global optimal particle GB and related: (8) in, and Defined respectively The maximum and minimum values, the operator Used to calculate particle PS i The number of bits in different encoding bits of the globally optimal particle GB. The dimension of the particle is the number of surviving underwater acoustic nodes. In formula (6), the function and They represent respectively with and Probability of particles PS i Perform cross operations: (9) (10) Among them, the cross operator Indicates the relationship between particles PS i And the best historical particle HB i Perform cross operations, proportional to P H Extract particle fragments and cross them. and These represent the current iteration number and the maximum iteration number, respectively. 2.6) In step 2.5), after each mutation or crossover operation, the particle needs to be checked to see if it meets the interval threshold condition of formula (1). If it does not meet the condition, the mutation or crossover operation is repeated until the condition is met. At the same time, the fitness value of the updated particle is calculated according to formula (2), still based on probability. Accept the particle as a new particle: (11) If a particle's fitness value is smaller, the particle is directly accepted as a new particle. After all particles have completed one round of updates, the historical best solution set HB and the globally best particle GB are updated, and the iteration continues. 2.7) After reaching the maximum number of iterations, the globally optimal particle GB is used as the result of cluster head selection, and the Sink node broadcasts the result to all CHs; 3) After selecting the cluster head node, it is necessary to determine the inter-cluster multi-hop transmission path, as follows: 3.1) Arrange the CHs in ascending order of depth. Starting with the shallowest CH, calculate the next-hop relay for each CH. The next hop of the shallowest CH must be a Sink node. For deeper CHs... i Select a CH or Sink node shallower than itself as the next-hop relay and calculate the transmission cost. : (12) in, The total energy consumption along the transmission path is calculated by CH. i The energy consumption generated by transmitting a 1-bit data packet to the Sink node along the selected path is obtained. After selecting this path, CH i The number of jumps, The weight values are those ranging from (0,1); CH i Choose to The smallest CH or Sink node is used as the next-hop relay; 3.2) Once the deepest CH completes the next hop selection, the inter-cluster multi-hop transmission path is constructed. The number of times each CH acts as a relay node (RN) is recorded. At the same time, the Sink node also needs to determine all available cooperative nodes on the next hop path of each CH based on the current node's location information, and broadcast the next hop node and available cooperative node information to the corresponding CH. 3.3) After receiving the broadcast information from the Sink node, the CH records the information of the next-hop node and available cooperating nodes in its local storage space; 4) Initial cluster range for each CH The cluster range of CH is defined by equation (1) and adjusted according to actual conditions. : (13) in, and Representing nodes CH respectively i The remaining energy and its distance from the Sink node, This is an approximation of the average energy consumption of CH after one round of data transmission. This represents the maximum distance between nodes in the network and the sink node. , and All values are weights within the range (0,1); additionally, the Sink node can also serve as a cluster head node, with its cluster range fixed at [value missing]. ; 5) Data Importance Divided into The levels represent unimportant, general, and important data, respectively. For each nCH node, the Sink node calculates the clustering cost according to the different data levels. : (14) in, Indicates nCH to CH i distance, For CH i The number of hops is used to select the cluster head node for each nCH node based on the minimum clustering cost under different data grading conditions: for unimportant data and general data, the cluster range selected by nCH is determined. Larger and closer CHs serve as cluster head nodes, forming a non-uniform clustering structure. The clustering result is called non-uniform clustering, and the corresponding CHs are denoted as CH. unequal For critical data, nCH selects cluster head nodes primarily based on their distance from the CH and the number of hops in the CH, ensuring the shortest transmission path. The clustering result is called shortest path clustering, and the corresponding CH is denoted as CH. shortest The Sink node broadcasts the clustering results, and the nCH node stores the information of the CH node corresponding to the two clustering results in its local storage space. 6) During data transmission, within the same cluster, due to the short distance between CH and nCH, it is assumed that CH can successfully receive and decode all signals from nCH without the assistance of cooperating nodes; when an nCH node collects data, it first determines the data classification of the data. ,like Then select CH unequal Transmit; if Then select CH shortest Transmit; 7) Inter-cluster transmission adopts a multi-hop method. Due to the long transmission path and large amount of data transmitted between CHs, the Dynamic Coding Cooperative (DCC) transmission mechanism is used to improve transmission reliability. The specific steps are as follows: 7.1) Assuming that data transmission between adjacent hops will not affect each other, let the i-th hop node be CH. i Let its next hop be CH. i+1 CH i Calculate the cooperative communication cost of all available cooperative nodes within its local storage space. : (15) in, The energy consumption of cooperative communication when node i is selected as the cooperative node. CH represents the remaining energy of node i. i The node with the lowest cost is selected as the cooperative node for cooperative communication. The cooperative node is denoted as R. i+1 ; 7.2) During the i-th hop transmission process, i.e., "CH i -R i+1 -CH i+1 Assuming the underwater acoustic communication physical layer uses OFDM modulation technology, let CH i Divide the data packet into N bl Blocks, each block modulated on an OFDM symbol; these N bl One OFDM symbol block was CH i Send, received by collaborating node R i+1 and the next hop CH i+1 take over; 7.3) Collaborative Node R i+1 Initially in the listening phase, R... i+1 They will all attempt to decode the entire data packet using accumulated OFDM blocks; in CH i If the transmitted data packet is successfully decoded before the transmission ends, R i+1 It will switch to the collaborative communication phase; 7.4) Define N li For R i+1 The number of OFDM blocks required to successfully decode the data packet, where Δ is an integer required for OFDM symbol block synchronization design; for the previous A symbol block, R i+1 CH does not participate in transmission as a cooperating node. i+1 Only CH received i Transmitted OFDM blocks; for the later A symbol block, employing a repetitive redundancy (RR) DCC cooperative scheme, with cooperative node R... i+1 Generate and transmit with CH i The same OFDM block, CH i+1 Received from CH i and R i+1 The superimposed signal; in two different stages, for node CH i+1 The input-output relationship of the received l-th OFDM symbol block is as follows: (16) in, It is CH i+1 The symbol block vector received on K subcarriers, It is the symbol block vector transmitted on K subcarriers. CH i With CH i+1 The channel matrix between, R represents i+1 With CH i+1 The channel matrix between, Represents the environmental noise vector; 7.5) CH i+1 After successfully receiving the data packet, it continues to transmit to the next-hop node specified in the local storage space until the data packet reaches the Sink node, at which point the transmission is complete.
Citation Information
Patent Citations
Wireless sensor network clustering collaborative routing algorithm based on cooperative game
CN105636143A
Dynamic calculation cluster head routing method for underwater acoustic network based on Q learning and data classification
CN115002865A
