Wireless ad hoc network data aggregation method and system

Abstract

The invention provides a wireless ad hoc network data aggregation method, which comprises the step of performing multi-round aggregation on data acquired by each sensor based on a preset time interval, each round of convergence mode comprises the following steps: S1, determining the level to which each sensor belongs according to the distance between each sensor in the wireless sensor network and the convergence node and a preset range value of the distance between each of a plurality of levels and the convergence node, and taking the level closest to the convergence node as the uppermost level; s2, dividing all sensors of each layer into a plurality of clusters according to a preset division rule, and determining a sensor serving as a cluster head and a sensor serving as a cluster member in each cluster; s3, converging data collected by all sensors in each cluster to a cluster head of the cluster to obtain converged data of each cluster; and S4, transmitting the aggregation data to the aggregation node layer by layer according to the cluster heads, including transmitting the aggregation data of the clusters not on the uppermost layer to the nearest cluster head on the upper layer and transmitting the aggregation data of the clusters on the uppermost layer to the aggregation node.

Description

Technical Field

[0001] The present invention relates to the field of wireless communication technology, in particular to the field of wireless sensor data acquisition networks, and more particularly to a wireless ad hoc network data aggregation method and system. Background Art

[0002] With the rise of the industrial Internet, building a low-energy data collection network has become a focus of research and industry. Wireless sensor networks (WSNs) can achieve self-organizing networks and communications between devices without the support of fixed infrastructure, and are widely used in environmental monitoring, equipment tracking, logistics tracking and other fields. However, due to the limited energy of WSNs, how to reduce network transmission energy consumption and extend network life has become a challenge.

[0003] In order to meet the above challenges, clustering is a common technical means of aggregation algorithm when aggregating data in WSN. Its core idea is to divide nodes into different clusters, and each cluster is managed by a cluster head (CH). Nodes in the cluster only need to send data to the cluster head, which compresses and aggregates the data and sends it to the gateway node. In this way, the overall communication energy consumption of the network can be reduced.

[0004] However, despite the benefits of clustering, there are still many challenges in practical applications. When selecting cluster heads, it is necessary to comprehensively evaluate multiple factors such as the remaining energy of the nodes, the geographical location, and the communication load. At the same time, the planning of the transmission path will also affect the energy consumption. Inappropriate path planning may increase the average transmission distance, resulting in increased energy consumption, thereby shortening the life of the network. The size of the cluster is also a factor that cannot be ignored. Inappropriate cluster size may increase the workload of the cluster head and cause the cluster head to die prematurely. Therefore, various factors need to be considered comprehensively during the design to ensure the optimization of energy consumption.

[0005] The existing wireless sensor network ad hoc aggregation method mainly adopts a single-level architecture. In this architecture, each cluster in the network is led by a cluster head node, which is responsible for collecting, integrating and processing data from each node in the cluster. Ordinary nodes are responsible for collecting environmental data and transmitting the data directly to the cluster head node to which they belong. After receiving the data, the cluster head node aggregates it to reduce the network transmission volume and energy consumption, and then transmits the aggregated data to the aggregation node in the network through a single hop, finally completing the data aggregation and analysis tasks. However, as the network coverage area expands, the overall network transmission energy consumption rises rapidly in the clustering of the traditional single-level architecture due to the increase in the average transmission distance between nodes. In addition, the distances from each cluster to the aggregation node are different, which leads to the problem of premature demise of clusters that are farther away. In order to reduce communication energy consumption, the existing clustering strategy mainly focuses on minimizing the distance from the nodes in the cluster to the cluster head. However, this method often ignores the relative position relationship between the cluster head node and other important nodes in the network (such as data aggregation points or target nodes), which makes the cluster head overloaded or increases the transmission distance between clusters, thus affecting the energy efficiency of the entire network.

[0006] Based on the above analysis, as the network coverage area increases, the existing single-layer architecture method has significant problems including: on the one hand, the selection of cluster heads is usually based on load balancing and transmission energy consumption considerations, which leads to uneven distribution of cluster heads in a large-scale network, so that some cluster heads may bear too many data transmission tasks, thus consuming energy quickly and shortening the life of nodes quickly. On the other hand, due to the single-hop transmission from the cluster head to the sink node, the nodes far away from the sink node are prone to overload, which not only increases energy consumption, but also may cause these nodes to run out of energy in a short period of time and cannot continue to operate effectively. Both of the above problems limit the sustainability and long-term stability of the network.

[0007] It should be noted that this background technology is only used to introduce the relevant information of the present invention to help understand the technical solution of the present invention, but it does not mean that the relevant information is necessarily the prior art. If there is no evidence that the relevant information has been disclosed before the application date of the present invention, the relevant information shall not be regarded as the prior art. Summary of the invention

[0008] Therefore, the object of the present invention is to overcome the above-mentioned defects of the prior art and provide a method and system for data aggregation in a wireless ad hoc network.

[0009] The objective of the present invention is achieved through the following technical solutions:

[0010] According to a first aspect of the present invention, a method for data aggregation in a wireless ad hoc network is provided, which is applied to a wireless sensor network, wherein the wireless sensor network includes a plurality of sensors, and the positions of at least some of the sensors are changeable. The method comprises: performing multiple rounds of aggregation on data collected by each sensor based on a predetermined time interval, wherein each round of aggregation process is performed in the following manner:

[0011] S1. Determine the layer to which each sensor belongs according to the distance between all sensors in the wireless sensor network and the sink node and the distance range values ​​between multiple preset layers and the sink node, wherein the layer closest to the sink node is the top layer; S2. Divide all sensors in each layer into multiple clusters according to preset division rules, and determine a sensor as a cluster head and a sensor as a cluster member in each cluster; S3. Aggregate the data collected by all sensors in each cluster to the cluster head of the cluster to obtain the aggregated data of each cluster; S4. Transmit the aggregated data to the sink node layer by layer according to the cluster head, including: transmitting the aggregated data of the clusters not in the top layer to the cluster head closest to the upper layer, and transmitting the aggregated data of the clusters in the top layer to the sink node.

[0012] In some embodiments of the present invention, in S2, the preset division rule includes: using a soft K-means clustering algorithm to perform clustering according to the position of each sensor in each layer, to obtain multiple cluster center positions corresponding to each layer and the membership probability values ​​of each sensor in the layer belonging to each cluster; according to the membership probability values ​​of each sensor belonging to each cluster, determining the cluster to which each sensor belongs, so as to divide all sensors of the layer into multiple clusters; wherein, the closer the layer is to the sink node, the greater the number of clusters divided.

[0013] In some embodiments of the present invention, the method of obtaining the multiple cluster center positions corresponding to each layer and the membership probability value of each sensor in the layer belonging to each cluster includes:

[0014] Each layer is divided multiple times, and each division is performed in the following manner: multiple cluster center positions are obtained, wherein the first time a plurality of randomly selected cluster center positions are used, and each time after the first time a plurality of cluster center positions calculated last time are used; according to the multiple cluster center positions and the positions of each sensor, the membership probability value of each sensor belonging to each cluster is determined; according to the multiple cluster center positions, the position of each sensor and the membership probability value, the multiple cluster center positions for the next time are calculated with minimizing the distance within each cluster as the optimization objective function, until the soft K-means clustering algorithm converges, and the final multiple cluster center positions and the membership probability value of each sensor in the layer belonging to each cluster are obtained.

[0015] In some embodiments of the present invention, the optimization objective function is as follows:

[0016] ,

[0017] in, represents minimizing the distance within each cluster, represents the distance within each cluster, , represents the location of all sensors in each layer, Indicates the position of the first sensor, Indicates the position of the second sensor, Indicates The location of the sensor, , Indicates The sensor belongs to The membership probability value of a cluster, , is the weight parameter of the membership probability value of each sensor, represents the number of clusters in each level, represents the total number of clusters in each level, Indicates The location of the sensor, Indicates the sensor number, represents the total number of sensors in each layer, Indicates The cluster center location, Indicates The distance between each sensor and the cluster center, Represents the set of all cluster center locations in each level.

[0018] In some embodiments of the present invention, the method for determining the convergence of the soft K-means clustering algorithm includes: calculating the membership probability change value based on the difference between the membership probability value of each sensor belonging to each cluster obtained this time and the membership probability value of each sensor belonging to each cluster obtained last time; obtaining a preset probability threshold, and determining that the soft K-means clustering algorithm converges when the membership probability change value is less than the preset probability threshold.

[0019] In some embodiments of the present invention, the method for determining the convergence of the soft K-means clustering algorithm includes: calculating the position distance change value based on the difference between the multiple cluster center positions obtained this time and the multiple cluster center positions obtained last time; obtaining a preset distance threshold, and when the position distance change value is less than the preset distance threshold, determining that the soft K-means clustering algorithm has converged.

[0020] In some embodiments of the present invention, the method of dividing into multiple clusters includes: according to the membership probability value of each sensor belonging to each cluster, dividing the sensor into the cluster corresponding to the maximum membership probability value to obtain multiple clusters initially divided at the level; judging whether the number of sensors contained in each of the multiple clusters initially divided is balanced; if not, balancing the number of sensors in the multiple clusters initially divided to obtain multiple clusters finally divided; otherwise, using the multiple clusters initially divided as the multiple clusters finally divided.

[0021] In some embodiments of the present invention, in S2, the method of determining one sensor as a cluster head and multiple sensors as cluster members in each cluster includes: calculating the distance between the position of each sensor in the cluster and the center position of the cluster, and calculating the residual energy and average residual energy of the sensors in the cluster in a preset manner; from all sensors whose residual energy is greater than the average residual energy, selecting a sensor with the smallest distance to the corresponding cluster center position as the cluster head, and the remaining sensors in the cluster are cluster members.

[0022] In some embodiments of the present invention, the remaining energy of the sensor is calculated as follows:

[0023] ,

[0024] in, Indicates the remaining energy of the sensor. represents the initial energy of the sensor in the previous round, Indicates the previous round The energy consumed by the sensor, Indicates the corresponding round, when the sensor was the cluster head in the previous round: , Indicates the number of packets that need to be sent. It indicates the proportion of aggregation when cluster heads gather data. , Indicates the amount of data transmitted per packet. Indicates the energy required to transmit a unit of data. represents the energy parameter of the free space model, represents the energy parameter of the multipath fading model, Indicates the distance between the sensor sending data and the sensor receiving data, Indicates the preset distance threshold, represents the energy consumption per unit data aggregation, Represents the energy consumption of the sensor receiving data when the sensor was a cluster member in the previous round: .

[0025] According to a second aspect of the present invention, a wireless ad hoc network data aggregation system is provided, the system comprising a plurality of sensors and aggregation nodes, wherein:

[0026] The aggregation node is configured to: determine the level to which each sensor belongs, the sensor as the cluster head and the sensor as the cluster member in each cluster of each level by the method described in the first aspect of the present invention during each round of aggregation, and instruct the sensor as the cluster member in each cluster of each level to transmit the data collected by it to the cluster head of the cluster to which it belongs, and instruct the sensor as the cluster head in each cluster of each level to aggregate the data collected by all sensors in each cluster and transmit the aggregated data to the aggregation node;

[0027] The sensor is configured to: according to the instruction transmitted from the sink node indicating that it is a cluster member, transmit the data collected by each sensor to the cluster head of the cluster to which it belongs; or according to the instruction transmitted from the sink node indicating that it is a cluster head, receive the data transmitted from the cluster members of the cluster to obtain the aggregated data of the cluster, wherein the aggregated data of the cluster not at the top layer is transmitted to the cluster head closest to the upper layer, and the aggregated data of the cluster at the top layer is transmitted to the sink node.

[0028] Compared with the prior art, the advantages of the present invention are:

[0029] In each round of convergence, the present invention, on the one hand, hierarchizes the network, divides multiple clusters in each layer, and determines the sensors as cluster heads and cluster members in each cluster, so that cluster heads are set at different layers away from the convergence node, ensuring that each cluster head is evenly distributed, so that each cluster head can share the data transmission task evenly, and improve the life of each cluster head. On the other hand, the data of the cluster members of each cluster are converged to the cluster head to obtain the converged data of each cluster, and the converged data of the cluster is transmitted to the nearest cluster head or convergence node in the upper layer, so that the transmission distance of the cluster head and the cluster member is controlled within the divided inter-layer distance, and when the converged data is transmitted, the nearest cluster head is used as the next hop for transmission, which greatly reduces the average transmission distance of the sensor, reduces the overall transmission energy consumption of the network, improves the overall energy utilization of the network, and simplifies the calculation of the transmission path. In addition, the present invention takes into account the mobility of the sensor, and performs multiple rounds of convergence, that is, re-stratification, clustering and cluster head selection at intervals to adapt to the position change of the sensor, so that the selection of the cluster head is in a changing state, and the data transmission tasks of each sensor are further balanced. That is, the method of the present invention is not only easy to deploy and implement, but also can effectively ensure the sustainability and long-term stability of the network. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] The embodiments of the present invention are further described below with reference to the accompanying drawings, in which:

[0031] Figure 1A schematic diagram of the principle of each round of convergence process according to an embodiment of the present invention;

[0032] Figure 2 A schematic diagram of a transmission structure principle using a first radio model according to an embodiment of the present invention;

[0033] Figure 3 A schematic diagram of a wireless ad hoc network data aggregation architecture principle according to an embodiment of the present invention;

[0034] Figure 4 is a schematic diagram of the overall process of performing each round of aggregation according to an embodiment of the present invention;

[0035] Figure 5 A schematic diagram of the structural principle of each sensor according to an embodiment of the present invention;

[0036] Figure 6 It is a schematic diagram of the comparison result of the overall network energy consumption between the method of the present invention according to an embodiment of the present invention and the existing method. DETAILED DESCRIPTION

[0037] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below through specific embodiments in conjunction with the accompanying drawings. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0038] As mentioned in the background technology section, as the network coverage area increases, the existing single-layer architecture method has significant problems, including: on the one hand, the selection of cluster heads is usually based on load balancing and transmission energy consumption considerations, which leads to uneven distribution of cluster heads in a large-scale network, so that some cluster heads may be burdened with too many data transmission tasks, thus consuming energy quickly and quickly shortening the life of nodes. On the other hand, due to the single-hop transmission from the cluster head to the aggregation node, the nodes far away from the aggregation node are prone to overload, which not only increases energy consumption, but also may cause these nodes to run out of energy in a short period of time and cannot continue to operate effectively. Both of the above problems limit the sustainability and long-term stability of the network.

[0039] Based on the above problems, the inventor proposes a method for data aggregation in a wireless ad hoc network. First, the method of the present invention is applied to a wireless sensor network (WSN for short), and the wireless sensor network includes multiple sensors, and the positions of at least some sensors can be changed. The method includes: performing multiple rounds of aggregation on the data collected by each sensor based on a predetermined time interval. Secondly, in each round of aggregation process, according to the distance between all sensors in the wireless sensor network and the aggregation node and the distance range values ​​between each of the preset multiple levels and the aggregation node, the level to which each sensor belongs is determined, and the network is hierarchical, and the level closest to the aggregation node is the top level, and multiple clusters are divided in each level and the sensors as cluster heads and sensors as cluster members in each cluster are determined, so that cluster heads are set at different levels away from the aggregation node, ensuring that each cluster head is evenly distributed, so that each cluster head can evenly share the data transmission task, and improve the life of each cluster head. Finally, since the existing single-hop transmission mode from cluster head to aggregation node easily leads to the problem of overload in cluster heads far from the aggregation node, the method of the present invention aggregates the data of cluster members of each cluster to the cluster head, obtains the aggregated data of each cluster, and transmits the aggregated data of the cluster to the nearest cluster head or aggregation node in the upper layer, so that the transmission distance of the cluster head and cluster members of the present invention is controlled within the divided inter-layer distance, and the nearest cluster head is used as the next hop for transmission when the aggregated data is transmitted, which greatly reduces the average transmission distance of the sensor, avoids the sensor from running out of energy in a short time and being unable to continue to operate effectively, reduces the overall transmission energy consumption of the network, improves the overall energy utilization of the network, and simplifies the transmission path. The present invention takes into account the mobility of the sensor, re-layers, clusters and cluster head selections at regular intervals to adapt to the location changes of the sensor, that is, the selection of the cluster head is in a changing state, further balancing the data transmission tasks of each sensor, and the method of the present invention is not only easy to deploy and implement, but also can effectively ensure the sustainability and long-term stability of the network.

[0040] According to one embodiment of the present invention, all sensors in a wireless sensor network can perform data collection, monitoring and control tasks, and can select sensors as common nodes (referred to as cluster members in the present invention) and sensors as cluster heads from all sensors in the form of self-organizing networks. Among them, the common nodes are responsible for collecting data and transmitting it to the cluster heads to which they belong, while the cluster heads are responsible for aggregating and transmitting the data of the cluster members to obtain aggregated data. The aggregated data is transmitted to the nearest aggregation node in a self-organized manner, and the aggregation node is generally an industrial gateway or a base station. Each sensor aggregates and transmits the collected data to the nearest aggregation node in a self-organized manner based on the transmission method of the present invention through 5G or short-range communication technology. Therefore, the method of the present invention is suitable for data acquisition systems, and can be used for data acquisition in application scenarios such as indoor production workshops, outdoor mines, and offshore oil fields. In such data acquisition systems, sensors are divided into fixed nodes and movable nodes. Taking data acquisition in indoor production workshop scenarios as an example, the core responsibility of fixed nodes is to monitor the operating status of equipment, workshop environment information, etc. in real time; movable nodes are installed on unmanned vehicles or handheld tools, and their monitoring positions can be changed as the equipment and staff move.

[0041] According to one embodiment of the present invention, see Figure 1 , which is a schematic diagram of the principle of each round of convergence process of the method of the present invention, and each round of convergence process includes steps S1, S2, S3 and S4. In order to better understand the present invention, each step is described in detail below in conjunction with specific embodiments.

[0042] Step S1: Determine the level to which each sensor belongs according to the distances between all sensors in the wireless sensor network and the sink node and the preset distance ranges between multiple levels and the sink node, wherein the level closest to the sink node is the top level.

[0043] According to an embodiment of the present invention, taking the application scenario of a production workshop as an example, since fixed nodes account for the majority in the workshop, and in a specific monitoring time period, the nodes installed on the maintenance equipment or handheld devices also show the characteristics of static distribution, the present invention mainly considers a static WSN network when performing each round of convergence. That is, the wireless sensor network has the following characteristics during each round of convergence: 1) Sensors are randomly distributed in the monitoring area, and there is at most one sensor in the same position. 2) Sensors can identify their location information by themselves, and identify the locations of surrounding nodes through the neighbor discovery mechanism. 3) Sensors all perform monitoring tasks, collect monitoring data in their areas, and transmit these data to the convergence node. 4) In the time slot of the current round, sensors are all regarded as static nodes. 5) All sensors are homogeneous, with consistent configurations and the same initial energy level. 6) The location of the convergence node is fixed, and it has unlimited energy supply, and can accurately know the specific location information of all sensors in the network. 7) The sensor has the function of data fusion. When acting as a cluster head, it can reduce the total amount of data transmission by fusing data packets of other nodes.

[0044] Therefore, based on the above analysis, the wireless sensor network has characteristics in each round of convergence. According to one embodiment of the present invention, in step S1, the method of determining the level to which each sensor belongs includes: obtaining a preset distance range value between each level and the convergence node in the preset multiple levels, wherein the distance range values ​​of the multiple levels are different and the distance range values ​​do not overlap; determining the distance between each sensor and the convergence node according to the current position of each sensor in the obtained wireless sensor network and the position of the convergence node; and dividing the sensor into the level corresponding to the preset distance range value according to the preset distance range value of the distance between each sensor and the convergence node. After the convergence node in each round obtains the positions of all sensors at the beginning, it is deemed that the positions of each sensor remain unchanged during the convergence time of this round.

[0045] In an illustrative manner, the preset distance range values ​​of the preset multiple levels from the sink node include: at the beginning of the data transmission phase, all sensors transmit their location information in a broadcast form, and after the sink node collects the location information of all sensors, a fixed threshold is set. , according to the distance from the sensor to the sink node, according to the fixed threshold The network is divided into multiple layers. The layer closest to the sink node is the top layer, and the distance range values ​​corresponding to the multiple layers arranged from the top layer to the bottom layer are: [0, )、[ , 2 )、[2 , 3 )..., and so on. When the distance between the sensor and the sink node is [0, ), the sensor is divided into the top layer, that is, the distance from each sensor to the sink node of all sensors contained in each layer is within the distance range value corresponding to the layer. Subsequently, multi-hop layer-by-layer transmission is performed by setting one hop per layer, avoiding complex transmission path calculation.

[0046] Step S2: Divide all sensors at each level into multiple clusters according to a preset division rule, and determine a sensor as a cluster head and sensors as cluster members in each cluster.

[0047] According to one embodiment of the present invention, in step S2, the preset division rule includes steps a1 and a2:

[0048] Step a1: Utilize the Soft K-Means algorithm to obtain the multiple cluster center positions corresponding to each layer and the membership probability values ​​of each sensor in the layer belonging to each cluster according to the positions of each sensor in each layer.

[0049] According to an embodiment of the present invention, a method for obtaining a plurality of cluster center positions corresponding to each layer and a membership probability value of each sensor in the layer belonging to each cluster includes: dividing each layer multiple times, and each division includes the following steps a11, a12 and a13:

[0050] Step a11: obtaining multiple cluster center positions, wherein the first time a plurality of randomly selected cluster center positions are used, and each time after the first time a plurality of cluster center positions calculated last time are used.

[0051] According to one embodiment of the present invention, the set of multiple cluster center positions in each level is recorded as , Indicates The cluster center location, represents the number of clusters in each level, Indicates the total number of clusters in each level.

[0052] According to an embodiment of the present invention, a cluster corresponds to a cluster center position, and the closer the level is to the aggregation node, the more cluster center positions are obtained, and the closer the level is to the aggregation node, the more clusters are divided. That is, the number of clusters divided from the lowest level to the highest level increases successively. Schematically, if the four levels divided from the lowest level to the highest level are divided into 6 clusters, 12 clusters, 16 clusters and 20 clusters respectively, the task flow of cluster heads in each cluster is balanced. The technical solution of this embodiment can at least achieve the following beneficial technical effects: different numbers of clusters can be achieved by setting different numbers of cluster centers, that is, the present invention takes into account that the levels close to the aggregation node will bear more tasks of forwarding aggregation data, so the number of clusters is increased in these layers to reduce the load of a single cluster, thereby optimizing the energy consumption distribution of the network.

[0053] Step a12: Determine the membership probability value of each sensor belonging to each cluster according to the positions of the center of multiple clusters and the positions of each sensor.

[0054] According to one embodiment of the present invention, the manner of determining the membership probability value of each sensor belonging to each cluster is as follows:

[0055] ,

[0056] in, Indicates The sensor belongs to The membership probability value of a cluster, is the weight parameter of the membership probability value of each sensor, represents the number of clusters in each level, represents the total number of clusters in each level, Indicates The location of the sensor, Indicates the sensor number, represents the total number of sensors in each layer, Indicates The cluster center location, Indicates Sensors and cluster center locations The distance

[0057] Step a13: According to the multiple cluster center positions, the position of each sensor and the membership probability value, the multiple cluster center positions for the next time are calculated with minimizing the distance within each cluster as the optimization objective function until the soft K-means clustering algorithm converges, and the final multiple cluster center positions and the membership probability value of each sensor in this layer belonging to each cluster are obtained.

[0058] According to one embodiment of the present invention, the optimization objective function is as follows:

[0059] ,

[0060] in, represents minimizing the distance within each cluster, represents the distance within each cluster, , represents the location of all sensors in each layer, Indicates the position of the first sensor, Indicates the position of the second sensor, Indicates The location of the sensor, , Represents the set of membership probability values ​​of each sensor to each cluster. The present invention minimizes the distance within the cluster through this function, which not only realizes faster data convergence within the cluster, but also reduces the transmission energy consumption within the cluster.

[0061] According to one embodiment of the present invention, the method for determining the convergence of the soft K-means clustering algorithm includes: calculating the membership probability change value according to the difference between the membership probability value of each sensor belonging to each cluster obtained this time and the membership probability value of each sensor belonging to each cluster obtained last time; obtaining a preset probability threshold, and determining that the soft K-means clustering algorithm has converged when the membership probability change value is less than the preset probability threshold. In addition, the method for determining the convergence of the soft K-means clustering algorithm also includes: calculating the position distance change value according to the difference between the multiple cluster center positions obtained this time and the multiple cluster center positions obtained last time; obtaining a preset distance threshold, and determining that the soft K-means clustering algorithm has converged when the position distance change value is less than the preset distance threshold.

[0062] According to one embodiment of the present invention, the set of membership probability values ​​of each sensor to each cluster obtained by any division is recorded as , the set of multiple cluster center positions in each level obtained by any division is recorded as If the set of membership probability values ​​of each sensor to each cluster obtained by this division is Compared with the previous division The change between Below the preset probability threshold, or the set of multiple cluster center positions in each level obtained by the current division Compared with the last time The change between If the distance is lower than the preset threshold, the soft K-means clustering algorithm ends and the final membership probability value of each sensor to each cluster and the center positions of multiple clusters are obtained. or change value If the value is greater than or equal to the corresponding preset threshold, the above steps a11, a12 and a13 are repeated using the soft K-means clustering algorithm to recalculate the new set of membership probability values. and the set of new cluster center locations Among them, the change value The calculation method includes: subtracting the set of membership probability values ​​from the set of previous membership probability values ​​to obtain a difference set, taking the absolute value of each difference in the difference set and adding them together to obtain the membership probability change value Similarly, the change value The calculation method can also use the change value If it is the first division, the change value is not calculated, and the second division is performed directly, that is, the new set of membership probability values ​​is recalculated. and the set of new cluster center locations .

[0063] Step a2: Determine the cluster to which each sensor belongs according to the membership probability value of each sensor belonging to each cluster, so as to divide all sensors at the level into multiple clusters.

[0064] According to one embodiment of the present invention, the method of dividing into multiple clusters includes: according to the membership probability value of each sensor belonging to each cluster, dividing the sensor into the cluster corresponding to the maximum membership probability value to obtain multiple clusters initially divided at the level; judging whether the number of sensors contained in each of the multiple clusters initially divided is balanced, if not balanced, balancing the number of sensors in the multiple clusters initially divided to obtain multiple clusters finally divided, otherwise, using the multiple clusters initially divided as the multiple clusters finally divided.

[0065] Schematically, if five clusters need to be divided for this level, and the membership probability values ​​of the sensors belonging to the center positions of the five clusters are 20%, 50%, 5%, 10% and 15% respectively. Then the cluster corresponding to the highest membership probability value of 50% is selected as the cluster to which it belongs. In this way, multiple clusters initially divided for this level are obtained. At this time, it can be further determined whether the number of sensors contained in each of the multiple clusters initially divided is balanced. Schematically, the judgment method can be calculated by calculating the difference in the number of sensors between any two clusters in all clusters. If the difference is greater than the preset difference threshold, it means that the number of sensors in the initially divided clusters is unbalanced, and balanced processing is required. The balanced processing method includes: re-dividing some sensors in the cluster with the largest number of sensors into clusters adjacent to the cluster. Among them, a circle is drawn with the center position of the cluster and with a preset radius threshold. The clusters within the circle and belonging to the same level as the cluster are clusters adjacent to it. Indicatively, if there are three clusters A, B and C after clustering at the same level, where the number of sensors in A, B and C are 20, 5 and 7 respectively, then the number of nodes in A is too large, so the number of nodes in cluster A should be reduced. If B and C are adjacent to A, and if the probability of x belonging to B and C is 30% and 20% respectively, then x should be divided into B. If balanced, the multiple clusters initially divided are used as the multiple clusters finally divided.

[0066] According to an embodiment of the present invention, in step S2, the method of determining a sensor as a cluster head and a plurality of sensors as cluster members in each cluster includes steps b1 and b2:

[0067] Step b1: Calculate the distance between the position of each sensor in the cluster and the center position of the cluster, and calculate the residual energy and average residual energy of the sensors in the cluster in a preset manner.

[0068] According to an embodiment of the present invention, the distance between the position of each sensor in the cluster and the center position of the cluster can be calculated using the Euclidean distance formula, which is not limited in the present invention, and the Manhattan distance calculation principle can also be used. The residual energy of all sensors in the cluster is added up and then divided by the total number of sensors in the cluster to obtain the average residual energy of the cluster. Among them, the calculation method of the remaining energy of the sensor is as follows:

[0069] ,

[0070] in, Indicates the remaining energy of the sensor. represents the initial energy of the sensor in the previous round, Indicates the previous round The energy consumed by the sensor, Indicates the corresponding round, when the sensor was the cluster head in the previous round: , Indicates the number of packets that need to be sent. It indicates the proportion of aggregation when cluster heads gather data. , Indicates the amount of data transmitted per packet. Indicates the energy required to transmit a unit of data. represents the energy parameter of the free space model, represents the energy parameter of the multipath fading model, Indicates the distance between the sensor sending data and the sensor receiving data, Indicates the preset distance threshold, represents the energy consumption per unit data aggregation, Represents the energy consumption of the sensor receiving data when the sensor was a cluster member in the previous round: Among them, the proportion of polymerization It is usually set to 0.1~0.5. In WSN, if =0.2, it means that every 5 data are aggregated into 1, which can reduce the amount of transmitted data and improve efficiency while retaining enough information for analysis.

[0071] Since the remaining energy of the sensor is calculated based on the energy it consumes, the calculation principle of the sensor's energy consumption is explained below:

[0072] 1) Analysis of the energy consumption calculation principles for sending and receiving data

[0073] ①Introduce the energy consumption model in the data collection process of wireless sensor networks. The transmission model of the wireless channel is based on the straight-line communication distance between the transmitter and the receiver, which can be divided into the free space propagation model and the multipath fading channel model. Less than threshold When using the Freeis free space propagation model (which belongs to the free space propagation model), the signal power at the receiving end is defined as:

[0074] ,

[0075] in, is the transmit power, is the transmit antenna gain, is the receiving antenna gain, is the RF signal carrier wavelength, is the system loss factor. Greater than threshold When the signal is reflected from the ground, due to the multipath effect caused by the ground reflection signal, the signal power at the receiving end needs to be calculated using the two-way ground reflection transmission model (belonging to the multipath fading channel model), which is defined as:

[0076] ,

[0077] in, is the distance between the transmitting antenna and the ground, is the distance between the receiving antenna and the ground.

[0078] ②Based on the above two definitions, quantitative analysis of energy consumption at the receiving and transmitting ends is performed. Figure 2 , which is a schematic diagram of the transmission structure principle using the first radio model. In this model, the energy consumption of the transmitter is divided into two parts: the energy consumption of the transmission circuit and the energy consumption of the RF power amplifier (i.e., the amplifier circuit). It is assumed that the transmitter can adjust the power amplifier according to the communication distance to ensure that the signal power received at the receiving end meets the decoding requirements. Therefore, the energy consumption of the transmitter sending a data packet is It can be expressed as:

[0079] .

[0080] For the receiving end, the model only considers the energy consumption of the receiving circuit. Therefore, the energy consumption of receiving data packets at the receiving end is .

[0081] 2) Analysis of the energy consumption calculation principle of cluster head aggregation data

[0082] Since the data detected by sensors, such as pressure, temperature, and humidity, usually have a high correlation, the cluster head can perform data compression after collecting the data collected by the cluster members of its cluster to remove data redundancy and reduce the amount of data transmitted. It can be expressed as follows:

[0083] ,

[0084] in, Indicates the number of packets that need to be sent, which is determined by the number of nodes in the cluster and the number of forwarding requirements of other cluster heads. Represents the energy consumption per unit of data aggregation.

[0085] Based on the above analysis of energy consumption calculation principle, the energy consumption of cluster head in the rth round is defined as , the energy consumption of cluster members in the rth round is defined as .

[0086] Step b2: From all sensors whose residual energy is greater than the average residual energy, select a sensor with the smallest distance to the corresponding cluster center as the cluster head, and the remaining sensors in the cluster are cluster members.

[0087] According to an embodiment of the present invention, all sensors in the cluster are sorted in ascending order according to the distance between each sensor and the center of the cluster, to obtain a sensor sequence, and the sensors in the sequence are traversed. When the residual energy of a certain sensor is higher than the average residual energy in the cluster, the sensor is selected as the cluster head. The method of selecting the cluster head of the present invention can reduce the overall energy consumption and ensure the load balance in the cluster.

[0088] In schematic form, assume that there are 5 sensors in the cluster, namely A, B, C, D and E, and their respective residual energies are 5, 10, 5, 15 and 10. F is the center of the cluster. The distances from F from near to far are A, D, C, E, and B. The average residual energy of the cluster is (5+10+5+15+10) / 5=9. After traversing A, D, C, E, and B, it is found that D is the first node whose residual energy is greater than the average residual energy of the cluster, so D is selected as the cluster head.

[0089] Step S3: Aggregate the data collected by all sensors in each cluster to the cluster head of the cluster to obtain aggregated data of each cluster.

[0090] According to an embodiment of the present invention, the data collected by the cluster members of the cluster are transmitted to the cluster head of the cluster by single-hop transmission. The cluster head aggregates the data received from all cluster members and the data collected by itself to obtain aggregated data of the cluster.

[0091] Step S4: transmitting the aggregated data to the sink node layer by layer according to the cluster heads, including: transmitting the aggregated data of the non-uppermost cluster to the nearest cluster head of the upper layer, and transmitting the aggregated data of the uppermost cluster to the sink node.

[0092] According to one embodiment of the present invention, see Figure 3 , which is a schematic diagram of the data aggregation architecture principle of wireless ad hoc networks. The figure is divided into two levels, namely Layer 1 (i.e. the top layer) and Layer 2. Both Layer 1 and Layer 2 show two clusters, the red sensors are cluster heads, and the black sensors are cluster members of the corresponding clusters. When performing data aggregation, first, the data stream within the cluster is transmitted in the manner of the above-mentioned embodiment, that is, the black arrow indicates that the cluster members within the cluster use a single-hop transmission method to transmit the collected data to the cluster head of the cluster, and obtain the aggregated data of each cluster. Then, according to the cluster head, the aggregated data is transmitted to the aggregation node layer by layer. As shown in Figure 3As shown, the cluster head of the second layer transmits the converged data to the nearest cluster head of the first layer, and the red arrow indicates the direction of the cluster head data flow. The cluster head of the first layer directly transmits the converged data to the sink node. The technical solution of this embodiment can at least achieve the following beneficial technical effects: the data of the cluster members of each cluster are directly converged to the cluster head through a single-hop transmission method, and the converged data of the cluster is transmitted to the nearest cluster head or sink node in the upper layer, so that the transmission distance of the cluster head and the cluster members are controlled within the divided inter-layer distance, which greatly reduces the average transmission distance of the sensor, reduces the overall transmission energy consumption of the network, and improves the overall energy utilization of the network. And the calculation of the transmission path is simplified through layer-by-layer transmission.

[0093] According to an embodiment of the present invention, the overall process of each round of aggregation is now described based on the above embodiment, see Figure 4 , which is a schematic diagram of the overall process of each round of convergence in the present invention, including the following steps:

[0094] 1. Divide the network into multiple layers according to the distance between the sensor and the sink node. Perform the following process on each layer:

[0095] 2. Clustering is performed at the current level using the soft K-means clustering algorithm to obtain the sensor's membership probability to each cluster and the center position of each cluster;

[0096] 3. Select the cluster head according to the residual energy of each sensor in the cluster and the distance of each sensor from the cluster center, including sorting the sensors in ascending order according to the distance of each sensor from the cluster center; and calculating the average residual energy of the sensors in the cluster; according to the sorting results of the sensors, traverse the sensors in turn, and if the residual energy of the sensor is greater than the average residual energy, select it as the cluster head;

[0097] 4. Transmit the aggregated data of the current level to the aggregation node, including:

[0098] 4.1. Determine whether the current layer is the top layer. If not, continue to execute the process of 4.2. Otherwise, set the sink node as the target node for direct transmission, and the transmission ends.

[0099] 4.2. Select the cluster head closest to the previous level of the current level as the next hop, and return to 4.1 to continue execution.

[0100] According to one embodiment of the present invention, for example, in an express warehousing system, there are a large number of AGVs transporting express delivery, and it is necessary to collect the location information of all AGVs to plan a reasonable transportation path for them. At this time, a large amount of location information needs to be transmitted to the data center for processing. Assume that there is only one data center, and this data center is a sink node. According to the distance between the AGV and the sink node, the AGV is divided into multiple levels, and the sink node is divided into the top layer; starting from the top layer, Soft K-Means is used for clustering in each level; when transmitting data, if the current level is the top layer, the sink node is set as the target node, otherwise, the cluster head with the closest distance in the previous level is selected as the next hop; when all levels have completed the cluster head selection, the path transmission of the sink data of each cluster in each level can be determined, and data aggregation can be started according to the path. At regular intervals, the above steps are repeated starting from the layer to adapt the movable sensor.

[0101] According to one embodiment of the present invention, a wireless ad hoc network data aggregation system is provided, which includes multiple sensors and an aggregation node, wherein the aggregation node is configured to: during each round of aggregation, determine the level to which each sensor belongs, the sensor serving as a cluster head and the sensor serving as a cluster member in each cluster of each level by the method of the above embodiment, and instruct the sensors serving as cluster members in each cluster of each level to transmit their collected data to the cluster head of the cluster to which they belong, and instruct the sensors serving as cluster heads in each cluster of each level to aggregate the data collected by all sensors in each cluster and transmit the aggregated data to the aggregation node; the sensor is configured to: receive the data transmitted by the cluster members of its cluster according to the instruction transmitted by the aggregation node indicating that it serves as the cluster head, and obtain the aggregated data of its cluster, wherein the aggregated data of the clusters not in the uppermost layer is transmitted to the cluster head closest to the uppermost layer, and the aggregated data of the cluster in the uppermost layer is transmitted to the aggregation node.

[0102] According to one embodiment of the present invention, see Figure 5 , which is a schematic diagram of the structural principle of each sensor, each sensor includes:

[0103] 1) Acquisition module: used to collect environmental data such as temperature and / or humidity. This module includes various sensing units, such as temperature, humidity or image sensing units, and the corresponding sensing unit can be selected according to monitoring requirements. The collected signal data is amplified, converted and pre-processed by the AD / DA conversion unit of this module;

[0104] 2) Processing module: It includes a processor and a memory. The processor is used to process the data collected by the acquisition module, perform data transmission tasks, manage clusters, determine the data transmission path, and control the power supply module. The memory is used to store the data collected by the acquisition module;

[0105] 3) Communication module: It includes a wireless transceiver and an AD / DA conversion unit. The wireless transceiver is used to transmit data and instructions between sensors, and the AD / DA conversion unit is used to convert received data or amplify data to be sent. This module enables the sensor to communicate wirelessly with other sensors or base stations in the network without physical connection, thus forming a collaborative network system;

[0106] 4) Power supply module: used to supply the necessary energy to the sensor, usually through a micro battery, or by integrating renewable energy modules such as solar panels or wireless charging modules to extend the working life of the sensor.

[0107] To verify the beneficial effects of the present invention, the inventors conducted the following experiments:

[0108] Parameter setting: Set the initial energy of each sensor in the wireless sensor network to 0.5J, set the data packet length and control packet length to 4000 bits and 100 bits respectively, and the energy required to transmit unit data , energy parameters of the free space model and the energy parameters of the multipath fading model Set to 50nJ / bit, 10pJ / bits / and 0.0013pJ / bits / , energy consumption per unit data aggregation The setting is 5nJ / bit, and 200 rounds of data acquisition are performed (i.e., 200 rounds of convergence are performed).

[0109] Based on the above parameter settings, the overall network energy consumption of the method of the present invention (called LBC) is compared with that of the existing methods under different data collection rounds. The existing methods include the R-Leach algorithm: a clustering algorithm based on residual energy and the OK-Means algorithm: a clustering algorithm based on optimal K-means. Figure 6 , which is a schematic diagram of the comparison of the overall network energy consumption of the method of the present invention and the existing method. The horizontal axis of the figure is the data collection round (i.e., the convergence round), and the vertical axis is the overall energy consumption, with the unit being J. According to the figure, as the round increases, the energy consumption of the method of the present invention increases the slowest. After 200 rounds of data collection, the energy consumption of the method of the present invention is reduced by 14.4% and 35.4% respectively compared with the OK-Means algorithm and the R-Leach algorithm, which illustrates the effectiveness of the method of the present invention in reducing energy consumption.

[0110] The existing methods are as follows:

[0111] R-Leach algorithm: Behera TM, Mohapatra SK, Samal UC, et al. ResidualEnergy-Based Cluster-Head Selection in WSNs for IoT Application[J]. IEEEInternet of Things Journal, 2019, 6(3): 5132-5139.

[0112] OK-Means algorithm: El Khediri S, Fakhet W, Moulahi T, et al. Improved nodelocalization using K-means clustering for Wireless Sensor Networks[J]. Computer Science Review, 2020, 37: 100284.

[0113] It should be noted that although the above describes the various steps in a specific order, it does not mean that the various steps must be executed in the above specific order. In fact, some of these steps can be executed concurrently or even in a different order as long as the required functions can be achieved.

[0114] The present invention may be a system, a method and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for causing a processor to implement various aspects of the present invention.

[0115] A computer-readable storage medium may be a tangible device that holds and stores instructions used by an instruction execution device. Computer-readable storage media may include, for example, but are not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination thereof. More specific examples (non-exhaustive list) of computer-readable storage media include: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a static random access memory (SRAM), a portable compact disk read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanical encoding device, such as a punch card or a protruding structure in a groove on which instructions are stored, and any suitable combination thereof.

[0116] The embodiments of the present invention have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or technical improvements in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. A wireless ad hoc network data aggregation method, applied to wireless sensor networks, wherein: The wireless sensor network includes a plurality of sensors, and the positions of at least some of the sensors are changeable. The method includes: performing multiple rounds of aggregation on the data collected by each sensor based on a predetermined time interval, wherein each round of aggregation process is performed in the following manner: S1, according to the distances between all sensors in the wireless sensor network and the sink node and the distance ranges between each of the preset multiple levels and the sink node, determine the level to which each sensor belongs, wherein the level closest to the sink node is the top level; S2, dividing all sensors at each level into multiple clusters according to a preset division rule, and determining a sensor as a cluster head and a sensor as a cluster member in each cluster; S3, aggregating the data collected by all sensors in each cluster to the cluster head of the cluster to obtain the aggregated data of each cluster; S4. Transmitting the aggregated data to the sink node layer by layer according to the cluster heads, including: transmitting the aggregated data of the non-uppermost cluster to the nearest cluster head of the upper layer, and transmitting the aggregated data of the uppermost cluster to the sink node.

2. The method according to claim 1, characterized in that In S2, the preset division rule includes: The soft K-means clustering algorithm is used to cluster the positions of each sensor in each layer, and multiple cluster center positions corresponding to each layer and the membership probability values ​​of each sensor in the layer belonging to each cluster are obtained; According to the membership probability value of each sensor belonging to each cluster, the cluster to which each sensor belongs is determined, so as to divide all sensors of the layer into multiple clusters; The closer the level is to the sink node, the greater the number of clusters.

3. The method according to claim 2, characterized in that The method of obtaining the multiple cluster center positions corresponding to each layer and the membership probability value of each sensor in the layer belonging to each cluster includes: Each level is divided multiple times, and each division is carried out in the following way: Obtain multiple cluster center positions, wherein the first time a plurality of randomly selected cluster center positions are used, and each time after the first time a plurality of cluster center positions calculated last time are used; According to the positions of the centers of the multiple clusters and the positions of the various sensors, the membership probability value of each sensor belonging to each cluster is determined; According to the positions of the center of multiple clusters, the position of each sensor and the membership probability value, the center positions of multiple clusters for the next time are calculated with minimization of the distance within each cluster as the optimization objective function until the soft K-means clustering algorithm converges, and the final center positions of multiple clusters and the membership probability value of each sensor in this layer belonging to each cluster are obtained.

4. The method according to claim 3, characterized in that: The optimization objective function is as follows: ， in, represents minimizing the distance within each cluster, represents the distance within each cluster, , represents the location of all sensors in each layer, Indicates the position of the first sensor, Indicates the position of the second sensor, Indicates The location of the sensor, , Indicates The sensor belongs to The membership probability value of a cluster, , is the weight parameter of the membership probability value of each sensor, represents the number of clusters in each level, represents the total number of clusters in each level, Indicates The location of the sensor, Indicates the sensor number, represents the total number of sensors in each layer, Indicates The cluster center location, Indicates The distance between each sensor and the cluster center, Represents the set of all cluster center locations in each level.

5. The method according to claim 3, characterized in that: The method of determining the convergence of the soft K-means clustering algorithm includes: Calculate the membership probability change value according to the difference between the membership probability value of each sensor belonging to each cluster obtained this time and the membership probability value of each sensor belonging to each cluster obtained last time; A preset probability threshold is obtained, and when the membership probability change value is less than the preset probability threshold, it is determined that the soft K-means clustering algorithm has converged.

6. The method according to claim 3, characterized in that The method of determining the convergence of the soft K-means clustering algorithm includes: Calculate the position distance change value according to the difference between the multiple cluster center positions obtained this time and the multiple cluster center positions obtained last time; A preset distance threshold is obtained, and when the position distance change value is less than the preset distance threshold, it is determined that the soft K-means clustering algorithm converges.

7. The method according to claim 2, characterized in that The method of dividing into multiple clusters includes: According to the membership probability value of each sensor belonging to each cluster, the sensor is divided into the cluster corresponding to the maximum membership probability value to obtain multiple clusters initially divided for the level; It is determined whether the number of sensors contained in each of the multiple clusters initially divided is balanced. If not, the number of sensors in the multiple clusters initially divided is balanced to obtain the multiple clusters finally divided. Otherwise, the multiple clusters initially divided are used as the multiple clusters finally divided.

8. The method according to claim 2 or 7, characterized in that: In S2, the method of determining a sensor as a cluster head and a plurality of sensors as cluster members in each cluster includes: Calculate the distance between the position of each sensor in the cluster and the center position of the cluster, and calculate the remaining energy and average remaining energy of the sensors in the cluster in a preset manner; From all sensors whose residual energy is greater than the average residual energy, a sensor with the smallest distance to the corresponding cluster center is selected as the cluster head, and the remaining sensors in the cluster are cluster members.

9. The method according to claim 8, characterized in that The calculation method of the remaining energy of the sensor is as follows: ， in, Indicates the remaining energy of the sensor. represents the initial energy of the sensor in the previous round, Indicates the previous round The energy consumed by the sensor, Indicates the corresponding round, when the sensor was the cluster head in the previous round: , Indicates the number of packets that need to be sent. It indicates the proportion of aggregation when cluster heads gather data. , Indicates the amount of data transmitted per packet. Indicates the energy required to transmit a unit of data. represents the energy parameter of the free space model, represents the energy parameter of the multipath fading model, Indicates the distance between the sensor sending data and the sensor receiving data, Indicates the preset distance threshold, represents the energy consumption per unit data aggregation, Represents the energy consumption of the sensor receiving data when the sensor was a cluster member in the previous round: .

10. A wireless ad hoc network data aggregation system, characterized in that: The system includes multiple sensors and aggregation nodes, among which, The sink node is configured as: When performing each round of aggregation, the layer to which each sensor belongs, the sensor serving as the cluster head and the sensor serving as the cluster member in each cluster of each layer are determined by the method described in any one of claims 1 to 9, and the sensors serving as the cluster members in each cluster of each layer are instructed to transmit the data collected by them to the cluster head of the cluster to which they belong, and the sensors serving as the cluster head in each cluster of each layer are instructed to aggregate the data collected by all sensors in each cluster and transmit the aggregated data to the aggregation node; The sensor is configured as: According to the instruction transmitted by the sink node indicating that it is a cluster member, the collected data is transmitted to the cluster head of the cluster to which it belongs; or According to the instruction transmitted by the aggregation node indicating that it serves as the cluster head, it receives the data transmitted by the cluster members of its cluster and obtains the aggregated data of its cluster, wherein the aggregated data of the non-top cluster is transmitted to the nearest cluster head of the upper layer, and the aggregated data of the top cluster is transmitted to the aggregation node.

Images