A method for data aggregation in underwater acoustic sensor networks

Abstract

The application provides a kind of underwater acoustic sensor network data convergence method, the spatial function model is established using the spatial variation characteristics of feature level data in UWSNs to characterize the sensor nodes in the network, a spatial function approximation algorithm with maximum approximation error priority transmission is proposed, the data of the node with maximum approximation error in UWSNs is transmitted in priority, the wireless broadcast characteristics of underwater acoustic signals are fully utilized, a sequential optimal subset selection method and optimal local error criterion are proposed, and through extreme value convergence and implementation, the optimal distributed approximation of spatial function is achieved with the least sensor feature level data, and then the data of other sensor nodes is characterized in the form of spatial function. The application can make the node with the largest contribution value to the spatial function approximation transmit data in priority, and the update of the spatial function can be completed without the transmission of data by all sensor nodes, achieving good effect of reducing energy consumption and time delay.

Description

Technical Field

[0001] This invention relates to the field of underwater acoustic sensors and is used as an efficient method for underwater target detection and perception. This invention can achieve efficient underwater data perception, greatly extend the lifespan of underwater acoustic sensor networks, and reduce latency. Background Technology

[0002] Reducing energy consumption and extending the lifespan of sensor networks through energy-efficient data aggregation methods has always been an important research topic in sensor networks. In recent years, the data aggregation problem in wireless sensor networks has attracted widespread attention. Researchers quickly realized that designing efficient data aggregation methods could significantly extend the lifespan of wireless sensor networks, and this has remained a crucial research area. Unlike the coexistence of multiple network structures in terrestrial wireless sensor networks, underwater acoustic sensor networks often employ clustered network structures to reduce energy consumption due to severe energy constraints in underwater environments. The core idea is to divide the entire underwater acoustic sensor network into several clusters, each consisting of a cluster head and several member sensor nodes. Cluster member sensors aggregate their data to the cluster head, and each cluster head transmits the aggregated data within its cluster to a fusion center using a specific routing algorithm. After receiving data from all cluster heads, the fusion center uses the sensor data within the network to complete a designated task according to a specific algorithm. While some research has been conducted on the data aggregation problem in underwater acoustic sensor networks, its depth is still significantly insufficient compared to terrestrial wireless sensor networks. Summary of the Invention

[0003] To overcome the shortcomings of existing technologies, this invention provides a data aggregation method for underwater acoustic sensor networks. Utilizing the spatial correlation of characteristic-level data from sensor nodes in the underwater acoustic sensor network, a spatial function is used to characterize the sensor data. An efficient distributed data transmission mechanism and an efficient adaptive distributed underwater channel allocation technique are designed to approximate the spatial function in real time, achieving a functional representation of the sensor network. This significantly reduces the number of sensor nodes requiring data transmission, greatly reducing energy consumption and latency.

[0004] First, a spatial function model is established using the spatial variation characteristics of feature-level data in UWSNs to characterize sensor nodes in the network, revealing the spatial variation characteristics of feature-level data in UWSN target detection and localization. Second, a spatial function approximation algorithm prioritizing the transmission of data with maximum approximation errors is proposed, enabling priority data transmission from nodes with maximum approximation errors in UWSNs, thus achieving distributed and rapid approximation of the spatial function. Then, fully utilizing the wireless broadcasting characteristics of underwater acoustic signals, a sequential optimal subset selection method and an optimal local error criterion are proposed. Through extreme value convergence, optimal distributed approximation of the spatial function is achieved using the minimum amount of sensor feature-level data, thereby characterizing the data of other sensor nodes in the form of a spatial function. To achieve optimal distributed approximation, this technique proposes an efficient adaptive distributed underwater channel allocation (ADAS) technology based on function approximation and characterization methods.

[0005] The technical solution adopted by this invention to solve its technical problem includes the following steps:

[0006] Step 1: Construct the space function;

[0007] Step 2: Receive data and calculate contribution value;

[0008] Assume the underwater acoustic sensor network contains N sensor nodes. After constructing the spatial function, X = [x1, x2, ..., x...]. i ,...,x n ] represents the location information of the sensor in the underwater acoustic sensor network. Given the feature-level data received by a node in the underwater acoustic sensor network as y(x i ;i∈S),

[0009] The sensor node collects feature-level data y(x) of the target. i Given the spatial function f(x; θ) and i∈S, calculate the local approximation error value D. m and contribution value

[0010] Step 3: Set the threshold value;

[0011] The node is based on the local approximation error value D m and contribution value Calculate the threshold value, where Door is the threshold value, and the threshold value is the data y acquired by sensor node i. i When the value exceeds the threshold Door, the sensor node transmits data; otherwise, it remains in sleep mode to save energy.

[0012] Step 4: Prioritize transmission of values ​​with higher contribution.

[0013] The local approximation error value D calculated based on the sensor node mThis is achieved by adjusting the backoff window value of nodes in the network;

[0014] Step 5: Jump to step 3, and continuously iterate through steps 3 and 4, checking whether the required accuracy is met;

[0015] During the data aggregation process in steps 3 and 4, the approximation of the spatial function is continuously monitored until the accuracy requirements are met, thus completing the real-time representation of the spatial function on the feature data of the underwater sensor network.

[0016] In step 1, the spatial function is constructed based on the data aggregation task. The spatial function model for the feature-level data of the underwater acoustic sensor network is as follows:

[0017]

[0018] Where f represents the spatial function, x represents spatial location information, θ represents other parameters of the kernel function, and κ(θ) p ;x) is the kernel function, the specific form of which is determined by the corresponding feature-level data space variation characteristics, θ p ω is the parameter of the corresponding kernel function. p ω represents the weights corresponding to the sensor nodes, p is the number of kernel functions, and ω represents the error at the sensor node.

[0019] In step 2, the local approximation error value D m and contribution value The calculation formula is:

[0020] D m =||y(x) m )-f(x m ;θ)|| 2

[0021] Where y(x) m () represents the actual data value at the node;

[0022]

[0023] in Let S(k) be the data transmission contribution value of node i, and the set of data already transmitted by k sensors.

[0024] In step 3, the threshold value is calculated:

[0025]

[0026] Door = D i (S(k))

[0027] Door is the threshold value.

[0028] In step 4, the local approximation error value D calculated by the sensor node is used. m This is achieved by adjusting the backoff window value of nodes in the network. The specific steps are as follows:

[0029] Backoff window CW configured for sensor node i i The value is set to:

[0030]

[0031] Where S all (i) represents the average distance S from node i to all remaining nodes in the underwater sensor network. allavg For all S all Calculate the average value, where cw is the initial backoff window value for node i.

[0032] The accuracy requirement is D. i <ε,D i ε represents the approximate error value for the current iteration, and ε is the accuracy standard for the current underwater sensor network data aggregation.

[0033] The beneficial effects of this invention are that by employing the underwater acoustic sensor network data aggregation method based on function approximation and representation proposed in this invention, and combining the characteristics of underwater sensor networks in channel allocation, the data aggregation of underwater sensor networks is integrated with the ideas of function approximation and representation. This allows underwater task-oriented data aggregation to be represented in real time by a continuously updated spatial function. Furthermore, by proposing a channel allocation method based on function approximation and representation, the node that contributes the most to the spatial function approximation can transmit data first, and the spatial function can be updated without all sensor nodes transmitting data, thus achieving a good effect of reducing energy consumption and latency. Attached Figure Description

[0034] Figure 1 This is a flowchart of the overall data aggregation process based on function representation and approximation.

[0035] Figure 2 This is a simulation diagram of the signal received by the sensor and the signal attenuation during signal propagation. Figure 2 (a) is the time-frequency diagram of the signal. Figure 2 (b) is a diagram showing the energy attenuation pattern during signal energy transmission.

[0036] Figure 3 This is a comparison chart of energy consumption and latency for ADAS with different numbers of nodes. Figure 3 Figure (a) is a comparison of energy consumption. Figure 3 (b) is a time delay comparison chart.

[0037] Figure 4 This is a comparison chart of ADAS throughput under different loads.

[0038] Figure 5 This is a comparison of the average throughput of ADAS with different numbers of nodes. Figure 5 Figure (a) shows a comparison of average throughput with 20 nodes. Figure 5 (b) shows the average throughput comparison under 40 nodes. Figure 5 (c) represents the average throughput comparison under 60 nodes.

[0039] Figure 6 This is a comparison of the average throughput of ADAS at different distances. Figure 6 (a) is a comparison chart of average throughput at 500m. Figure 6 (b) is a comparison chart of average throughput at 1000m. Figure 6 (c) is a comparison chart of average throughput at 1500m. Detailed Implementation

[0040] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0041] Step 1: The underwater acoustic sensor network consists of N sensor nodes, all within communication distance of each other. Therefore, each sensor node is a one-hop neighbor to any other node in the cluster. Topologically, this is equivalent to assuming all sensor nodes within the cluster form a clique in their connectivity graph. A spatial function model of the underwater acoustic sensor network's feature-level data is established. Different feature-level data are formally included, and a weighted superposition modeling method is used to model the spatial function of the underwater acoustic sensor network's feature-level data as follows:

[0042]

[0043] Where x represents spatial location information, κ(θ) p ;x) is the kernel function, the specific form of which is determined by the corresponding feature-level data space variation characteristics, θ p ω is the parameter of the corresponding kernel function. p ω represents the corresponding weight, p is the number of kernel functions, and ω represents the error.

[0044] Step 2: Sensor nodes in the underwater acoustic network acquire feature data of underwater acoustic signals through receivers and calculate the maximum contribution value of each node based on the acquired feature data. Where X = [x1, x2, ..., x...] i ,...,x n ] represents the location information of the sensor in the underwater acoustic sensor network. Given the feature-level data y(x) of the underwater acoustic sensor network. i ;i∈S), To achieve efficient and rapid aggregation of feature-level data, the overall mean square error shown in the following formula needs to be minimized while consuming the least amount of energy:

[0045]

[0046] R * D represents the overall mean square error. x Error can be measured using the average deviation or maximum deviation of the data, and is expressed as follows:

[0047]

[0048] Where D A D represents the average deviation of the available data. M The maximum deviation of the available data;

[0049] Given a given error precision D ≤ ε, where ε is the error precision, how can we find S such that |S| is minimized?

[0050] K = min{k|||D k -R * ||≤ε}

[0051] Where K satisfies |||D k -R * The smallest value of k ≤ ε, where k is the number of sensor nodes transmitting data. The method of sequentially selecting nodes so that each selected sensor node for transmitting data minimizes the error function can be expressed as:

[0052] like

[0053] m∈S c ={i; 1≤i≤N\S}

[0054] Where m is the sensor node number transmitting data, S c This is a collection of sensor nodes that have already transmitted data.

[0055] and

[0056] f m (x;θ)=E{f(x;θ)|(x m ,y(x m ))∪{(x i ,y(x i ));i∈S}}

[0057] Where f m (x; θ) is the current spatial function.

[0058] Search

[0059]

[0060] Where m * To obtain the sensor node number that minimizes the local approximation error;

[0061]

[0062] Repeat until D ≤ ε is satisfied.

[0063] For the error D at each sensor node m For unknown problems, local approximation errors are used as substitutes, considering two local error criteria. The first is the local approximation error criterion, which is the approximation error of the existing spatial function approximation formula to the local sensor feature-level data; the second is the posterior approximation error criterion, which is the overall approximation error of the updated spatial function approximation formula to the already transmitted data if the local sensor transmits data. They are expressed as follows:

[0064] Local approximation error D m for:

[0065] D m =||y(x) m )-f(x m ;θ)|| 2

[0066] Posterior approximation error D m for:

[0067]

[0068] Each sensor node can update its spatial function approximation expression using data from other nodes it hears, and simultaneously calculate local approximation errors using its own data. Whether to transmit data is determined based on the local approximation error and the required error accuracy. This process does not require information exchange between sensor nodes, making it a fully distributed data aggregation method.

[0069] If k sensors have already transmitted their data, denoted as S(k), then the contribution of node i is... Defined as the decrease in error caused by transmitting its data, that is:

[0070]

[0071] Therefore, given that k sensors have already transmitted their data S(k), the optimal sensor for the next data transmission is the node that is not among these k sensors and can make the largest contribution:

[0072]

[0073] Note the following relationship

[0074]

[0075] in

[0076]

[0077] Therefore, the local approximation error D is made m This serves as the basis for determining data transmission priority. If the local approximation error D at a certain sensor... m A large value indicates that the existing spatial function approximation expression cannot fully represent the local sensor data. This is achieved by addressing the local approximation error D. m Prioritizing the transmission of larger data sets can accelerate the convergence of the function approximation algorithm. Clearly, if each transmitted data set consists of sensor data with extremely large local approximation errors, then each transmission will maximize the reduction of the function approximation error, thereby maximizing the convergence of the function approximation algorithm.

[0078] This further transforms the data aggregation problem into a problem of sensor priority design and priority transmission of the largest contribution value. However, due to the varying distances between sensor nodes, fairness is difficult to guarantee during channel contention. Therefore, to address both the issues of fair channel contention and priority transmission of the largest contribution value, an adaptive aggregation strategy for ADAS is proposed.

[0079] Step 3: Effectively obtain the feature-level data with the largest approximation error in the network, and achieve function approximation of the underwater acoustic sensor network using a smaller number of sensor nodes. By collecting and calculating data at the sensor nodes and setting thresholds, the thresholds are continuously updated based on the data updates from each node, allowing each sensor node to autonomously update its local threshold. Utilizing the idea of ​​threshold separation, by performing threshold separation on the feature-level data in the underwater acoustic sensor network and setting thresholds, efficient function approximation is achieved while transmitting less data from fewer sensors.

[0080] Given the actual value y(x) at sensor node i i The approximate error at the sensor node is as follows:

[0081] D i =||y(x) i )-f(x i ;θ)|| 2

[0082] The local mean square error obtained at the sensor node is:

[0083]

[0084] To satisfy the given error accuracy D iUnder the condition of <ε (ε is the error accuracy), when the approximate error value at the sensor node is greater than the local mean square error of the sensor, the sensor transmits data, and the data transmitted by the sensor contributes more to the entire network. When the approximate error value at the sensor node is less than or equal to the local mean square error of the sensor, the sensor does not transmit data and enters a sleep state to save energy. This process is repeated until the spatial variation function of the underwater acoustic sensor network is approximated.

[0085] Step 4: To prioritize information transmission by nodes with higher contribution values ​​in the network, a high-efficiency adaptive distributed underwater channel allocation (ADAS) technique based on function approximation and representation is proposed. In underwater networks, underwater acoustic communication speeds are limited, and the distances between nodes vary significantly; therefore, achieving channel fairness is the primary task of priority control. First, a backoff timer is configured for the sensor nodes that will transmit data. Assume the backoff time for the i-th transmitting node is T. n The definition is as follows:

[0086] T n =cw n ·δ

[0087] Among them, cw n The initial timing value is randomly generated, ranging from [0, CW], where CW is the maximum value across all sensor nodes. δ represents a fixed time slot within which the transmitting node can listen to the channel's status information via a virtual carrier. The CW value configured for each sensor node needs to be optimized based on its location. This invention proposes an adaptive window strategy: a strategy that adjusts the contention window based on the distance to the remaining nodes (referring to transmitting nodes that have not yet gained channel control).

[0088] Assume that the arrival of data packets at each node follows a Poisson distribution with parameter λ, and nodes E and F are respectively located at... The probability that there will be no data packet collision with node C within the time period is:

[0089]

[0090]

[0091] Where D C,E D C,F Let P(C, E) represent the propagation delay of the RTS data packet between nodes CE and CF, respectively. P(C, E) represents the probability that C and E will not collide during this time interval, and P(C, F) represents the probability that C and F will not collide during this time interval. Now, assuming M is the total number of sending nodes, we know that the probability that node i will not collide with any of the remaining nodes is Pi. i :

[0092]

[0093] Where node j represents the remaining node excluding i, and D i,j This represents the propagation delay of RTS data packets between nodes i and j. From the formula above, we can see that... The smaller this value, the better P i The larger the value, the lower the collision rate of a single data packet transmission. Therefore, it can be concluded that the smaller the sum of the delays required for a node to send RTS packets to all remaining nodes, that is, the smaller the average distance between the node and all remaining nodes, the lower the collision rate of the node.

[0094] To reduce the impact of packet collisions, this invention proposes a contention window (CW) strategy based on distances to remaining nodes. This protocol sets the CW value of the sending node based on the distance between the sending node and the remaining nodes. The principle for setting the CW value is: sending nodes with smaller average distances to all remaining nodes have lower collision rates and can adopt a more aggressive sending strategy, i.e., appropriately reducing the CW value set for that node; conversely, a more conservative sending strategy is adopted, i.e., appropriately increasing the CW value set for that node. Now, let S be the average distance from sending node i to all remaining nodes. all (i):

[0095]

[0096] For all S all The average is:

[0097]

[0098] According to the CW configuration principles, the CW is configured for node i. i The proposed calculation formula is as follows:

[0099]

[0100] Where node j represents the remaining nodes excluding i, S i , j represents the distance between nodes i and j, and cw represents the initial value of the contention window originally set for the sending node.

[0101] Having resolved the issue of fair channel utilization among nodes at different locations, each sensor node receives feature data y(x) from the target in real time. i ;i∈S), and calculate the local approximation error value D based on the spatial function in UWSN. mWe know that the larger the approximation error value, the greater the contribution of that node to the network function approximation, and the more priority it should be for transmission. Therefore, based on the original cw, we make the window size inversely proportional to the approximation error value. In this way, after the node with the largest contribution value in the network transmits data, each node updates its own data, performs data processing and judgment. If the spatial function meets the accuracy requirement (D≤ε) at this time, each sensor node only collects target information and does not transmit data.

[0102] Based on the data aggregation mechanism described above, the CW configured for node i i The value is set to:

[0103]

[0104] Step 5: Repeat steps 3 and 4.

[0105] Step 6: Continuously check the approximation of the spatial function expression f(x; θ) during the process of continuously updating the threshold values ​​of the nodes and calculating the contribution values, until the underwater acoustic sensor network meets the accuracy requirement D. i <ε, to complete the real-time representation of the feature data of underwater sensor network by spatial functions.

[0106] Simulations were performed using the data aggregation method of this invention. The energy information received by the underwater sensor nodes was aggregated, and the resulting graph is shown below. Figures 3 to 6 As shown, it can be seen that among various data aggregation methods and under different circumstances, the method of the present invention is optimal.

Claims

1. A method for data aggregation in an underwater acoustic sensor network, characterized in that... Includes the following steps: Step 1: Construct the space function; Step 2: Receive data and calculate contribution value; The underwater acoustic sensor network is configured to contain N sensor nodes. After constructing the spatial function... This represents the location information of sensors in an underwater acoustic sensor network. Given the feature-level data received by a given underwater acoustic sensor network node, the data is... , ; Sensor nodes collect feature-level data of the target. and space functions Calculate the local approximation error value and contribution value ; Step 3: Set the threshold value; The node is based on the local approximate error value and contribution value Calculate the threshold value. The threshold value is the data acquired by sensor node i. Greater than the threshold value When the sensor node transmits data, it does so; otherwise, it remains in sleep mode to conserve energy. Step 4: Prioritize transmission of values ​​with higher contribution values; The local approximation error value calculated based on the sensor node This is achieved by adjusting the backoff window value of nodes in the network; The local approximation error value calculated based on the sensor node This is achieved by adjusting the backoff window value of nodes in the network. The specific steps are as follows: For sensor nodes Configured backoff window The value is set to: ； in For nodes The average distance to the remaining nodes in the entire underwater sensor network. For all Find the average value. For nodes The initial backoff window value; Step 5: Jump to step 3, and continuously iterate through steps 3 and 4, checking whether the required accuracy is met; During the data aggregation process in steps 3 and 4, the approximation of the spatial function is continuously monitored until the accuracy requirements are met, thus completing the real-time representation of the spatial function on the feature data of the underwater sensor network.

2. The underwater acoustic sensor network data aggregation method according to claim 1, characterized in that: In step 1, the spatial function is constructed based on the data aggregation task. The spatial function model for the feature-level data of the underwater acoustic sensor network is as follows: ； in Represents a space function. Indicates spatial location information, Other parameters representing the kernel function, The kernel function, whose specific form is determined by the characteristics of the corresponding feature-level data space variation, For the parameters of the corresponding kernel function, The weights corresponding to the sensor nodes. The number of kernel functions. This indicates the error at the sensor node.

3. The underwater acoustic sensor network data aggregation method according to claim 1, characterized in that: In step 2, the local approximation error value and contribution value The calculation formula is: ； in This represents the actual data value at the node. ； in The contribution value to data transmission for node i. The set of data transmitted by each sensor is .

4. The underwater acoustic sensor network data aggregation method according to claim 1, characterized in that: In step 3, the threshold value is calculated: ； ； in This is the threshold value.

5. The underwater acoustic sensor network data aggregation method according to claim 1, characterized in that: The accuracy requirement is: , This is the approximate error value for the current iteration. This serves as the accuracy standard for current underwater sensor network data aggregation.

Images