A method for collecting underwater acoustic data based on queuing model and genetic algorithm

Abstract

The application relates to an underwater acoustic network and discloses a water acoustic data collection method based on a queuing model and a genetic algorithm. In view of the problem of limited cluster head capacity in the underwater acoustic network, a single-service mixed queuing theory model is applied to the data transmission process between a cluster head and a sensor node in a data collection scene of an underwater node by using an AUV, so that the time when different cluster heads reach the upper limit of the cluster head capacity can be obtained. The time parameter and the distance between the cluster heads are used as the basis for designing the fitness function of the genetic algorithm. The time parameter and the time difference between the AUV and the cluster head are considered to cause data packet loss, and the weight between the two is adjusted according to the need. Then, the genetic algorithm is used to plan the travel path of the AUV. The purpose is to reduce the data loss caused by the limited node capacity under the premise of considering the energy loss of the AUV.

Description

TECHNICAL FIELD

[0001] The present application relates to underwater data collection, and in particular to an underwater acoustic data collection method based on a queuing model and a genetic algorithm. BACKGROUND

[0002] As human beings exploit marine resources more frequently, it is of great significance to establish an efficient and reliable underwater acoustic data collection network for human exploration of the ocean. However, the establishment of an underwater acoustic data collection network faces many difficulties such as complex and variable underwater environments, limited computing power, storage capacity and battery energy of sensor nodes. Therefore, how to effectively collect and transmit data is a very challenging task, and is also the main problem of many current underwater acoustic routing protocol researches. Routing selection plays a crucial role in effective data transmission, and many protocols have been proposed to make better path selection for different data transmission needs. Various algorithms and mathematical models have been used in underwater acoustic data collection networks.

[0003] Due to the complexity of the seabed environment, the data collection intensity in different areas is not the same. When collecting data from multiple underwater nodes, in addition to considering path loss, the data collection upper limit of the node also needs to be considered. However, the current commonly used underwater acoustic routing protocol lacks consideration of the capacity of different nodes, so it is necessary to use appropriate models and algorithms to design an underwater acoustic routing protocol that can reduce data loss and improve information integrity.

[0004] Establishing a system with cluster heads as the data collection center and multiple sensor nodes distributed around it for underwater acoustic data collection is a relatively efficient and energy-saving node deployment method. Using autonomous underwater vehicles (AUVs) for underwater acoustic data collection and transmission to buoy nodes is more flexible than data collection without AUVs, and AUVs have a wide range of activities, replaceable batteries, and are commonly used in underwater acoustic data collection. F. A. Alfouzan et al. (F. A. Alfouzan, et al., “A Novel Cross-layer Mobile Data-gathering Protocol for Underwater Sensor Networks,” 2020 IEEE 91st Vehicular Technology Conference (VTC2020-Spring), 2020.) proposed a cross-layer mobile data collection protocol for underwater sensor networks, which uses a group of cluster heads to aggregate data packets and is regularly visited by AUVs to keep the trip length short and reduce latency.

[0005] For the use of mathematical models, there are already many literatures that apply queuing theory to many resource allocation problems such as workshop material supply, urban road traffic, airport security check and taxi dispatching. However, the application of queuing theory in underwater acoustic data collection networks is less. Z. Fang et al. (Z. Fang, et al., “AoI Inspired Collaborative Information Collection for AUV Assisted Internet of Underwater Things,” IEEE Internet of Things Journal, 2021, 8: 14559-14571.) applied the queuing theory model to the data collection model of V-AUV (vertical motion AUV) and multiple H-AUV (horizontal motion AUV), derived the optimal upper limit of the number of AUVs serving in the queuing model system, and further ensured the freshness of the collected data. Y. J. Song (Y. J. Song, “Underwater Acoustic Sensor Networks With Cost Efficiency for Internet of Underwater Things,” IEEE Transactions on Industrial Electronics, 2021, 68: 1707-1716.) modeled each sensor as an M / G / 1 queue, derived the average queuing delay of underwater sensors to evaluate the communication performance, and further optimized the overall performance of three-dimensional UWSN.

[0006] In the routing protocol for optimal path selection, reinforcement learning algorithm, deep neural network algorithm, ant colony algorithm, genetic algorithm and other learning algorithms are proposed and used to solve optimization problems. Among them, genetic algorithm is favored in route planning problem because of its good global optimization ability. When facing the problem of underwater node data collection, S. Mahmoudzadeh et al. (S. Mahmoudzadeh, et al., “Optimal Route Planning with Prioritized Task Scheduling for AUV Missions,” 2015 IEEE International Symposium on Robotics and Intelligent Sensors (IRIS), 2015.) proposed a solution to the joint problem of large-scale route planning and task allocation of AUV, proved the feasibility of finding the best path for underwater tasks under given constraints, and concluded that route planning based on genetic algorithm can produce better results.

[0007] It can be seen that the node capacity reaches the upper limit, resulting in data loss problem has not been proposed in the literature on the optimal path planning of underwater data collection. The capacity of underwater nodes may be different, and the frequency of effective data generated in different areas of different nodes is also different. Therefore, it is necessary to consider the time difference between the time when the node reaches the capacity upper limit and the time when the AUV travels when designing the path optimization of underwater acoustic data collection. SUMMARY

[0008] The purpose of the present application is to solve the problem of data loss caused by limited storage capacity of cluster head in the process of underwater data collection in complex marine environment. A method of underwater acoustic data collection based on queuing model and genetic algorithm is provided. The mixed queuing theory model is applied to the process of underwater data collection, and the estimated time when the different cluster heads reach the capacity upper limit is obtained. The time parameter is used in the process of genetic algorithm path planning, and the distance of cluster head is also considered. The best collection path of AUV considering path loss and data loss is obtained.

[0009] The present application comprises the following steps:

[0010] 1) A certain number of sensor nodes, referred to as "nodes", are arranged on the seabed for underwater information data collection. The nodes are divided into several clusters, and a suitable node in each cluster is selected as the cluster head of the data collection center, which is responsible for collecting the data of the nodes deployed around it. During the data collection process of the cluster head, the AUV periodically visits the cluster head in a certain visiting order and collects data. Due to the limited capacity and energy of the AUV, it cannot work underwater for a long time. After visiting all the cluster heads, the AUV will return to the initial node and transmit the collected data to the sink node on the water surface. Since the cluster head is arranged on the seabed, the data collection process can be considered to be carried out in a two-dimensional scene.

[0011] 2) The process of each node in the cluster transmitting data packets to the cluster head is analogous to the process of a single-service mixed queuing model M / M / 1 / K. M represents an exponential distribution, and K is the system capacity. The process of nodes transmitting data to the cluster head is analogous to the process of customers arriving at the service counter one after another. The behavior of nodes transmitting data to the cluster head can be regarded as a service, and the corresponding data transmission time is the service time. The time when the nodes arrive at the cluster head and wait for service is the waiting time in the queuing theory. The number of cluster heads in each cluster is 1, which can be regarded as a single-service model. The data that the cluster head can store has a certain capacity upper limit, which can be regarded as the space capacity K of the queuing system. When the data that the cluster head can store reaches the capacity upper limit, it means that the space is fully occupied, and the newly arrived node data will be lost without being collected. When the data that the cluster head can store does not reach the capacity upper limit, the node will continue to transmit data.

[0012] 3) In the actual underwater scene, each data packet itself has different length; therefore, even if the cluster head capacity is fixed as K, the number of nodes that the cluster head can collect each time m is not the same; in addition, after the node transmits data to the cluster head, the data will occupy a certain capacity space in the cluster head, which is equivalent to the customer occupying the system space after receiving the service; it can be assumed that the process of the node transmitting data to the cluster head obeys the negative exponential distribution with parameter λ1, the time of the node transmitting data to the cluster head and the node exchanging data with the cluster head obeys the negative exponential distribution with parameter λ2, and the length of each data packet obeys the negative exponential distribution with parameter λ3, the lengths of the data packets are L1, L2, L3...L m ; the relationship between the cluster head capacity K and the data packet length L i satisfies:

[0013]

[0014] wherein, L i is the length of the i-th data packet, and m is the number of nodes that the cluster head can collect;

[0015] Since in the node-cluster head queuing system, the number of queued nodes is 0 to m-1, and satisfies the negative exponential distribution with parameter λ1, the process of the node transmitting data to the cluster head satisfies:

[0016]

[0017] And the time of the node exchanging data with the cluster head satisfies the negative exponential distribution with parameter λ2, that is:

[0018] λ 2n = λ2, n = 1, 2,..., m

[0019] According to the balance equation , the queue length distribution of the single-service mixed queuing system in the steady state is derived as:

[0020]

[0021] wherein, L s is the average queue length of the system, which corresponds to the space occupied by the node transmitting data packets in the cluster head in this scenario; Pn represents the degree of busyness of the system; Pn is the probability of n customers queuing in the queue;

[0022] 4) Since the queuing system has limited capacity, there are only m-1 positions, therefore, when the system space is occupied, the coming customers cannot enter the system to queue, that is, it cannot be guaranteed that all arriving customers can enter the system to wait for service; the customer arrival rate is λ1, the customers cannot enter the system when the system is in the m state, that is, the probability of the customers entering the system is 1-pm ; so the average number of customers that can actually enter the system per unit of time is:

[0023] λ e = λ1(1 - p m )

[0024] where p m is the loss rate of customers; so the average residence time is:

[0025]

[0026] In this scenario, the time corresponding to the data exchange between the node and the cluster head and the time corresponding to the data transmission from the node to the cluster head can be regarded as the service time T s ; the time corresponding to the waiting of the node for the data exchange with the cluster head can be regarded as the queuing time T w ; it can be considered that in the node-cluster head system, the average residence time is:

[0027] W s = T s + T w

[0028] Since after the data transmission from the node to the cluster head, the data will still occupy a certain capacity space in the cluster head, the estimated time for the cluster head to reach the capacity limit is:

[0029] T = m * W s

[0030] 5) The path of the AUV is planned by using the genetic algorithm, and the coding form is that the visited cluster heads are numbered and arranged in combination, which is used as the representative of the order of the visited cluster heads in the path planning process; different arrangement combinations are different individuals in the genetic algorithm, which can be represented by X i , where i represents the ith individual; the random arrangement combination with a quantity of N is defined as the initialization population, and the population size affects the population diversity and the efficiency of the algorithm;

[0031] 6) The time parameter T derived above is applied in the objective function and the fitness function of the genetic algorithm, and the fitness function of the individual X i can be set as the reciprocal of the individual objective function value; since the AUV will be affected by the sea current factor in the actual underwater navigation process, the position and state of the AUV will change, and then the energy loss will be caused; therefore, under the assumption that the direction and speed of the sea current are certain, the energy loss can be represented by the motion model under the disturbance of the sea current:

[0032] J(m, n) = {(v * cos θ - v x ) 2 + (v * sin θ - vy ) 2}*d mn / v

[0033] E =∑J(m, n)

[0034] where J(m, n) is the energy consumption between cluster head m and cluster head n, v is the AUV travel speed, θ is the angle between the sailing speed and the x-axis, v x and v y are the velocity components of the sea current on the x-axis and y-axis respectively, d mn is the distance between cluster head m and cluster head n;

[0035] The target function value is composed of three parts, which are the total distance of AUV traveling in the order of individual X i , the absolute value of the time difference between the time loss of AUV reaching the cluster head and the estimated time T of the cluster head reaching the upper limit of the capacity, and the energy loss caused by the sea current factor; the three are added in a certain proportion to obtain the target function value, that is:

[0036]

[0037] where β, α and γ are the weights of the total distance, the total time difference and the sea current loss respectively, len(X i ) is the total path with the visiting order of individual X i , D is the path experienced before visiting cluster head j in individual X i , and E(X i ) is the total energy consumption of AUV caused by the sea current with the visiting order of individual X i ;

[0038] Then the fitness function is:

[0039] fitness(X i ) = I / F(X i )

[0040] Through genetic operations such as crossover and mutation, the paths of different visiting orders of AUV can be obtained, and the fitness is calculated to obtain the probability of selecting the path;

[0041] 7) The roulette selection strategy is adopted to calculate the fitness of each individual and the probability of being inherited to the next generation, the probability values are accumulated and normalized, and the individual is selected by comparing the randomly generated value in the interval [0, 1] with the cumulative probability; the probability P(X i ) that individual X i is selected is:

[0042]

[0043] Where N is the number of cluster heads, fitness(x) i For individual X i The fitness function; then, the cumulative probability for:

[0044]

[0045] 8) Continuously try to find reasonable settings for the initial population size, crossover probability, and mutation probability; in the initial population design, if the number of individuals in the population is too small, it will lead to a reduction in diversity, while if the number of individuals is too large, it will lead to a reduction in algorithm efficiency; when setting the parameters of genetic operations, the magnitude of crossover probability and mutation probability will affect the convergence of the algorithm, leading to premature convergence or non-convergence.

[0046] This invention can effectively reduce path loss and data packet loss during underwater data collection, thereby reducing energy consumption and further improving the integrity of data information.

[0047] This invention takes into account the complex marine environment, the varying data collection intensity in different areas, and the limited storage capacity of cluster heads. It utilizes a single-server hybrid queuing model suitable for underwater data collection scenarios to derive the estimated time for different cluster heads to reach their capacity limits. Furthermore, it uses a genetic algorithm to plan the AUV's travel path, adjusting the weight ratio of path loss and data loss according to different scenarios to obtain the optimal AUV path, thus further optimizing the underwater data collection scheme.

[0048] The present invention has the following outstanding advantages:

[0049] 1) The single-server hybrid queuing theoretical model is applied to the underwater data collection scenario, and the theoretical formula is derived according to the actual scenario to obtain the expected time parameters for the queue leader and the cluster head to reach the upper limit. Then, the parameters are used to analyze the data collection process and optimize resource allocation.

[0050] 2) By taking the estimated time for the cluster head to reach the upper limit and the distance between the cluster heads as factors influencing the path planning of the genetic algorithm, a path that comprehensively considers path loss and data integrity can be selected, thus solving the problem of data packet loss caused by the limited storage capacity of the cluster heads in underwater scenarios. Attached Figure Description

[0051] Figure 1 This is a scene diagram illustrating the underwater data collection method based on queuing models and genetic algorithms of the present invention.

[0052] Figure 2 The queuing theory model of the underwater acoustic data collection method based on queuing model and genetic algorithm of this invention is consistent with the underwater data collection scenario.

[0053] Figure 3 Figure 1 is a flow chart of the genetic algorithm of the method for collecting underwater acoustic data based on the queuing model and the genetic algorithm according to the present application.

[0054] Figure 4 Figure 2 is a convergence chart of the genetic algorithm of the method for collecting underwater acoustic data based on the queuing model and the genetic algorithm according to the present application.

[0055] Figure 5 Figure 3 is a random path route chart of the embodiment of the method for collecting underwater acoustic data based on the queuing model and the genetic algorithm according to the present application.

[0056] Figure 6 Figure 4 is an optimization route chart of the genetic algorithm without using the queuing model in the embodiment of the method for collecting underwater acoustic data based on the queuing model and the genetic algorithm according to the present application.

[0057] Figure 7 Figure 5 is an optimization route chart of the genetic algorithm with using the queuing model in the embodiment of the method for collecting underwater acoustic data based on the queuing model and the genetic algorithm according to the present application. DETAILED DESCRIPTION

[0058] The present application will be described in detail below in combination with the drawings and specific embodiments.

[0059] The embodiment of the present application comprises the following steps:

[0060] 1) A certain number of sensor nodes, referred to as "nodes", are arranged on the seabed for underwater information data collection; the nodes are divided into several clusters, and a suitable node is selected as the cluster head of the data collection center in each cluster, which is responsible for collecting the data of the nodes arranged around it. During the data collection of the cluster head, the AUV regularly visits the cluster head in a certain visiting order and collects data. Since the capacity and energy of the AUV are limited, it cannot work underwater for a long time, and it is stipulated that after visiting each cluster head, the AUV will return to the initial node and transmit the collected data to the sink node on the water surface. Since the cluster head is arranged on the seabed, it can be considered that the data collection process is carried out in a two-dimensional scene.

[0061] 2) The process of each node in the cluster transmitting data packets to the cluster head is analogous to the process of a single-service mixed queuing model M / M / 1 / K. M represents an exponential distribution, and K is the system capacity. The process of the node transmitting data to the cluster head is analogous to the process of customers arriving at the service counter one after another. The behavior of the node transmitting data to the cluster head can be regarded as a service, and the corresponding data transmission time is the service time. The time of the node arriving at the cluster head and waiting for service is the waiting time in the queuing theory. The number of cluster heads in each cluster is 1, which can be regarded as a single-service model. The cluster head has a certain upper limit of the storage capacity, which can be regarded as the space capacity K of the queuing system. When the data that the cluster head can store reaches the upper limit of the capacity, it means that the space is fully occupied, and the newly arrived node data will be lost without being collected; when the data that the cluster head can store does not reach the upper limit of the capacity, the node will continue to transmit data.

[0062] 3) In the actual underwater scene, each data packet itself has a different length. Therefore, even if the cluster head capacity is fixed as K, the number of nodes that the cluster head can collect each time m is not the same. In addition, after the node transmits data to the cluster head, the data occupies a certain capacity space in the cluster head, which is equivalent to the customer occupying the system space after receiving the service. It can be assumed that the process of the node transmitting data to the cluster head obeys a negative exponential distribution with parameter λ1. The time of the node transmitting data to the cluster head and the node exchanging data with the cluster head obeys a negative exponential distribution with parameter λ2. The length of each data packet obeys a negative exponential distribution with parameter λ3. The lengths of the data packets are L1, L2, L3,..., Lm respectively. Then the relationship between the cluster head capacity K and the data packet length Lm satisfies: m i

[0063]

[0064] wherein L i is the length of the i-th data packet, and m is the number of nodes that the cluster head can collect.

[0065] Since the number of nodes in the node-cluster head queuing system is 0 to m-1, and satisfies the negative exponential distribution with parameter λ1, the process of the node transmitting data to the cluster head satisfies:

[0066]

[0067] The time of the node exchanging data with the cluster head, then obeys a negative exponential distribution with parameter λ2, that is:

[0068] λ 2n = λ2, n = 1, 2,..., m

[0069] According to the balance equation ​​The queue length distribution of the single-server mixed queuing system in the steady state is derived as follows:

[0070]

[0071] wherein L s is the average queue length of the system, corresponding to the space occupied by the node transmission data packet in the cluster head in this scenario; denotes the busy degree of the system; Pn is the probability of n customers queuing in the queue.

[0072] 4) Since the queuing system has a limited capacity, only m-1 positions, when the system space is full, the coming customers cannot enter the system for queuing, that is, it cannot be guaranteed that all the arriving customers can enter the system for waiting for service. The customer arrival rate is λ1, and the customers cannot enter the system when the system is in the m state, that is, the probability of customers entering the system is 1-p m . Therefore, the average number of customers that can actually enter the system per unit time is:

[0073] λ e = λ1(1-p m )

[0074] wherein p m is the customer loss rate. Therefore, the average stay time is:

[0075]

[0076] In this scenario, the time for the node to exchange data with the cluster head and the time for the node to transmit data to the cluster head can be regarded as the service time T s ; the time for the node to wait for data exchange with the cluster head can be regarded as the queuing time T w ; it can be considered that in the node-cluster head system, the average stay time is:

[0077] W s = T s + T w

[0078] Since after the node transmits data to the cluster head, the data will still occupy a certain capacity space in the cluster head, the estimated time for the cluster head to reach the capacity upper limit is:

[0079] T = m * W s

[0080] 5) The path of AUV is planned by genetic algorithm. The encoding form is to number the visited cluster heads and arrange the serial numbers to represent the order of the cluster heads visited in the path planning process. Different arrangement of the sequence is different individual in the genetic algorithm, which can be represented by Xi, where i represents the ith individual. Define the random arrangement of N as the initial population, and the population size affects the population diversity and the efficiency of the algorithm.

[0081] 6) The time parameter T derived above is applied to the objective function and fitness function of genetic algorithm. The fitness function of individual X i can be set as the inverse of the individual objective function value. Since AUV will be affected by the sea current during the actual underwater navigation, the position and state of AUV will change, which will cause energy loss. Therefore, under the assumption that the direction and speed of the sea current are constant, the energy loss can be represented by the motion model under the disturbance of the sea current as follows:

[0082] J(m, n) = {(v * cos θ - v x ) 2 + (v * sin θ - v y ) 2} * d mn / v

[0083] E = ∑J(m, n)

[0084] Where J(m, n) is the energy consumption between cluster head m and cluster head n, v is the AUV travel speed, θ is the angle between the navigation speed and the x-axis, v x and v y are the velocity components of the sea current on the x-axis and y-axis, respectively, and d mn is the distance between cluster head m and cluster head n.

[0085] The objective function value includes three parts, which are the total distance of AUV traveling in the order of individual X i , the absolute value of the time difference between the time loss of AUV to the cluster head and the estimated time T of the cluster head to the upper limit of the capacity, and the energy loss caused by the sea current factor. The three are added in a certain proportion to obtain the objective function value, that is:

[0086]

[0087] Where β, α and γ are the weights of the total distance, the total time difference and the sea current loss, respectively, len(X i ) is the total path with the order of individual X i as the visiting order, D is the path experienced before visiting cluster head j in individual X i , and E(X iis an individual X i The total energy consumption of the AUV caused by the sea current in the access sequence.

[0088] The fitness function is:

[0089] fitness(X i ) is an individual X i )

[0090] Through genetic operations such as crossover and mutation, the path of different access sequences of the AUV can be obtained, and the fitness is calculated to obtain the probability of selecting the path.

[0091] 7) The roulette selection strategy is adopted, the fitness of each individual and the probability of being inherited to the next generation are calculated, the probability values are accumulated and normalized, and whether the individual is selected is obtained by comparing the randomly generated value in the [0, 1] interval with the cumulative probability. The individual X i The selected probability P(X i ) is:

[0092]

[0093] Wherein, N is the number of cluster heads, fitness(x i ) is the fitness function of the individual X i , and the cumulative probability is:

[0094]

[0095] 8) The reasonable setting values of the initial population number, crossover and mutation probability are constantly tried to find out. In the design of the initial population, if the number of population individuals is too small, the diversity will be reduced, and if the number of individuals is too large, the algorithm efficiency will be reduced. When the parameters of genetic operation are set, the size of the crossover and mutation probability will affect the convergence of the algorithm, resulting in the results of premature convergence or non-convergence.

[0096] Figure 1 The underwater data collection scene diagram of the underwater acoustic data collection method based on the queuing model and the genetic algorithm is given. Figure 2 The underwater acoustic data collection method based on the queuing model and the genetic algorithm conforms to the queuing theory model of the underwater data collection scene. Figure 3 The genetic algorithm flowchart of the underwater acoustic data collection method based on the queuing model and the genetic algorithm is given.

[0097] The feasibility of the method described in the application is verified by computer simulation.

[0098] The simulation scene is set as a two-dimensional underwater scene of 10kmx10km, in which 8 cluster heads are randomly distributed, 5-10 sensor nodes around the cluster head transmit data to the cluster head, and the AUV travels to the cluster head at a speed of 2m / s to collect data. In order to verify the feasibility of the method proposed in the application, the time T at which the cluster head reaches the upper limit of the capacity is set to be different. T is related to the arrival rate of the customer successive arrival system, i.e. the frequency of the node transmitting data to the cluster head and the cluster head capacity, etc. Therefore, the simulation scene set in this embodiment has two types: the first type is to arrange cluster heads with different capacities in each region when deploying nodes and cluster heads, and the cluster head capacity is set to be 100-300MB, and the data successive arrival parameter λ1=0.014 / s; the second type is to arrange cluster heads with a capacity of 200MB in regions with different data generation frequencies, and the data successive arrival parameter is set to be in the range of 0.013-0.023 / s.

[0099] Other parameters are set as follows: population size M=200, iteration number C=2000, elimination acceleration index k=2, crossover probability p c =0.4, mutation probability p m =0.2, time difference TD proportion α=0.5, path proportion β=0.3, current loss proportion γ=0.2, data transmission rate v b =20kbps.

[0100] The simulation steps are as follows:

[0101] (1) Initialize the population: randomly generate 8 points as cluster heads in the space of 10000x10000 and number the cluster heads, arrange the 200 individuals generated by combining the 8 cluster head numbers as the initial population M, and define the above parameter values one by one;

[0102] (2) Queuing theory application: present the queuing theory formula derivation process in the form of Matlab code, calculate by computer, and obtain the expected upper limit time T of the cluster head under different cluster head capacities or different marine environments;

[0103] (3) The individual distance calculation function, the individual time difference calculation function and the sea current loss calculation function are added in the proportion of time difference proportion a=0.5, path proportion β=0.3 and sea current loss proportion γ=0.2, and the reciprocal is obtained to obtain the fitness function. The individual distance calculation function is to obtain the total path of the AUV, the individual time difference calculation function is the difference between the total time cost before the AUV reaches a cluster head and the estimated time of the cluster head reaching the upper limit of the capacity, and the sum of the differences of all cluster heads under the corresponding path of the individual is obtained. The normalized fitness function value is used as the individual selection probability, a certain number of individuals are selected by the roulette strategy, and other individuals are obtained by genetic operation of crossover and mutation, and the fitness function is calculated, and the individual with the best fitness is selected as the best planning path.

[0104] Through Matlab simulation, a simulation result graph is obtained. Figure 4 The genetic algorithm convergence graph is used for path planning by using the genetic algorithm, and the convergence of the genetic algorithm needs to be ensured, that is, the algorithm converges to a fixed value, that is, the global optimal solution, as the state migrates.

[0105] Figures 5 to 7respectively, the path trajectory diagram of the random path, the genetic algorithm optimized path without using the queuing theory and the genetic algorithm optimized path using the queuing theory can be seen. It can be seen from the displayed data on the comparison diagram that before optimization using the queuing theory and the genetic algorithm, the total path of the AUV running and the absolute value TD of the time difference between the AUV running process and the upper limit of the cluster head arrival capacity are 39.906km and 39.1266h respectively, and the path energy consumption and the time difference result in a large amount of data packet loss. Under the optimization trajectory of the queuing model without considering the path loss, the path loss and TD are 27.9248km and 30.8257h respectively; the total path of the AUV running is obviously reduced, and the path loss is about 30.0% less than that of the random path; TD, as a parameter for comparing the path time consumption and the upper limit of the cluster head arrival capacity, can reflect the change of the relative packet loss rate of different schemes, and it can be seen that compared with the random path, the optimization trajectory considering only the path loss reduces the packet loss rate by about 21.2%. Under the optimization method considering the path loss and the time difference loss with the queuing theory, the total path and TD are 28.5095km and 14.828h respectively, although the path loss is increased by about 2.1% compared with that before using the queuing theory, the packet loss rate caused by the time difference loss is reduced by 48.1% compared with that before using the queuing theory; compared with the random running trajectory, the optimization effect is very obvious, and the path loss and the packet loss rate are reduced to 71.4% and 37.9% of the random path respectively. Therefore, the optimization method proposed in the application can effectively reduce the information loss degree while ensuring a small path loss.

[0106] The application models the underwater data collection scene by using the single-service mixed queuing theory model, transforms the queuing theory according to the actual underwater scene, and makes corresponding theoretical formula derivation and transformation, to obtain the estimated time of the upper limit of the arrival capacity of each cluster head in the underwater data collection scene. The time parameter and the time consumed by the AUV running to the corresponding cluster head have a difference, and the time difference will affect the completeness of the data information, and its size can be used as an important index for measuring the amount of data packet loss. The index and the AUV energy loss caused by the distance between the cluster heads are added in a certain proportion, and the result is used as an influencing factor of the fitness function of the genetic algorithm; the greater the time difference and the path loss, the smaller the fitness. By comprehensively considering the path loss and the genetic algorithm of data loss, the AUV path is planned, and the best data collection path can be obtained. The application considers the complex marine environment, the different data collection intensities in different regions and the limited storage capacity of the cluster head, combines the queuing theory model and the genetic algorithm, and can effectively reduce the path loss and the data packet loss in the underwater data collection process, reduce the energy consumption and further improve the completeness of the data information.

Claims

1. A method for collecting underwater acoustic data based on queuing model and genetic algorithm, characterized in that The method comprises the following steps: 1) a plurality of sensor nodes are arranged on the seabed for underwater information data collection; the nodes are divided into a plurality of clusters, and a suitable node is selected as a cluster head of a data collection center in each cluster, which is responsible for collecting data of the nodes arranged around the cluster head; during data collection of the cluster head, an AUV regularly visits the cluster head in a certain visiting order and collects data; due to the limited capacity and energy of the AUV, the AUV cannot work underwater for a long time, and is required to return to the initial node after visiting each cluster head once, and transmit the collected data to a surface buoy node sink; since the cluster head is arranged on the seabed, the data collection process is considered to be carried out in a two-dimensional scene; 2) the process of transmitting a data packet from each node in the cluster to the cluster head is analogous to the process of a single-service mixed queuing model M / M / 1 / K; wherein M represents an exponential distribution, and K is the cluster head space capacity; the process of transmitting data from the node to the cluster head is regarded as the process of customers arriving at the service station one after another, and the behavior of the node transmitting data to the cluster head is regarded as a service, and the corresponding data transmission time is the service time, and the time of the node arriving at the cluster head and waiting for service is the waiting time in the queuing theory; the number of cluster heads in each cluster is 1, which is regarded as a single-service model; the data that can be stored by the cluster head has an upper limit of capacity, and the upper limit is the cluster head space capacity K of the queuing system; when the data that can be stored by the cluster head reaches the upper limit of capacity, it indicates that the space is fully occupied, and the newly arrived node data will be lost without being collected; when the data that can be stored by the cluster head does not reach the upper limit of capacity, the node will continue to transmit data; 3) In real-world underwater scenarios, each data packet has a different length; even if the cluster head space capacity is fixed at K, the number of nodes that the cluster head can collect at one time is N. c They are also different; after a node transmits data to the cluster head, the data occupies a certain amount of capacity space in the cluster head, just as a customer still occupies system space after receiving a service; assuming that the process of a node transmitting data to the cluster head follows a negative exponential distribution with parameter λ1, the time for a node to transmit data to the cluster head and the time for a node to exchange data with the cluster head follow a negative exponential distribution with parameter λ2, and the length of each data packet follows a negative exponential distribution with parameter λ3, and the lengths of the data packets are L1, L2, L3...L Nc The cluster head space capacity K is related to the data packet length L. i The relationship satisfies: wherein L i is the length of the ith data packet, N c is the number of nodes that the cluster head can collect; In the node-cluster head queuing system, the number of queued nodes is 0 to N c -1, and satisfies the negative exponential distribution with parameter λ1, so that the process of the node transmitting data to the cluster head satisfies: The time of data exchange between the node and the cluster head is subject to a negative exponential distribution with a parameter λ2, that is: λ 2n = λ2,n = 1, 2, …, N c According to the balance equation The queue length distribution of the single-server mixed queuing system in the steady state is derived as where L s is the average queue length of the system, which corresponds to the space occupied by the data packets transmitted by the nodes in the cluster head in this scenario; represents the degree of busyness of the system; P i is the probability of having i customers in the queue; 4) Since the queuing system capacity is limited, there are only Nc-1 positions, therefore, when the system space is occupied, the next customer cannot enter the system queue, i.e. it cannot be guaranteed that all arriving customers can enter the system to wait for service; the customer arrival rate is λ1, and the customer cannot enter the system when the system is in a full state, i.e. the probability of the customer entering the system is 1-p Nc ; therefore, the average number of customers that can actually enter the system per unit time is: λ e = λ1(1 - p Nc ) where p Nc is the customer loss rate; the average dwell time is: Corresponding in this scenario, the time of the node exchanging data with the cluster head and the time of the node transmitting data to the cluster head as service time T s ; the time of the node waiting for data exchange with the cluster head as the queuing time T w ; in the node-cluster head system, the average residence time is: W s = T s + T w After the node transmits data to the cluster head, the data still occupies a certain capacity space in the cluster head, and therefore, the estimated time for the cluster head to reach the capacity upper limit is: T = Nc*W s 5) The path planning of AUV by genetic algorithm, the encoding form is to arrange the cluster head number and the serial number combination as the representative of the cluster head order in the path planning process; the different arrangement combination sequence is the different individual in the genetic algorithm, represented by X i where i represents the ith individual; define the random arrangement combination of the number N as the initialization population, the population size affects the population diversity and the algorithm efficiency; 6) The time parameter T derived above is applied in the objective function and fitness function of the genetic algorithm, and the fitness function of the individual X i is set as the inverse of the individual objective function value; since the AUV will be affected by the sea current factor during actual underwater navigation, resulting in changes in the position and state of the AUV, and further causing energy loss; therefore, under the assumption that the direction and speed of the sea current are constant, the energy loss is represented by the motion model under the disturbance of the sea current: J(i,j) = {(v*cos θ - v x ) 2 +(v*sin θ - v y ) 2}*d ij / v E=∑J(i,j) where J(i, j) is the energy consumption between cluster head i and cluster head j, v is the AUV travel speed, θ is the angle between the travel speed and the x-axis, v x , v y are the velocity components of the sea current in the x-axis and y-axis, respectively, and d ij is the distance between cluster head i and cluster head j. The target function value includes three parts, respectively: the total distance of the AUV in the individual X i order, the absolute value of the time difference between the time loss of the AUV to the cluster head and the estimated time T of the cluster head to the upper limit of the capacity, and the energy loss caused by the sea current factor; the three are added in a certain proportion to obtain the target function value, that is: where β, α and γ are the weights of total distance, total time difference and current loss, respectively, len(X i ) is the total path with the sequence of individual X i , D is the path experienced before accessing the cluster head j in individual X i , E(X i ) is the total energy consumption of AUV caused by current with the sequence of individual X i . The fitness function is: fitness(X i ) = 1 / F(X i ) Through genetic operations such as crossover and mutation, the path of different visiting orders of the AUV is obtained, and the probability of selecting the path is calculated by calculating the fitness; 7) Roulette wheel selection strategy is adopted to calculate the fitness of each individual and the probability of being inherited to the next generation. The probability values are accumulated and normalized. By randomly generating a value in the interval [0, 1] and comparing it with the cumulative probability, it is determined whether the individual is selected or not; individual X i The probability P(X i ) of being selected is: Where N is the number of cluster heads, fitness(X i ) is the fitness function of individual X i ; then, the cumulative probability is: 8) the reasonable setting values of the initial population size, crossover and mutation probabilities are constantly tried to find; in the initial population design, too small a number of population individuals will lead to reduced diversity, and too many population individuals will lead to reduced algorithm efficiency; when the parameters of the genetic operation are set, the sizes of the crossover and mutation probabilities will affect the convergence of the algorithm, leading to the results of premature convergence or non-convergence.

Images