A multi-objective optimization-based underwater acoustic task intelligent scheduling method

Abstract

This invention provides an intelligent scheduling method for underwater acoustic tasks based on multi-objective optimization, comprising: upon receiving an underwater acoustic task, determining the data dependencies between subtasks and the resource requirements of each subtask; based on the resource requirements of each subtask, querying resources that meet the requirements and generating a list of available resources for each subtask; based on the list of available resources for each subtask, reading the physical connection relationships and performance parameters of the corresponding available resources; and based on the data dependencies between subtasks, the list of available resources for each subtask, the physical connection relationships of the available resources, and the performance parameters of the available resources, employing a multi-objective optimization algorithm to obtain a resource scheduling strategy that meets the resource requirements. By implementing this scheme, the probability of task execution stuttering due to insufficient consideration of data dependencies can be reduced.

Description

Technical Field

[0001] This invention belongs to the field of resource scheduling technology, specifically relating to an intelligent scheduling method for underwater acoustic tasks based on multi-objective optimization. Background Technology

[0002] With the deepening development of applications such as underwater acoustic signal processing, target recognition, and tracking, the computational tasks exhibit high heterogeneity and data dependency. A typical underwater acoustic processing task chain usually consists of multiple stages, including signal preprocessing, beamforming, feature extraction, and classification. These tasks not only have differentiated requirements for computing resources (CPU, memory, GPU), but also involve frequent and large-scale data interaction between tasks, making them extremely sensitive to network communication bandwidth and latency. Traditional scheduling strategies often only focus on the static allocation of computing resources, ignoring the network overhead caused by data flow between tasks, which can easily lead to network congestion becoming a performance bottleneck in the data processing pipeline. Summary of the Invention

[0003] In view of this, the purpose of this invention is to provide an intelligent scheduling method for underwater acoustic tasks based on multi-objective optimization, so as to meet the need to reduce the probability of task execution lag.

[0004] To achieve the above objectives, the present invention provides the following technical solution:

[0005] According to a first aspect, the present invention provides an intelligent scheduling method for underwater acoustic tasks based on multi-objective optimization, comprising:

[0006] Upon receiving an underwater acoustic task, the system determines the data dependencies between subtasks and the resource requirements of each subtask. Based on the resource requirements of each subtask, it queries for resources that meet the requirements and generates a list of available resources for each subtask. Based on the list of available resources for each subtask, it reads the physical connection relationships and performance parameters of the corresponding available resources. Based on the data dependencies between subtasks, the list of available resources for each subtask, the physical connection relationships of the available resources, and the performance parameters of the available resources, a multi-objective optimization algorithm is used to obtain a resource scheduling strategy that meets the resource requirements.

[0007] According to a second aspect, the present invention provides an intelligent scheduling device for underwater acoustic tasks based on multi-objective optimization, comprising:

[0008] The task splitting module is used to determine the data dependencies between subtasks and the resource requirements of each subtask when an underwater acoustic task is received. The resource determination module is used to query resources that meet the resource requirements of each subtask and generate a list of available resources for each subtask. The reading module is used to read the physical connection relationships and performance parameters of the available resources based on the list of available resources for each subtask. The scheduling strategy decision module is used to obtain a resource scheduling strategy that meets the resource requirements by using a multi-objective optimization algorithm based on the data dependencies between subtasks, the list of available resources for each subtask, the physical connection relationships of available resources, and the performance parameters of available resources.

[0009] According to a third aspect, embodiments of the present invention provide an electronic device, the device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the steps of the intelligent scheduling method for underwater acoustic tasks based on multi-objective optimization as described in the first aspect or any embodiment of the first aspect.

[0010] According to a fourth aspect, embodiments of the present invention provide a computer storage medium storing computer instructions thereon, which, when executed by a processor, implement the steps of the intelligent scheduling method for underwater acoustic tasks based on multi-objective optimization as described in the first aspect or any embodiment of the first aspect.

[0011] This embodiment provides an intelligent scheduling method for underwater acoustic tasks based on multi-objective optimization. It integrates sub-task data dependencies, resource performance adaptability, and resource physical connectivity, generating a globally optimal scheduling strategy through a multi-objective optimization algorithm. This embodiment reduces the probability of task execution stuttering due to insufficient consideration of data dependencies and, through precise resource selection and multi-objective optimization, is applicable to complex and diverse underwater acoustic task scenarios.

[0012] Other advantages, objectives, and features of the invention will be set forth in the following description and will be apparent to those skilled in the art in some respects, or may be learned by practice of the invention. The objectives and other advantages of the invention can be realized and obtained through the following description. Attached Figure Description

[0013] To make the objectives, technical solutions, and beneficial effects of this invention clearer, the following figures are provided for illustration:

[0014] Figure 1 This is a flowchart illustrating a specific example of an intelligent scheduling method for underwater acoustic tasks based on multi-objective optimization in this invention.

[0015] Figure 2This is a schematic diagram of a module structure of an intelligent underwater acoustic task scheduling device based on multi-objective optimization according to the present invention;

[0016] Figure 3 This is a schematic block diagram of a specific example of an electronic device in an embodiment of the present invention. Detailed Implementation

[0017] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can also refer to the internal connection of two components; and they can refer to a wireless connection or a wired connection. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0019] Furthermore, the technical features involved in the different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0020] This invention provides an intelligent scheduling method for underwater acoustic tasks based on multi-objective optimization, such as... Figure 1 As shown, it includes:

[0021] S101, When an underwater acoustic task is received, determine the data dependencies between the subtasks in the underwater acoustic task and the resource requirements of each subtask.

[0022] S102, based on the resource requirements of each subtask, query the resources that meet the requirements and generate a list of available resources for each subtask;

[0023] S103, based on the available resource list of each subtask, read the physical connection relationship of the corresponding available resources and the performance parameters of the available resources;

[0024] S104. Based on the data dependencies between subtasks, the list of available resources for each subtask, the physical connection relationships of available resources, and the performance parameters of available resources, a multi-objective optimization algorithm is used to obtain a resource scheduling strategy that meets resource requirements.

[0025] For example, when the system receives an underwater acoustic task, it first needs to break it down into several sub-tasks. For instance, an underwater acoustic task for underwater environment monitoring can be broken down into data acquisition sub-tasks, data preprocessing sub-tasks, data transmission sub-tasks, and data analysis sub-tasks. Then, a task analysis tool is used to analyze the data dependencies between the sub-tasks. This tool stores sub-task breakdown data, data dependencies between sub-tasks, and resource requirements for various typical underwater acoustic tasks. Upon receiving an underwater acoustic task, the system reads the data dependencies between the sub-tasks and the resource requirements of each sub-task from the matched underwater acoustic task list.

[0026] Taking the aforementioned monitoring task as an example, the data preprocessing subtask relies on the raw monitoring data output by the data acquisition subtask, while the data transmission subtask relies on the standardized data processed by the data preprocessing subtask. This dependency can be represented by a directed graph, where nodes represent subtasks and directed edges represent the data dependency directions. Different subtasks have different resource requirements. The data acquisition subtask may require underwater sensors with specific sampling frequencies, measurement accuracy, and endurance; the data processing subtask may have requirements for processor speed and memory capacity. These resource requirements are parameterized to form a resource requirement list for each subtask.

[0027] Based on the resource requirement list for each subtask, the system's resource management module is invoked to query resources. The resource management module stores information on all available underwater acoustic resources, including resource type, resource number, performance parameters, and current operating status. For example, when querying resources required for a data acquisition subtask, the resource management module will filter out sensors whose sampling frequency, measurement accuracy, and endurance parameters meet the requirements. The underwater acoustic resources here should include resource information at the same granularity as the subtask. For instance, if the subtask is a comprehensive data acquisition subtask with resource requirements of underwater sensor sampling frequency, measurement accuracy, and endurance, the corresponding granularity of resource information is found in the resource management module, where sensor resources are stored as overall equipment information, specifying parameters such as the overall sampling frequency, overall measurement accuracy, and equipment endurance for a particular sensor.

[0028] If a matching sensor is found to be currently occupied, the priority information of the task being performed by that occupied sensor needs to be further obtained. For example, if the currently queried sensor is performing a low-priority underwater equipment inspection task, while the data acquisition subtask to be scheduled belongs to a high-priority emergency environmental monitoring task, then the sensor is added to the available resource list of the data acquisition subtask. If the occupied sensor is performing a higher-priority task, it is not added to the available resource list of the current subtask. Through the above query and filtering process, a corresponding available resource list is generated for each subtask, containing information such as resource number, resource type, and key performance parameters.

[0029] Then, the physical connections of the corresponding resources in the available resource list for each subtask are read from the system's resource topology database. This database records the physical location distribution of all underwater acoustic resources and their interconnections in the form of a topology map. For example, sensor A is directly connected to processor B via optical fiber, and processor B is connected to data transmission node C via a wireless communication module. Simultaneously, the detailed performance parameters of these available resources are reconfirmed and read, such as the sensor's sampling delay, the processor's task processing response time, and the communication module's transmission bandwidth and bit error rate.

[0030] The data dependencies between subtasks, the list of available resources for each subtask, the physical connections between available resources, and the performance parameters of available resources are input into a pre-defined multi-objective optimization algorithm model. This algorithm model constructs an objective function with the optimization objectives of maximizing resource utilization and minimizing communication overhead.

[0031] During algorithm execution, the input data is first preprocessed, transforming data dependencies into mathematical constraints and resource performance parameters and physical connectivity into weight coefficients or variables that the algorithm can recognize. Then, through iterative calculations, the resource allocation scheme is continuously adjusted. For example, different sensors are allocated to the data acquisition subtask, and different processors are allocated to the data processing subtask. The task execution time, resource utilization, and communication overhead under different allocation schemes are calculated in conjunction with the physical connectivity.

[0032] When the algorithm iterates to a preset number of times or the calculation results meet a preset convergence condition, the iteration stops, and an optimal resource allocation scheme is output. This scheme is the underwater acoustic task resource scheduling strategy that meets the resource requirements. For example, the final allocation is determined as follows: data acquisition subtask is assigned to sensor A, data preprocessing subtask is assigned to processor B, data transmission subtask is assigned to communication module C, and data analysis subtask is assigned to processor D. The execution order and resource usage time periods of each subtask are also clearly defined.

[0033] This invention provides an intelligent scheduling method for underwater acoustic tasks based on multi-objective optimization. It integrates sub-task data dependencies, resource performance adaptability, and resource physical connectivity, generating a globally optimal scheduling strategy through a multi-objective optimization algorithm. This method reduces the probability of task execution stuttering due to insufficient consideration of data dependencies and, through precise resource selection and multi-objective optimization, is applicable to complex and diverse underwater acoustic task scenarios.

[0034] As an optional implementation, the data dependency relationship includes communication relationship data. Based on the data dependency relationship between each subtask, the available resource list of each subtask, the physical connection relationship of the available resources, and the performance parameters of the available resources, a multi-objective optimization algorithm is used to obtain a resource scheduling strategy that meets the resource requirements, including:

[0035] Based on communication relationship data, each subtask is divided into multiple task groups, and each task group includes at least one subtask.

[0036] Based on the data dependencies between various subtasks, the critical path is determined. The critical path represents the path with the longest total execution time among all paths from the start point to the end point in the data dependency relationship.

[0037] Based on available resource performance parameters and the critical path, resources are pre-allocated for each sub-task in the critical path to obtain a pre-allocation scheme.

[0038] Based on the physical connection relationship of optional resources, the pre-allocation scheme, the available resource list of each subtask, and task grouping, a multi-objective optimization algorithm is used to obtain a resource scheduling strategy that meets resource requirements.

[0039] For example, communication relationship data, including data exchange volume and communication frequency, is extracted from the data dependencies of each subtask. For instance, data acquisition subtask 1 and data acquisition subtask 2 need to frequently exchange environmental data of adjacent areas, with a communication frequency of 5 times per minute and a single data exchange volume of 10MB; data preprocessing subtask A and data analysis subtask B communicate with each other twice per hour, with a single data exchange volume of 50MB.

[0040] Based on the above communication relationship data, the communication density score between subtask i and subtask j is calculated using the communication density calculation formula. The calculation formula can be set as follows:

[0041]

[0042] in, , The weights can be 0.4 and 0.6 respectively. This is a standardized value for the amount of data exchanged between subtask i and subtask j. The communication frequency between subtask i and subtask j.

[0043] The total data volume of a subtask is the total amount of data generated or processed by that subtask throughout the entire task cycle, and the total task duration is the estimated execution time of the entire underwater acoustic task. After calculating the communication density score between each subtask using this formula, a preset score threshold (e.g., 0.6) is set, and subtasks with a communication density score greater than or equal to this threshold are grouped into task groups.

[0044] Based on the data dependencies between subtasks, a task dependency network graph is constructed. Nodes in the graph represent subtasks, directed edges represent the sequential dependencies between subtasks, and each directed edge is labeled with the estimated execution time of the corresponding subtask. For example, subtask A has an execution time of 2 hours; subtask B must be executed after subtask A is completed, with an execution time of 3 hours; subtask C must be executed after subtask A is completed, with an execution time of 1 hour; and subtask D must be executed after both subtasks B and C are completed, with an execution time of 2 hours.

[0045] The critical path method is used to calculate the total execution time of all possible paths from the starting subtask (e.g., subtask A) to the ending subtask (e.g., subtask D) in the task dependency network graph. In the example above, there are two paths: A→B→D, with a total execution time of 2+3+2=7 hours; and A→C→D, with a total execution time of 2+1+2=5 hours. The path with the longest total execution time (A→B→D) is the critical path of the underwater acoustic task. The total execution time of the critical path determines the shortest completion time of the entire underwater acoustic task. Therefore, resource scheduling should prioritize ensuring the resource needs of subtasks on the critical path.

[0046] Based on the performance parameters of available resources and the resource requirements of each subtask on the critical path, resources are pre-allocated for the subtasks on the critical path. Taking the data acquisition subtask on the critical path as an example, its resource requirements are a sampling frequency of 20Hz, a measurement accuracy of 0.05℃, and a battery life of 96 hours. From the list of available resources for this subtask, the sensor with the best performance and currently idle is selected for pre-allocation. Assuming that the list of available resources contains sensor X and sensor Y, with sensor X having a sampling frequency of 25Hz, a measurement accuracy of 0.03℃, and a battery life of 100 hours, and sensor Y having a sampling frequency of 20Hz, a measurement accuracy of 0.05℃, and a battery life of 96 hours, then sensor X is preferentially selected for pre-allocation to this data acquisition subtask.

[0047] The multi-objective optimization algorithm is run again, taking into account the physical connectivity of the available resources, the pre-allocation scheme, the available resource list for each subtask, and the task grouping data. During the algorithm's computation, the characteristics of task groups are fully considered, and subtasks within the same task group are preferentially assigned to physically connected resources to reduce communication latency and overhead between subtasks within the group.

[0048] For example, a task group may contain a data acquisition subtask 3 and a data preprocessing subtask 4. Both subtasks have intensive communication, and the sensors pre-assigned to data acquisition subtask 3 are physically directly connected to processor E. Therefore, when allocating resources for data preprocessing subtask 4, the algorithm will prioritize assigning it to processor E or a processor with a close physical connection to processor E and high communication bandwidth. Simultaneously, combined with the pre-allocation scheme, the algorithm ensures that the pre-allocated resources for critical path subtasks are not arbitrarily changed, and only optimizes the resource allocation for non-critical path subtasks. Through iterative calculation and optimization, a resource scheduling strategy for underwater acoustic tasks that satisfies resource requirements while considering multiple optimization objectives is ultimately obtained.

[0049] This invention provides an intelligent scheduling method for underwater acoustic tasks based on multi-objective optimization. Tasks are grouped based on communication relationship data, and subtasks with high-frequency data interaction are grouped together, significantly reducing communication latency within a group and improving data flow efficiency. By identifying the critical path (the path with the longest total execution time) and prioritizing resource pre-allocation, the core execution chain of the entire underwater acoustic task is ensured to remain uninterrupted, avoiding overall task delays due to insufficient resources in critical subtasks. Finally, multi-objective optimization is performed by combining task grouping and pre-allocation schemes, preserving resource guarantees for the critical path while also considering resource optimization for non-critical subtasks.

[0050] As an optional implementation, based on the physical connection relationships of optional resources, the pre-allocation scheme, the available resource list for each subtask, and task grouping, a multi-objective optimization algorithm is used to obtain a resource scheduling strategy that meets resource requirements, including:

[0051] S1. Initialize the population. The population contains multiple individuals, each representing a resource allocation scheme. The resource allocation scheme retains the pre-allocation scheme of each subtask in the critical path. Non-critical path subtasks randomly select resources from the available resource list of the corresponding subtask.

[0052] S2, based on the resource allocation scheme of each individual and the physical connection relationship of the optional resources, simulates the underwater acoustic task execution process and extracts target index data;

[0053] S3, based on the target index data and the pre-set fitness function, determines the fitness of each individual. The fitness function is constructed from communication overhead, resource utilization and task grouping achievement rate. The task grouping achievement rate represents the closeness between the resources allocated to tasks in the same task group in the allocation scheme.

[0054] S4, based on the fitness of each individual, select at least one primary target individual and multiple secondary target individuals;

[0055] S5, perform crossover mutation on multiple second target individuals to generate new offspring individuals, and take the first target individual and the new offspring individuals as the total offspring individuals. During the crossover mutation process, the pre-allocated resources bound to each subtask in the critical path are preserved.

[0056] Repeat steps S2-S5 until the termination condition is met, and obtain a resource scheduling strategy that satisfies the resource requirements.

[0057] For example, the population size is first determined based on the complexity of the underwater acoustic task and the number of subtasks. For instance, when there are 10 subtasks, the population size is set to 50, meaning the population contains 50 individuals, each representing a complete resource allocation scheme.

[0058] When constructing the resource allocation scheme for each individual, the pre-allocation schemes for each subtask in the critical path are strictly preserved; that is, the resource allocation of subtasks on the critical path remains fixed in the individual scheme. For non-critical path subtasks, resources are randomly selected from the available resource list of the corresponding subtask for allocation. For example, for the data backup subtask in the non-critical path, whose available resource list includes Backup Server 1, Backup Server 2, and Backup Server 3, one of the backup servers is randomly assigned to this subtask when constructing the individual scheme. Through the above method, the initialization of the entire population is completed, resulting in 50 individuals with different resource allocation schemes.

[0059] For each individual in the population, i.e., for each resource allocation scheme, and in conjunction with the physical connections of the available resources, a simulation environment for underwater acoustic task execution is built using task simulation software. Within this simulation environment, resources are allocated to each subtask according to the resource allocation scheme, and the entire execution process of the underwater acoustic task is simulated based on the data dependencies and physical connections between the subtasks.

[0060] During the simulation, target indicator data is recorded and extracted in real time, mainly including the following three types of indicators:

[0061] Communication overhead: Based on the communication bandwidth, transmission distance, and data exchange volume between subtasks in the physical connection relationship of resources, calculate the time cost and energy consumption cost of data transmission between each subtask, and sum them up to obtain the total communication overhead during the entire task execution process. For example, if subtask A and subtask B transmit 500MB of data through a communication link with a bandwidth of 100Mbps, the transmission time = 500 × 8 ÷ 100 = 40 seconds. Combined with the communication energy consumption per unit time, calculate the energy consumption cost of this transmission, and then sum up to obtain the total communication overhead.

[0062] Resource utilization rate: Calculate the actual working time and total idle time of each resource during the entire task execution cycle. For example, if processor C works for 120 minutes and is idle for 30 minutes during the task execution cycle, its resource utilization rate = (120 ÷ 150) × 100% = 80%. Summarize the utilization rates of all resources to calculate the overall resource utilization rate.

[0063] Task grouping compliance rate: The tightness between resources assigned to tasks within the same task group in the allocation scheme. Tightness can be evaluated from three dimensions: the first dimension is communication bandwidth, the maximum data transmission bandwidth between two resources; the higher the bandwidth, the tighter the connection. The second dimension is physical topology distance, the number of hops between two resources in the underwater hardware topology; the fewer the hops, the tighter the connection. The third dimension is communication latency, the average latency for transmitting a unit of data between two resources; the lower the latency, the tighter the connection. The task grouping compliance rate can also be determined by a weighted sum of these three factors.

[0064] A fitness function is constructed based on three types of target index data. The fitness function is constructed using a weighted summation method, as shown in the following formula:

[0065]

[0066] in, This represents the individual fitness value, which ranges from [0,1]. The larger the value, the better the resource allocation scheme for that individual; , , These are the weighting coefficients for communication overhead, resource utilization, and task grouping achievement rate, respectively, set according to actual task requirements. For example, in underwater acoustic tasks where real-time communication is critical, setting... =0.4, =0.3, =0.3;

[0067] The communication overhead for the current individual solution, This represents the maximum communication overhead across all individual schemes.

[0068] For the current individual plan's resource utilization rate, The maximum resource utilization rate among all individual solutions;

[0069] G represents the current individual's task grouping achievement rate. This represents the maximum achievement rate for task groups across all individual schemes.

[0070] Using the above formula, the fitness value of each individual in the population is calculated to obtain the fitness score of each individual, which provides a basis for subsequent individual selection.

[0071] All individuals in the population are sorted in descending order based on their fitness values. A selection ratio is set; for example, the top 10% of individuals with the highest fitness values ​​are selected as the first target individuals. That is, the top 5 individuals are selected from 50 individuals as the first target individuals. These individuals represent the better resource allocation scheme in the current population and are directly retained for the next generation of the population.

[0072] From the remaining 45 individuals, a roulette wheel selection algorithm is used to select multiple second target individuals. The number of individuals selected is determined based on the population size and the number of first target individuals; for example, 30 second target individuals may be selected. In the roulette wheel selection algorithm, the probability of each individual being selected is proportional to its fitness value; individuals with higher fitness values ​​have a greater probability of being selected. This method selects individuals with certain superior genes as second target individuals for subsequent crossover and mutation operations.

[0073] Crossover operations are performed on the 30 selected second-target individuals. A single-point crossover method is used, randomly selecting a crossover point. This crossover point corresponds to a resource allocation item in a non-critical path subtask. For example, the resource allocation item of the 5th subtask (a non-critical path subtask) is selected as the crossover point. The resource allocation schemes of the two second-target individuals in the non-critical path subtasks after this crossover point are swapped, generating two new offspring individuals. During the crossover process, the pre-allocated resources bound to each subtask in the critical path are strictly preserved; only the resource allocation schemes of the non-critical path subtasks are crossovered.

[0074] After the crossover operation is completed, a mutation operation is performed on the generated offspring individuals. A mutation probability is set (e.g., 0.05). For the resource allocation items of non-critical path subtasks in each offspring individual, a non-critical path subtask is randomly selected according to the mutation probability, and its currently allocated resources are replaced with other resources from the available resource list of that subtask. For example, if the data backup subtask in the offspring individual is currently allocated to Backup Server 1, it will be randomly replaced with Backup Server 2 or Backup Server 3 during mutation. Similarly, the pre-allocated resources of critical path subtasks are not changed during the mutation process.

[0075] The offspring individuals generated after crossover mutation are merged with the five previously selected primary target individuals to form a new population, completing one iteration process.

[0076] Repeat steps S2 to S5 to continuously update and optimize the population. The iteration termination condition can be set by the number of iterations reaching a preset number, such as 100, or by the change in the fitness value of the best individual in the population being less than a preset threshold, such as 0.001, during 10 consecutive iterations, indicating that the algorithm has converged.

[0077] When the iteration termination condition is met, the iteration stops, and the individual with the highest fitness value is selected from the final population. The resource allocation scheme corresponding to this individual is the underwater acoustic task resource scheduling strategy that meets the resource requirements.

[0078] This invention provides an intelligent scheduling method for underwater acoustic tasks based on multi-objective optimization. In the population initialization phase, the pre-allocation scheme for critical paths is retained, while resources are randomly allocated only to non-critical path subtasks. This ensures the stability of the core link and provides ample search space for subsequent optimization. By simulating task execution, three target indicators—communication overhead, resource utilization, and task grouping achievement rate—are extracted, and a fitness function is constructed based on these indicators. This makes the optimization objectives more aligned with actual needs, avoiding a disconnect between algorithm optimization and actual execution. During the selection and crossover mutation process, the binding of critical path resources is retained to ensure that the allocation of core resources is not disrupted. Simultaneously, iterative population optimization is achieved through the generation of offspring individuals, gradually selecting a better scheduling scheme.

[0079] As an optional implementation, the communication relationship data includes data exchange volume and communication frequency. Based on the communication relationship data, each subtask is divided into multiple task groups, including:

[0080] Based on communication relationship data, calculate the communication density score between subtasks;

[0081] Preliminary task grouping is determined based on the communication density score between subtasks and a preset score threshold.

[0082] If the number of subtasks in the initial task group exceeds the target number, analyze the overlap of resource requirements of the subtasks in the initial task group.

[0083] Under the premise of meeting the target number, task groups are constructed based on subtasks with low overlap in resource requirements.

[0084] For example, firstly, the specific values ​​of data exchange volume and communication frequency in the communication relationship data are defined. The system's task monitoring module collects data exchange records of each subtask in real time during its historical execution, or presets the data exchange volume and communication frequency between subtasks according to task requirements. For instance, subtask M and subtask N exchange data every 30 minutes during task execution, with each exchange involving 8MB of data; subtask P and subtask Q exchange data every 2 hours, with each exchange involving 20MB of data.

[0085] Next, determine the total data volume of each subtask and the total duration of the entire underwater acoustic task. The total data volume of a subtask is the total amount of data generated, received, and transmitted by that subtask during the entire task execution cycle, which can be determined through task requirements analysis and historical data statistics. Following the corresponding formulas described above, calculate the communication density score between all pairs of subtasks to obtain the complete subtask communication density matrix.

[0086] Based on the actual needs and resource allocation of the underwater acoustic mission, a communication density scoring threshold is set. The subtask communication density matrix is ​​traversed, and subtasks with communication density scores greater than or equal to the set threshold are grouped together to form preliminary task groups. Then, the number of subtasks in each preliminary task group is checked. If the number of subtasks in a preliminary group exceeds the target number (e.g., a preliminary group contains 5 subtasks (the target number is 3), the group needs to be split and adjusted. At this point, the resource overlap of each subtask within the preliminary group is analyzed. The resource overlap is determined by calculating the overlap ratio of the subtask resource requirement parameters.

[0087] Taking sensor resource requirements as an example, subtask 1 requires a sensor with a sampling frequency of 15Hz and a measurement accuracy of 0.1℃; subtask 2 requires a sensor with a sampling frequency of 15Hz and a measurement accuracy of 0.2℃; subtask 3 requires a sensor with a sampling frequency of 20Hz and a measurement accuracy of 0.1℃; subtask 4 requires a sensor with a sampling frequency of 18Hz and a measurement accuracy of 0.3℃; and subtask 5 requires a sensor with a sampling frequency of 15Hz and a measurement accuracy of 0.15℃. Calculating the resource overlap between subtasks, subtask 1 and subtask 2 overlap in sampling frequency (50%); subtask 1 and subtask 5 overlap in sampling frequency (50%); subtask 1 and subtask 3 only partially overlap in measurement accuracy (30%); and subtask 1 and subtask 4 have the lowest resource overlap, at only 10%.

[0088] While meeting the target number of tasks, prioritize grouping subtasks with minimal resource overlap into task groups. This adjustment ensures that each task group has the target number of subtasks, while also guaranteeing complementary resource needs among subtasks within the group, thus reducing resource competition.

[0089] This invention provides an intelligent scheduling method for underwater acoustic tasks based on multi-objective optimization. First, communication density is calculated based on data exchange volume and communication frequency to ensure that grouping logic aligns with the data interaction needs of subtasks. Second, when the initial number of subtasks in a group exceeds the target number, resource overlap (i.e., the proportion of overlapping resource requirements among subtasks) is analyzed to split the group, prioritizing subtasks with low resource overlap into new groups. This effectively avoids resource contention within the same group caused by conflicting resource requirements. For example, if two subtasks both require high-sampling-frequency sensors, splitting them into different groups avoids sensor resource contention and improves resource utilization efficiency. This grouping method ensures communication efficiency for subtasks within a group while reducing resource contention within the group.

[0090] As an optional implementation, based on resource requirements, resources that meet the requirements are queried, and a list of available resources for each subtask is generated, including:

[0091] If a resource that meets the resource requirements of any subtask is found to be in use, obtain the priority of the task that is in use.

[0092] If the priority of the occupied task is lower than that of the subtask, then the resource will be added to the available resource list of the corresponding subtask.

[0093] For example, when querying resources in the resource management module based on the resource requirements of a subtask, the current working status of each candidate resource is first obtained, with status categorized as either idle or occupied. For resources that are idle and meet the resource requirements, they are directly added to the available resource list for that subtask.

[0094] For resources that are in an occupied state but whose performance parameters meet the resource requirements of the subtask, the priority information of the currently executing tasks for that resource can be obtained through the task association query function of the resource management module. The system has a pre-established task priority classification system, which divides tasks into multiple levels according to their importance and urgency. For example, from high to low, they are divided into four levels: urgent critical tasks, important tasks, routine tasks, and low priority tasks. Each task is assigned a corresponding priority level when it is created and stored in the task information database.

[0095] Obtain the priority level of the subtask to be scheduled. Assume subtask A is an urgent and critical task (priority level 1). Compare the priority level of the subtask to be scheduled with the priority level of the currently executing task on the occupied resource:

[0096] If the priority of the subtask to be scheduled is higher than the priority of the currently executing task on the occupied resource, such as subtask A (level 1) having a higher priority than the regular task currently being executed by processor X (level 3), then the occupied resource will be added to the available resource list of the subtask to be scheduled (subtask A). Simultaneously, the resource's occupancy status and the estimated completion time of the currently executing task will be marked in the resource list, providing reference information for task preemption or resource handover during subsequent resource scheduling.

[0097] If the priority of the subtask to be scheduled is lower than or equal to the priority of the task currently being executed on the occupied resource, such as the subtask to be scheduled being a regular task (level 3) while the occupied resource is currently being executed as an important task (level 2), then the occupied resource will not be included in the list of available resources for the subtask to be scheduled. This is to avoid interruption of high-priority task execution due to resource preemption and to ensure the stability and reliability of overall task execution.

[0098] This invention provides an intelligent scheduling method for underwater acoustic tasks based on multi-objective optimization. When resources that meet the requirements of a subtask are occupied, the priority of the occupied task is compared with the priority of the current subtask. If the priority of the occupied task is lower, the resource is added to the available list to complete resource preemption and ensure that high-priority tasks are completed smoothly.

[0099] As an optional implementation, a multi-objective optimization-based intelligent scheduling method for underwater acoustic tasks further includes:

[0100] If the priority of the occupied task is higher than that of the subtask, the subtask is split into multiple subtasks.

[0101] Query the resources that meet the requirements of each sub-task to obtain a list of available resources for each sub-task;

[0102] Select an available resource from the available resource list corresponding to each subtask and add it to the available resource list of the corresponding subtask.

[0103] For example, when the priority of the subtask to be scheduled is lower than the priority of the currently executing task on the occupied resource, and there are no other idle resources or preemptible resources to meet the resource requirements of the subtask, the subtask splitting mechanism is activated. First, the task structure of the subtask to be scheduled is analyzed to clarify the core functional modules and execution flow of the subtask.

[0104] For example, the subtask to be scheduled is the underwater data comprehensive analysis subtask. Its core functions include four modules: data format conversion, data anomaly detection, data feature extraction, and data trend analysis. The execution flow is to complete the processing of these four modules sequentially. Based on the independence of the functional modules and the order of data processing, this subtask is divided into four subtasks: subtask 1 - data format conversion, subtask 2 - data anomaly detection, subtask 3 - data feature extraction, and subtask 4 - data trend analysis.

[0105] During the decomposition process, ensure that each sub-task has relatively independent functional and resource requirements, and that the execution order between sub-tasks is consistent with the functional module execution flow of the atomic task. At the same time, record the data dependencies between each sub-task, such as sub-task 2 needing to depend on the standardized data output by sub-task 1, and sub-task 3 needing to depend on the abnormal data output by sub-task 2.

[0106] For each sub-task after splitting, resource queries are performed again in the resource management module based on its independent resource requirements. It should be noted that the granularity of the resource information description here should be the same as the granularity of the sub-task. For example, a data preprocessing sub-task can be split into a data denoising sub-task, an outlier removal sub-task, and a format standardization sub-task. The corresponding resource information description granularity can be refined to processor speed, multi-core or single-core, and memory. Because the resource requirements of sub-tasks are more granular and may be lower than those of the original parent task, it is easier to find available resources that meet the requirements.

[0107] From the list of available resources for each secondary task, the optimal available resource is selected and allocated to the corresponding secondary task based on factors such as resource performance, current idle time, and physical connection relationship with other secondary task resources. For example, processor A, which has the longest idle time and a close physical connection with the candidate resources of secondary task 2, is selected for secondary task 1; processor B, which has the fastest data processing response speed, is selected for secondary task 2; processor C, which has the highest matrix operation efficiency, is selected for secondary task 3; and processor D, which has the most comprehensive statistical analysis functions, is selected for secondary task 4.

[0108] The resources selected for each sub-task are integrated to form an available resource combination scheme for the original sub-task to be scheduled (parent task), and this combination scheme is included in the available resource list of the parent task. The list details the resource number, resource type, key performance parameters, execution order and data dependencies of each sub-task, providing a complete resource selection scheme for subsequent resource scheduling of the parent task.

[0109] This invention provides an intelligent scheduling method for underwater acoustic tasks based on multi-objective optimization. When the priority of occupied resources is higher than that of the current subtask and there are no other suitable resources, the subtask is split into multiple secondary tasks (each secondary task has independent resource requirements). This reduces the resource requirement threshold of a single secondary task, makes it easier to find suitable idle resources, overcomes the limitation of insufficient resources, and improves the executability of underwater acoustic tasks.

[0110] This invention provides an intelligent scheduling device for underwater acoustic tasks based on multi-objective optimization, such as... Figure 2 As shown, it includes:

[0111] The task splitting module 201 is used to determine the data dependencies between the subtasks in the underwater acoustic task and the resource requirements of each subtask when an underwater acoustic task is received.

[0112] The resource determination module 202 is used to query resources that meet the resource requirements of each subtask and generate a list of available resources for each subtask.

[0113] The reading module 203 is used to read the physical connection relationship and performance parameters of the available resources based on the list of available resources for each subtask.

[0114] The scheduling strategy decision module 204 is used to obtain a resource scheduling strategy that meets resource requirements by adopting a multi-objective optimization algorithm based on the data dependencies between subtasks, the list of available resources for each subtask, the physical connection relationship of available resources, and the performance parameters of available resources.

[0115] This application also provides an electronic device, such as... Figure 3 As shown, processor 501 and memory 502 are connected via a bus or other means.

[0116] Processor 501 can be a central processing unit (CPU). Processor 501 can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, or combinations of the above types of chips.

[0117] The memory 502, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs, non-transitory computer-executable programs, and modules, such as the program instructions / modules corresponding to a multi-objective optimization-based intelligent scheduling method for underwater acoustic tasks in this embodiment of the invention. The processor executes various functional applications and data processing by running the non-transitory software programs, instructions, and modules stored in the memory.

[0118] Memory 502 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created by the processor, etc. Furthermore, the memory may include high-speed random access memory and non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory 502 may optionally include memory remotely located relative to the processor, which can be connected to the processor via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0119] The one or more modules are stored in the memory 502, and when executed by the processor 501, they perform actions such as... Figure 1 The embodiment shown presents an intelligent scheduling method for underwater acoustic tasks based on multi-objective optimization.

[0120] For specific details regarding the aforementioned electronic devices, please refer to the relevant documentation. Figure 1 The relevant descriptions and effects in the illustrated embodiments are for understanding purposes only and will not be repeated here.

[0121] This embodiment also provides a computer storage medium storing computer-executable instructions that can execute a multi-objective optimization-based intelligent scheduling method for underwater acoustic tasks in any of the above method embodiments. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk drive (HDD), or solid-state drive (SSD), etc.; the storage medium may also include combinations of the above types of memory.

[0122] Finally, it should be noted that the above preferred embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail through the above preferred embodiments, those skilled in the art should understand that various changes can be made to it in form and detail without departing from the scope defined by the claims of the present invention.

Claims

1. A method for intelligent scheduling of underwater acoustic tasks based on multi-objective optimization, characterized in that, include: When an underwater acoustic task is received, the data dependencies between the various subtasks in the underwater acoustic task and the resource requirements of each subtask are determined. The resource requirements of the subtasks are the sampling frequency, measurement accuracy, and endurance of the underwater sensor. Based on the resource requirements of each subtask, query the resources that meet the requirements and generate a list of available resources for each subtask. Based on the list of available resources for each subtask, read the physical connection relationship and performance parameters of the corresponding available resources; Based on the data dependencies between subtasks, the available resource list of each subtask, the physical connection relationship of the available resources, and the performance parameters of the available resources, a multi-objective optimization algorithm is adopted to obtain a resource scheduling strategy that meets the resource requirements. Data dependencies include communication relationship data. Based on the data dependencies between subtasks, the available resource lists for each subtask, the physical connections of available resources, and the performance parameters of available resources, a multi-objective optimization algorithm is used to obtain a resource scheduling strategy that meets resource requirements, including: Based on communication relationship data, each subtask is divided into multiple task groups, and each task group includes at least one subtask. Based on the data dependencies between various subtasks, the critical path is determined. The critical path represents the path with the longest total execution time among all paths from the start point to the end point in the data dependency relationship. Based on available resource performance parameters and the critical path, resources are pre-allocated for each sub-task in the critical path to obtain a pre-allocation scheme. Based on the physical connection relationship of optional resources, the pre-allocation scheme, the available resource list of each subtask, and task grouping, a multi-objective optimization algorithm is used to obtain a resource scheduling strategy that meets resource requirements. Based on the physical connection relationships of optional resources, the pre-allocation scheme, the available resource list for each subtask, and task grouping, a multi-objective optimization algorithm is used to obtain a resource scheduling strategy that meets resource requirements, including: S1. Initialize the population. The population contains multiple individuals, each representing a resource allocation scheme. The resource allocation scheme retains the pre-allocation scheme of each subtask in the critical path. Non-critical path subtasks randomly select resources from the available resource list of the corresponding subtask. S2, based on the resource allocation scheme of each individual and the physical connection relationship of the optional resources, simulates the underwater acoustic task execution process and extracts target index data; S3, based on the target index data and the pre-set fitness function, determines the fitness of each individual. The fitness function is constructed from communication overhead, resource utilization and task grouping achievement rate. The task grouping achievement rate represents the closeness between the resources allocated to tasks in the same task group in the allocation scheme. S4, based on the fitness of each individual, select at least one primary target individual and multiple secondary target individuals; S5, perform crossover mutation on multiple second target individuals to generate new offspring individuals, and take the first target individual and the new offspring individuals as the total offspring individuals. During the crossover mutation process, the pre-allocated resources bound to each subtask in the critical path are preserved. Repeat steps S2-S5 until the termination condition is met, and obtain a resource scheduling strategy that satisfies the resource requirements.

2. The intelligent scheduling method for underwater acoustic tasks based on multi-objective optimization according to claim 1, characterized in that, Communication relationship data includes data exchange volume and communication frequency. Based on this data, the subtasks are divided into multiple task groups, including: Based on communication relationship data, calculate the communication density score between subtasks; Preliminary task grouping is determined based on the communication density score between subtasks and a preset score threshold. If the number of subtasks in the initial task group exceeds the target number, analyze the overlap of resource requirements of the subtasks in the initial task group. Under the premise of meeting the target number, task groups are constructed based on subtasks with low overlap in resource requirements.

3. The intelligent scheduling method for underwater acoustic tasks based on multi-objective optimization according to claim 1, characterized in that, Based on resource requirements, query resources that meet those requirements and generate a list of available resources for each subtask, including: If a resource that meets the resource requirements of any subtask is found to be in use, obtain the priority of the task that is in use. If the priority of the occupied task is lower than that of the subtask, then the resource will be added to the available resource list of the corresponding subtask.

4. The intelligent scheduling method for underwater acoustic tasks based on multi-objective optimization according to claim 3, characterized in that, Also includes: If the priority of the occupied task is higher than that of the subtask, the subtask is split into multiple subtasks. Query the resources that meet the requirements of each sub-task to obtain a list of available resources for each sub-task; Select an available resource from the available resource list corresponding to each subtask and add it to the available resource list of the corresponding subtask.

5. The intelligent scheduling method for underwater acoustic tasks based on multi-objective optimization according to claim 1, characterized in that, The fitness function is: Where F is the individual fitness value, which ranges from [0,1]. The larger the F value, the better the resource allocation scheme corresponding to that individual. These are the weighting coefficients for communication overhead, resource utilization, and task grouping achievement rate, respectively. M represents the communication overhead of the current individual scheme. U represents the maximum communication overhead across all individual schemes; U is the resource utilization rate of the current individual scheme. The maximum resource utilization rate among all individual solutions; G represents the task grouping achievement rate of the current individual solution. This represents the maximum achievement rate for task groups across all individual schemes.

6. A smart scheduling device for underwater acoustic tasks based on multi-objective optimization, characterized in that, include: The task splitting module is used to determine the data dependencies between the subtasks in the underwater acoustic task and the resource requirements of each subtask when an underwater acoustic task is received. The resource requirements of the subtasks are the sampling frequency, measurement accuracy, and endurance of the underwater sensor. The resource determination module is used to query resources that meet the resource requirements of each subtask and generate a list of available resources for each subtask. The reading module is used to read the physical connection relationship and performance parameters of the available resources based on the list of available resources for each subtask. The scheduling strategy decision module is used to obtain a resource scheduling strategy that meets resource requirements based on the data dependencies between subtasks, the list of available resources for each subtask, the physical connection relationship of available resources, and the performance parameters of available resources, using a multi-objective optimization algorithm. Data dependencies include communication relationship data. Based on the data dependencies between subtasks, the available resource lists for each subtask, the physical connections of available resources, and the performance parameters of available resources, a multi-objective optimization algorithm is used to obtain a resource scheduling strategy that meets resource requirements, including: Based on communication relationship data, each subtask is divided into multiple task groups, and each task group includes at least one subtask. Based on the data dependencies between various subtasks, the critical path is determined. The critical path represents the path with the longest total execution time among all paths from the start point to the end point in the data dependency relationship. Based on available resource performance parameters and the critical path, resources are pre-allocated for each sub-task in the critical path to obtain a pre-allocation scheme. Based on the physical connection relationship of optional resources, the pre-allocation scheme, the available resource list of each subtask, and task grouping, a multi-objective optimization algorithm is used to obtain a resource scheduling strategy that meets resource requirements. Based on the physical connection relationships of optional resources, the pre-allocation scheme, the available resource list for each subtask, and task grouping, a multi-objective optimization algorithm is used to obtain a resource scheduling strategy that meets resource requirements, including: S1. Initialize the population. The population contains multiple individuals, each representing a resource allocation scheme. The resource allocation scheme retains the pre-allocation scheme of each subtask in the critical path. Non-critical path subtasks randomly select resources from the available resource list of the corresponding subtask. S2, based on the resource allocation scheme of each individual and the physical connection relationship of the optional resources, simulates the underwater acoustic task execution process and extracts target index data; S3, based on the target index data and the pre-set fitness function, determines the fitness of each individual. The fitness function is constructed from communication overhead, resource utilization and task grouping achievement rate. The task grouping achievement rate represents the closeness between the resources allocated to tasks in the same task group in the allocation scheme. S4, based on the fitness of each individual, select at least one primary target individual and multiple secondary target individuals; S5, perform crossover mutation on multiple second target individuals to generate new offspring individuals, and take the first target individual and the new offspring individuals as the total offspring individuals. During the crossover mutation process, the pre-allocated resources bound to each subtask in the critical path are preserved. Repeat steps S2-S5 until the termination condition is met, and obtain a resource scheduling strategy that satisfies the resource requirements.

7. An electronic device, the device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor performs the steps of the intelligent scheduling method for underwater acoustic tasks based on multi-objective optimization as described in any one of claims 1-5.

8. A computer storage medium storing computer instructions thereon, characterized in that, When executed by the processor, this instruction implements the steps of the intelligent scheduling method for underwater acoustic tasks based on multi-objective optimization as described in any one of claims 1-5.

Images