Heterogeneous satellite resource allocation method for maritime internet of things scene

Abstract

The application relates to a heterogeneous satellite resource allocation method for a maritime Internet of Things scene, which comprises the following steps: obtaining a target function with the least heterogeneous satellite resources and constraint conditions corresponding to the target function; the target function comprises a multi-dimensional scheduling matrix, which represents the use of resource blocks of heterogeneous satellites by an edge master station; the multi-dimensional scheduling matrix is subjected to optimization processing by using a genetic algorithm and a particle swarm algorithm respectively to obtain a candidate genetic scheduling matrix and a candidate particle scheduling matrix; the candidate genetic scheduling matrix and the candidate particle scheduling matrix are subjected to constraint processing according to the constraint conditions to obtain a constrained genetic scheduling matrix and a constrained particle scheduling matrix; a target scheduling matrix is determined from the constrained genetic scheduling matrix and the constrained particle scheduling matrix according to an evaluation function; and the target scheduling matrix is used to allocate resource blocks of corresponding satellites to the edge master station through the maritime Internet. The method can improve the utilization rate of satellite resources.

Description

Technical Field

[0001] This application relates to the field of marine satellite resource allocation technology, and in particular to a heterogeneous satellite resource allocation method for marine Internet of Things scenarios. Background Technology

[0002] As the maritime Internet of Things (IoT) evolves towards wider coverage, multi-service concurrency, and intelligent operation, satellite communication has become a crucial infrastructure supporting marine sensing, information backhaul, and remote collaboration. Especially in the open and deep sea areas, unmanned maritime platforms, communication buoys, and various edge master stations heavily rely on space-based satellite communication systems for data acquisition, transmission, and processing. Given the sparse distribution of services, the sudden nature of missions, and significant differences in service quality requirements in the marine environment, the rational allocation of limited satellite communication resources directly impacts the overall service capability and resource utilization efficiency of the maritime IoT system. Therefore, satellite resource allocation technology for maritime IoT scenarios has become one of the key technical challenges restricting system performance improvement.

[0003] Heterogeneous satellites refer to satellite systems that can exist in multiple types. In maritime satellite IoT, after the cloud completes the task planning for the edge master stations, it is necessary to uniformly plan and schedule the communication resources of heterogeneous satellites, so as to transmit the data generated by the edge master stations back to the remote cloud for centralized processing on demand. Traditional methods use heuristic algorithms to solve resource allocation problems; however, traditional methods are prone to getting trapped in local optima, resulting in low utilization of satellite resources. Summary of the Invention

[0004] Therefore, it is necessary to provide a heterogeneous satellite resource allocation method for maritime IoT scenarios to address the above-mentioned technical problems, which can avoid getting trapped in local optima and improve the utilization rate of satellite resources.

[0005] A method for allocating heterogeneous satellite resources for a maritime Internet of Things (IoT) scenario, the method comprising: Obtain the objective function that minimizes the use of heterogeneous satellite resources and the corresponding constraints of the objective function; the objective function includes a multi-dimensional scheduling matrix, which represents the use of resource blocks of heterogeneous satellites by the edge master station; A genetic algorithm is used to optimize the multidimensional scheduling matrix to obtain candidate genetic scheduling matrices; The candidate genetic scheduling matrix is ​​constrained according to the constraints to obtain the constrained genetic scheduling matrix. The multidimensional scheduling matrix is ​​optimized using a particle swarm optimization algorithm to obtain a candidate particle scheduling matrix; The candidate particle scheduling matrix is ​​constrained according to the constraints to obtain the constrained particle scheduling matrix. Based on the evaluation function, a target scheduling matrix is ​​determined from the constrained genetic scheduling matrix and the constrained particle scheduling matrix; the target scheduling matrix is ​​used to allocate resource blocks of corresponding satellites to the edge master station via the maritime Internet.

[0006] A computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements steps of a heterogeneous satellite resource allocation method for various maritime Internet of Things scenarios.

[0007] A computer program product includes a computer program that, when executed by a processor, implements steps for a heterogeneous satellite resource allocation method for various maritime Internet of Things (IoT) scenarios.

[0008] The aforementioned heterogeneous satellite resource allocation method for maritime IoT scenarios traditionally uses formulas as the objective form. However, the embodiments of this application employ a multidimensional matrix as the objective form, which is more suitable for processing multidimensional data related to heterogeneous satellite resource allocation and significantly improves the computational efficiency of the objective function. While genetic algorithms excel at global exploration and particle swarm optimization excels at local development, the continuous use of multiple optimization operators may lead to the final optimization process destroying the excellent individuals obtained by the previous optimization operator, resulting in a local optimum. However, by using genetic algorithms and particle swarm optimization in parallel to optimize the multidimensional scheduling matrix, the negative impact of excessive interaction can be avoided, preserving the optimal scheduling schemes obtained by each algorithm. The candidate scheduling matrices are constrained according to constraints to ensure that the obtained constrained scheduling matrices meet business requirements. Based on the evaluation function, the target scheduling matrix is ​​determined from the constrained scheduling matrix and the constrained particle scheduling matrix, ultimately satisfying business requirements with fewer resource blocks and improving the overall utilization rate of heterogeneous satellite network resources. Attached Figure Description

[0009] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0010] Figure 1 This is an application scenario diagram of a heterogeneous satellite resource allocation method for a maritime Internet of Things scenario in one embodiment; Figure 2 This is a flowchart illustrating a heterogeneous satellite resource allocation method for a maritime Internet of Things scenario in one embodiment. Figure 3 This is a schematic diagram of multidimensional scheduling matrix encoding in one embodiment; Figure 4 for Figure 3 The coding mapping diagram of the scheduling scheme in the middle; Figure 5 Here is a flowchart of MDMHA in one embodiment; Figure 6 This is a scenario diagram illustrating the resource scheduling problem of heterogeneous satellites and maritime satellite IoT in one embodiment. Figure 7 This is a scheme for associating the edge master station and the satellite at a certain moment in one embodiment; Figure 8 This illustrates how the number of satellite resource blocks occupied changes with the number of algorithm iterations in one embodiment. Figure 9 This is a schematic diagram illustrating the process of change in the number of satellite resource blocks in one embodiment; Figure 10 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0011] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0012] The heterogeneous satellite resource allocation method for maritime IoT scenarios provided in this application can be applied to, for example... Figure 1 The application environment shown. Figure 1 Figure 1 illustrates an application scenario of a heterogeneous satellite resource allocation method for a maritime IoT scenario. The scenario includes a maritime edge master station, heterogeneous satellite resources, an operations and maintenance control center, and a cloud (which includes gateway stations). The heterogeneous satellite communication network comprises multiple satellite types, such as GEO, MEO, and LEO. These satellites need to provide services to edge master stations with different resource requirements. While satellite multi-beam (multi-resource block) services can be utilized in the same time slot to serve multiple edge master stations, beam isolation constraints are needed to reduce interference and improve overall system capacity.

[0013] Considering the varying sensitivities of different edge master stations to latency, especially in tasks with high real-time requirements, it is essential to ensure that mission transmission latency remains within acceptable limits. Simultaneously, considering the capability constraints of each edge master station, insufficient transmission power may prevent the station from establishing a stable communication link with the satellite; therefore, it is necessary to ensure that power allocation meets the minimum access requirements. Throughout the operation of the heterogeneous satellite system, the status quo information of each edge master station and mission is perceived in real-time through edge nodes. Cloud nodes are responsible for extracting, cleaning, and evaluating the collected status quo information before reporting it to the system operation control center, thereby monitoring the operational status of the entire heterogeneous satellite network. When measuring the resources of the heterogeneous satellite system, based on the operational characteristics of satellite communication systems and adhering to the principles of systematicity, objectivity, and hierarchy, and using transmission rate, user latency, and beam interference as core factors, a multi-level communication resource measurement system for the heterogeneous satellite system is established to achieve resource measurement of the communication satellite system. The cloud can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services.

[0014] The embodiments of this application focus on the resource allocation optimization problem in heterogeneous satellite system scenarios for maritime internet. The goal is to design optimization schemes to maximize resource utilization efficiency and meet mission requirements, taking into account the characteristics of heterogeneous satellites and the diverse latency QoS (Quality of Service) requirements of terminals. In the embodiments of this application, the target scheduling matrix can be calculated by the cloud or the operation and maintenance control center and sent to the relevant satellite resource management unit and edge master station, respectively. The satellite side provides corresponding resource blocks in the corresponding time slots according to the scheduling scheme, and the edge master station accesses the designated satellite and performs data transmission in the designated time slot according to the received access plan. Therefore, the transmission path of the target scheduling matrix can be from the cloud, then to the operation and maintenance control center, then to the satellite and the edge master station; or it can be from the cloud to the satellite and then transmitted to the edge master station.

[0015] In one exemplary embodiment, such as Figure 2 The diagram shown is a flowchart illustrating a heterogeneous satellite resource allocation method for a maritime IoT scenario in one embodiment. This method is applied to… Figure 1 Taking the cloud as an example, the explanation includes the following steps 202 to 212. Wherein: Step 202: Obtain the objective function that minimizes the use of heterogeneous satellite resources and the corresponding constraints of the objective function; the objective function includes a multi-dimensional scheduling matrix, which represents the use of resource blocks of heterogeneous satellites by the edge master station.

[0016] Specifically, heterogeneous satellites refer to satellite systems that can contain multiple types of satellites. Specifically, it can be a hybrid constellation system composed of satellites of different types, orbits (such as GEO / MEO / LEO), payload capacities, or from different operators. Because of the differences in satellite orbital periods, coverage areas, data transmission rates, and computing capabilities, it is called "heterogeneous."

[0017] The objective function is a mathematical function designed to minimize the overall resource cost of the entire heterogeneous satellite system while meeting the operational needs of the edge master stations. The constraints are the rules that the objective function must satisfy.

[0018] Optionally, in a heterogeneous satellite communication system, multiple GEO, MEO, and LEO satellites are used to form a multi-layered satellite network. The total number of GEO satellites is [number missing]. The total number of MEO satellites is The total number of LEO satellites is Different satellites carry different numbers of beams; GEO satellites carry... Each MEO satellite carries a beam, and each MEO satellite has a... Each MEO satellite carries a beam, and each MEO satellite has a... One beam. The total number of edge master stations transmitting data to the satellite is used. This means that each edge master station can only be served by one satellite in the same time slot. Therefore, it can be used... This is a multidimensional matrix, representing satellite resource utilization and used as an optimization variable, representing the edge master station. In the time slot Using satellites The resource block information, with a value of 1 indicating a satellite. time slot Assigned to the edge master station The optimization objective is to minimize the total number of resource blocks used in the entire system.

[0019] like Figure 3 The diagram shown illustrates the encoding of a multidimensional scheduling matrix in one embodiment. Row vectors represent resource blocks for each satellite, and column vectors represent satellites (row and column vectors can be interchanged). The dimension of the row vectors is the total number of satellites, the dimension of the column vectors is the number of satellite resource blocks, and the dimension of the multidimensional scheduling matrix is ​​the number of edge master stations. As shown in Figure 3, the edge master stations are served by satellites 1, 2, and 3 at different times. After being served by satellite 3, the data transmission requirements are met, and the station satisfies its own latency constraints (total transmission volume constraints and latency constraints). Figure 4 for Figure 3 The coding mapping diagram of the scheduling scheme has a value of 1 when the satellite's time resource block is occupied by the edge master station, and 0 otherwise.

[0020] Step 204: Use a genetic algorithm to optimize the multidimensional scheduling matrix to obtain a candidate genetic scheduling matrix.

[0021] The candidate genetic scheduling matrix refers to a multi-dimensional scheduling matrix obtained by optimization using a genetic algorithm. The number of candidate genetic scheduling matrices must be at least two, and can be set according to requirements.

[0022] Specifically, the Genetic Algorithm (GA) is used to simulate natural evolution in search of optimal solutions. First, an initial population is generated, with each individual representing a multi-dimensional scheduling matrix and indicating possible solutions. The GA performs fitness evaluation, selection, crossover, and mutation on the initial population to generate multiple candidate genetic scheduling matrices. The GA excels at global exploration. The cloud-based GA periodically optimizes the multi-dimensional scheduling matrices to obtain candidate genetic scheduling matrices. This period can be determined based on the data upload cycle of the edge master station.

[0023] Step 206: Perform constraint processing on the candidate genetic scheduling matrix according to the constraint conditions to obtain the constrained genetic scheduling matrix.

[0024] Specifically, since the candidate genetic scheduling matrix may not meet the constraints, the cloud performs constraint processing on the matrix elements in the candidate genetic scheduling matrix according to the constraints of the objective function to obtain the constrained genetic scheduling matrix.

[0025] Step 208: The particle swarm optimization algorithm is used to optimize the multidimensional scheduling matrix to obtain the candidate particle scheduling matrix.

[0026] Specifically, Particle Swarm Optimization (PSO) is used to efficiently find optimal solutions in a search space, excelling at local optimization. In each iteration, the algorithm updates the individual optimal solutions and the global optimal solution. The cloud uses PSO to optimize the multidimensional scheduling matrix, obtaining a candidate particle scheduling matrix, which includes both individual and global optimal solutions. The cloud periodically uses PSO to optimize the multidimensional scheduling matrix again, obtaining the candidate particle scheduling matrix.

[0027] Step 210: Perform constraint processing on the candidate particle scheduling matrix according to the constraint conditions to obtain the constrained particle scheduling matrix.

[0028] Specifically, since the candidate particle scheduling matrix may not meet the constraints, the cloud performs constraint repair processing on the matrix elements in the candidate particle scheduling matrix according to the constraints of the objective function to obtain the constrained particle scheduling matrix.

[0029] Step 212: Determine the target scheduling matrix from the constrained genetic scheduling matrix and the constrained particle scheduling matrix according to the evaluation function; the target scheduling matrix is ​​used to allocate resource blocks of corresponding satellites to the edge master station through the maritime Internet.

[0030] The evaluation function includes evaluation terms corresponding to the objective function value. A smaller objective function value results in a better evaluation.

[0031] Specifically, the cloud platform calculates the evaluation function values ​​of each constrained genetic scheduling matrix and constrained particle scheduling matrix based on the evaluation function; the scheduling matrix with the best evaluation among the constrained genetic scheduling matrix and constrained particle scheduling matrix is ​​selected as the target scheduling matrix. The target scheduling matrix is ​​used to allocate resource blocks of corresponding satellites to the edge master station in a certain time slot through the maritime Internet, enabling the edge master station to use the corresponding satellite resource blocks at the corresponding time through the maritime Internet of Things.

[0032] In this embodiment, instead of using formulas as the objective form in traditional methods, a multidimensional matrix is ​​adopted, which is more suitable for processing multidimensional data on heterogeneous satellite resource allocation and greatly improves the computational efficiency of the objective function. Genetic algorithms excel at global exploration, while particle swarm optimization excels at local development. However, using multiple optimization operators consecutively may lead to the last optimization process destroying the excellent individuals obtained by the previous optimization operator, resulting in a local optimum. By using genetic algorithms and particle swarm optimization algorithms in parallel to optimize the multidimensional scheduling matrix, the negative impact of excessive interaction can be avoided, and the optimal scheduling schemes obtained by each algorithm can be preserved. The candidate scheduling matrix is ​​constrained according to the constraints to ensure that the obtained constrained scheduling matrix meets the business requirements. Based on the evaluation function, the target scheduling matrix is ​​determined from the constrained scheduling matrix and the constrained particle scheduling matrix, ultimately meeting the business requirements with fewer resource blocks and improving the overall utilization rate of heterogeneous satellite network resources.

[0033] In one embodiment, the constraints include one or more of the following: edge master station transmission resource constraints, latency constraints, total satellite resource constraints, power matching constraints, beam spatial isolation constraints, and single-timeslot single-satellite access constraints.

[0034] Specifically, the objective function for heterogeneous satellite resource scheduling is as follows: This is a multidimensional matrix, representing satellite resource utilization and used as an optimization variable, representing the edge master station. In the time slot Using satellites The status of resource blocks.

[0035] This refers to the resource constraints for edge master stations, meaning that all edge master stations must meet the resource requirements, i.e., the actual amount of data transmitted must be greater than or equal to the data volume requirement. This indicates that the edge master station uses satellites. The data transfer rate of a single resource block. Indicates the edge main station (Target data volume requirements). Constraints The physical meaning is to ensure that the total amount of data transmitted by the edge master station within the allowed service time window meets its business data requirements. It should be noted that in this embodiment... This is not the traditional "instantaneous transmission rate (bit / s)," but rather a unified abstraction of the data carrying capacity that a single time-frequency resource block can provide. Since each resource block implicitly contains a fixed-length time slot and spectrum resource, it essentially represents "the amount of data that a single resource block can transmit within a scheduling cycle." Therefore, constraints... middle: The summation can be understood as the cumulative data transmission capacity obtained by the edge master station across multiple time-frequency resource blocks, rather than a simple "direct summation of rates". Its unit essentially corresponds to the data volume, so there is no need to multiply by the time length. In other words, this model uses a discrete resource block scheduling modeling approach, rather than a continuous time rate allocation model; therefore, the cumulative resource block capacity is greater than or equal to the user's data demand.

[0036] The delay constraint indicates that the delay of the edge master station meets the delay requirements. Different edge master stations have different sensitivities to delay, especially in tasks with high real-time requirements. It is necessary to ensure that the transmission delay of the task is within an acceptable range. Therefore, in order to meet the service quality requirements of different terminals, it is necessary to constrain the delay QoS (Quality of Service) of the edge master stations. This model assumes that the data packets of each edge master station need to be processed within a certain timeframe. The latest time for the edge master station to complete the service is calculated by adding the data transmission start time to the latency QoS indicator (e.g., the transmission needs to be completed within 3 seconds). .

[0037] The constraint on the total amount of satellite resources means that the total amount of satellite resources used cannot exceed the available satellite resources. Since the resources of each satellite are limited, and considering that each beam of a satellite has the same capability and can only serve one edge master station within a time slot, the current model uses... Indicates satellite In the time slot The number of available beams.

[0038] The power matching constraint indicates that the master station's transmit power meets the satellite's minimum access power requirement. Different types of satellites have minimum access power requirements. If the master station's transmit power is insufficient, the edge master station cannot establish a stable communication link with the satellite. Therefore, the edge master station needs to select a suitable satellite for access. This model uses... Indicates satellite The minimum access power threshold, and through Indicates the edge main station The transmission power.

[0039] The beam spatial isolation constraint represents the spatial isolation constraint between beams of a single satellite. In multi-beam scenarios, interference can exist between beams of the same satellite. If two beams simultaneously serve different edge master stations, excessively small beam angles may cause interference. Therefore, beam isolation constraints are needed to reduce interference and ensure communication stability and reliability. This model uses... Indicates the edge main station and edge main site Simultaneously by satellite The beam interference relationship during service is set to 1, indicating the presence of interference. This can be determined based on the angle between the two users and the satellite beam isolation angle threshold.

[0040] This is a single-timeslot, single-satellite access constraint, used to restrict each user to access only one satellite in a single timeslot.

[0041] In this embodiment, the scheduling matrix is ​​constrained by conditions such as edge master station transmission resource constraints, latency constraints, total satellite resource constraints, power matching constraints, beam spatial isolation constraints, and single-time slot single-satellite access constraints, so that the obtained scheduling matrix meets the business requirements of the heterogeneous satellite resource allocation scenario of the maritime Internet of Things.

[0042] In one embodiment, the constraint includes a repair constraint; The candidate genetic scheduling matrix is ​​constrained according to the constraints to obtain the constrained genetic scheduling matrix, including: The matrix elements in the candidate genetic scheduling matrix are subjected to repair constraint processing according to the repair constraint conditions to obtain the repaired genetic scheduling matrix. The constrained genetic scheduling matrix is ​​determined based on the repaired genetic scheduling matrix.

[0043] Among them, the repair constraint is used to repair the matrix elements in the candidate genetic scheduling matrix so that the matrix satisfies the constraint.

[0044] Specifically, the cloud performs repair constraint processing on the matrix elements in the candidate genetic scheduling matrix according to the repair constraint conditions, that is, it modifies the matrix elements in the candidate genetic scheduling matrix so that the candidate genetic scheduling matrix meets the repair constraint conditions.

[0045] In this embodiment, since the candidate genetic scheduling matrix generated by the genetic algorithm may not meet the constraints, it is necessary to repair the matrix elements in the candidate genetic scheduling matrix according to the repair constraints, so that the obtained scheduling matrix meets the service requirements of heterogeneous satellite resource allocation.

[0046] In one embodiment, constraining the candidate particle scheduling matrix according to constraints to obtain a constrained particle scheduling matrix includes: performing constraint repair on the matrix elements in the candidate particle scheduling matrix according to the repaired constraints to obtain a repaired particle scheduling matrix; and determining the constrained particle scheduling matrix based on the repaired particle scheduling matrix.

[0047] In one embodiment, the matrix elements in the candidate genetic scheduling matrix are subjected to repair constraint processing according to repair constraints to obtain a repaired genetic scheduling matrix, including: Based on the power constraints and the transmit power of each edge master station, a genetic scheduling matrix that matches the transmit power between the edge master station and the satellite is selected from the candidate genetic scheduling matrices; In the genetic scheduling matrix with matching transmit power, matrix elements that exceed the latency requirements of the edge master station are set to not be occupied, matrix elements that exceed the total amount of satellite resources are set to not be occupied, and matrix elements where the number of satellites accessed by the edge master station in each time slot exceeds 1 are set to not be occupied, thus obtaining the repaired genetic scheduling matrix.

[0048] The power constraint includes a transmit power matching matrix, which represents the transmit power matching status between the satellite and the edge master station at different times. For example, the matrix element is 0 when the transmit power is mismatched and 1 when it is matched.

[0049] Since the requirements of each heterogeneous satellite are different, various constraints are needed to ensure that the edge master station can connect to the satellite. Specific constraints include power matching constraints, latency constraints, total satellite resource constraints, and single-timeslot single-satellite access constraints.

[0050] Specifically, if the master station's transmission power is insufficient, it may be unable to establish a stable communication link with the satellite. Therefore, it is necessary to ensure that the transmission power of the edge master stations meets the minimum satellite access requirements. Thus, based on the transmission power matching matrix and the transmission power of each edge master station, the cloud eliminates candidate genetic scheduling matrices from the candidate genetic scheduling matrices that do not meet the minimum satellite access requirements for the edge master stations, leaving only the power-matching genetic scheduling matrices where the transmission power of the edge master stations meets the minimum satellite access requirements.

[0051] The latency constraint is implemented by designing a transmission demand repair operator based on the latency requirements of the edge master station. Specifically, if the scheduling matrix contains elements exceeding the latency requirements of the edge master station, the corresponding element in the edge master station's encoding matrix is ​​set to unoccupied (e.g., 0) to ensure the latency constraint is met. For example, suppose the latest service completion time for edge master station u1 is the 3rd time slot. In a candidate genetic scheduling matrix, u1 is allocated satellite resource blocks in the 1st, 3rd, and 5th time slots, corresponding to the matrix value [1,0,1,0,1]. Since the 5th time slot exceeds u1's latency requirement, the repair operator sets the corresponding value of the 5th time slot to 0, resulting in [1,0,1,0,0]. Thus, the edge master station only retains resource allocations within the latency allowable range.

[0052] The total satellite resource constraint is a repair operator designed based on the satellite's available resource constraints. When the scheduling scheme exceeds the total satellite resources, the matrix elements of the edge master stations allocated beyond the total limit are set to 0 in the power-matching genetic scheduling matrix to ensure that the overall satellite resource constraints are met. For example, suppose satellite s1 can only serve a maximum of 2 edge master stations in a certain time slot, but the candidate scheduling matrix allocates that time slot to 3 edge master stations: u1=1, u2=1, u3=1. In this case, the allocation of 3 is greater than the satellite's available resources of 2. According to the repair method, the allocations that do not exceed the total resources are retained, and the allocations "beyond the total limit" are set to 0. For example, u1 and u2 are retained according to the encoding traversal order, and the corresponding value of u3 is set to 0; then u1=1, u2=1, u3=0. In this way, the number of edge master stations served by the satellite in that time slot no longer exceeds the total available resources.

[0053] The single-timeslot, single-satellite access constraint stipulates that an edge master station can only access one satellite per timeslot, and a matching repair operator is set accordingly. In the candidate genetic scheduling matrix, if the number of satellites accessed by the edge master station in each timeslot exceeds one, the matching scheme is changed to access the first satellite and not the others. The objective function value is calculated. Then, the matching scheme is changed again to access the second satellite and not the others, and the objective function value is calculated again. The scheduling schemes for accessing all satellites are iterated, and the optimal one is selected as the repaired scheduling scheme corresponding to the power matching genetic scheduling matrix. For example, if edge master station u1 accesses s1, s2, and s3 simultaneously in the same timeslot, three candidate repair schemes are formed: accessing only s1, accessing only s2, and accessing only s3. After calculating the objective function values ​​for each of the three methods, the scheme with the optimal objective function value is selected as the repair result corresponding to the power matching genetic scheduling matrix.

[0054] It is understandable that matrix elements exceeding the latency requirements of the edge master station are set to unoccupied, matrix elements exceeding the total amount of satellite resources are set to unoccupied, and matrix elements where the edge master station accesses more than 1 satellite at any given time are set to unavailable. The order of these three can be arbitrarily changed and is not limited here.

[0055] In this embodiment, based on power constraints and the transmit power of each edge master station, a genetic scheduling matrix with matching transmit power is selected from the candidate genetic scheduling matrices to ensure that each edge master station can connect to the satellite and to exclude scheduling matrices that cannot be connected. In the genetic scheduling matrix with matching transmit power, matrix elements that exceed the latency requirements of the edge master station are set to not be occupied, matrix elements that exceed the total amount of satellite resources are set to not be occupied, and matrix elements that allow the edge master station to access more than 1 satellite at any given time are set to not be occupied. This ensures that the edge master station can successfully connect to and transmit data successfully with heterogeneous satellites.

[0056] In one embodiment, the matrix elements in the candidate particle scheduling matrix are subjected to repair constraint processing according to repair constraints to obtain a repaired particle scheduling matrix, including: Based on the power constraints and the transmit power of each edge master station, a power matching particle scheduling matrix is ​​selected from the candidate particle scheduling matrix to match the transmit power between the edge master station and the satellite. In the power matching particle scheduling matrix, set the matrix elements that exceed the latency requirements of the edge master station to be unoccupied, set the matrix elements that exceed the total amount of satellite resources to be unoccupied, and set the matrix elements that allow the edge master station to access more than 1 satellite at each time to be unavailable, to obtain the repaired particle scheduling matrix.

[0057] In one embodiment, the constraints include penalty constraints, which include edge master station transmission resource constraints; determining the constraint genetic scheduling matrix based on the repaired genetic scheduling matrix includes: Obtain the target data volume for each edge master station; The amount of allocated data for each edge master station is determined based on the repaired genetic scheduling matrix. The penalty value of the repaired genetic scheduling matrix is ​​determined based on the difference between the target data volume and the allocated data volume; The constraint genetic scheduling matrix is ​​determined based on the penalty value.

[0058] The penalty constraint is used to calculate the penalty value, limiting the opportunity for the candidate genetic scheduling matrix to participate in subsequent iterative optimization. The target data volume of the edge master station refers to the amount of data the edge master station needs to transmit to the satellites. The allocated data volume refers to the total amount of satellite resource blocks corresponding to the edge master station in the repaired genetic scheduling matrix. For example, if edge master station u1 allocates resource block 1 from satellite 1 and resource block 1 from satellite 2, then the sum of the resources of these two resource blocks is the allocated data volume corresponding to the edge master station.

[0059] Specifically, the resource demand constraints of edge master stations are treated as soft constraints. The cloud obtains the target data volume for each edge master station and determines the allocated data volume for that edge master station based on the repaired genetic scheduling matrix. The cloud can determine a penalty value based on the difference between the actual value represented by the candidate genetic scheduling matrix and the set target value; the larger the difference, the larger the penalty value. Specifically, the cloud determines the penalty value of the repaired genetic scheduling matrix based on the absolute value of the difference between the target data volume and the allocated data volume; and the absolute value of this difference can be positively correlated with the penalty value.

[0060] The cloud can select N genes from the repaired genetic scheduling matrix as the constraint genetic scheduling matrix based on the penalty value. Alternatively, the cloud can arrange the repaired genetic scheduling matrix in ascending order of penalty value and select the first N (preset number) repaired genetic scheduling matrices to determine the constraint genetic scheduling matrix. Alternatively, the cloud can arrange the repaired genetic scheduling matrix in ascending order of penalty value without selection to obtain the constraint genetic scheduling matrix.

[0061] In this embodiment, when the transmission demand of a certain edge master station in the repaired genetic scheduling matrix is ​​not met, or the transmission resources are allocated excessively, the difference between the target data volume and the allocated data volume can be used as a penalty value. The constraint genetic scheduling matrix is ​​determined based on the penalty value, thereby reducing the chance of the repaired genetic scheduling matrix being used as the target scheduling matrix and eliminating scheduling schemes where data has not been fully transmitted.

[0062] In one embodiment, the constraint includes a penalty constraint, which includes a beam space isolation constraint. The penalty value of the repaired genetic scheduling matrix is ​​determined based on the penalty constraints, including: The beam angle between each edge master station is determined based on the edge master station corresponding to the satellite in the repaired genetic scheduling matrix; The penalty value of the repaired genetic scheduling matrix is ​​determined based on the beam angle and beam isolation angle threshold between each edge master station; The constraint genetic scheduling matrix is ​​determined based on the penalty value.

[0063] Among them, the beam isolation angle threshold refers to the minimum value of the beam angle stored in the cloud in a pre-set manner.

[0064] Specifically, different satellites may carry different numbers of beams. In multi-beam scenarios, interference can occur between beams of the same satellite. If two beams simultaneously serve different edge master stations, excessively small beam angles may cause interference. Therefore, beam isolation constraints are needed to reduce interference and ensure the stability and reliability of communication.

[0065] Based on the positions of the edge master stations corresponding to the same satellite in the repaired genetic scheduling matrix, the cloud can determine the beam angle between each edge master station. When the beam angle between two stations is less than the beam isolation angle threshold, it indicates that there may be interference in their communication, and the penalty value increases. Alternatively, by determining the number of edge master station pairs with beam angles less than the beam isolation angle threshold (i.e., the number of beam angles), the penalty value is positively correlated with this number. For example, suppose that in a certain time slot, satellite s1 simultaneously serves edge master stations u1, u2, and u3. Based on the beam angle and the beam isolation angle threshold, it is determined that: u1 and u2 do not meet the spatial isolation requirement; u2 and u3 also do not meet the spatial isolation requirement; u1 and u3 meet the spatial isolation requirement. Then the number of matching pairs that do not meet the spatial isolation constraint is 2, so the spatial isolation penalty value of this candidate scheme is positively correlated with 2. Therefore, the penalty value P_{beam} = 2; or: P_{beam} = \alpha × 2; where \alpha is the preset penalty coefficient. The larger the penalty value, the less likely the repaired genetic scheduling matrix will be retained in subsequent fitness selections.

[0066] The cloud can select N genes from the repaired genetic scheduling matrix as the constraint genetic scheduling matrix based on the penalty value. Alternatively, the cloud can arrange the repaired genetic scheduling matrix in ascending order of penalty value and select the first N (preset number) repaired genetic scheduling matrices to determine the constraint genetic scheduling matrix. Alternatively, the cloud can arrange the repaired genetic scheduling matrix in ascending order of penalty value without selection to obtain the constraint genetic scheduling matrix.

[0067] In this embodiment, the beam angle between each edge master station is determined based on the edge master station corresponding to the satellite in the repaired genetic scheduling matrix. The penalty value of the repaired genetic scheduling matrix is ​​determined based on the beam angle between each edge master station and the beam isolation angle threshold. The constraint genetic scheduling matrix is ​​determined based on the penalty value. The degree of interference can be judged by the beam angle, and the interference can be reduced by the beam isolation constraint, ensuring the stability and reliability of communication. Subsequently, unreliable communication schemes are eliminated as much as possible.

[0068] In one embodiment, constraining the candidate particle scheduling matrix according to constraints to obtain a constrained particle scheduling matrix includes: The candidate particle scheduling matrix is ​​then normalized, and then repaired and penalized according to the constraints to obtain the penalty value. The constraint particle scheduling matrix is ​​determined based on the penalty value.

[0069] Specifically, since the particle swarm algorithm may generate non-zero and non-1 values ​​during the update process, when the value in the candidate particle scheduling matrix is ​​greater than 1, it is set to 1, and when it is less than 0, it is set to 0. Values ​​between 0 and 1 are rounded to achieve normalization.

[0070] Similarly, the candidate particle scheduling matrix is ​​subjected to repair constraint processing and penalty constraint processing according to the constraint conditions to obtain the penalty value, including: repairing the matrix elements in the candidate particle scheduling matrix according to the repair constraint conditions to obtain the repaired particle scheduling matrix; and determining the penalty value of the repaired particle scheduling matrix according to the penalty constraint conditions.

[0071] In this embodiment, the candidate particle scheduling matrix is ​​normalized sequentially, and then repaired and penalized according to the constraints to obtain a constrained particle scheduling matrix, so that the candidate particle scheduling matrix can meet the requirements of heterogeneous satellite resource allocation.

[0072] In one embodiment, determining the target scheduling matrix from the constrained genetic scheduling matrix and the constrained particle scheduling matrix based on the evaluation function includes: In each iteration, the evaluation function values ​​of the constraint genetic scheduling matrix and the constraint particle scheduling matrix for the current round are determined according to the evaluation function. The scheduling matrix set for the current round is obtained by sorting and filtering the evaluation function values ​​of the constraint genetic scheduling matrix and constraint particle scheduling matrix of the current round, as well as the evaluation function values ​​of the constraint genetic scheduling matrix and constraint particle scheduling matrix of the previous round. When the iteration termination condition is not met, construct the multidimensional scheduling matrix for the next round based on the current round's scheduling matrix set, and return to execute the step of using a genetic algorithm to optimize the multidimensional scheduling matrix and obtain candidate genetic scheduling matrices; Until the iteration termination condition is met, the multidimensional scheduling matrix with the optimal evaluation function value is determined from the current round's scheduling matrix set, and the target scheduling matrix is ​​obtained.

[0073] Here, "current round" refers to the current iteration, and "previous rounds" refers to the iterations preceding the current round. "Previous rounds" can specifically refer to all previous rounds, several previous rounds, or the previous round, etc., without limitation. The iteration termination condition is used to end the iteration and can be set as needed, including but not limited to reaching a preset number of iterations or the evaluation function value reaching a preset value.

[0074] Specifically, the cloud platform determines the evaluation function values ​​of each constraint genetic scheduling matrix and each constraint particle scheduling matrix for the current round based on the evaluation function. The cloud platform then sorts and filters the constraint genetic scheduling matrices and constraint particle scheduling matrices for the current round, along with the evaluation function values ​​from previous rounds, selecting the scheduling matrices with better evaluations to obtain the scheduling matrix set for the current round.

[0075] When the iteration termination condition is not met, the cloud constructs a multi-dimensional scheduling matrix for the next round based on the current round's scheduling matrix set, which serves as input for the genetic algorithm and particle swarm optimization algorithm. The genetic algorithm optimizes the multi-dimensional scheduling matrix, followed by constraint processing to obtain a constrained genetic scheduling matrix; the particle swarm optimization algorithm optimizes the multi-dimensional scheduling matrix, followed by constraint processing to obtain a constrained particle scheduling matrix. This process continues until the iteration termination condition is met, at which point the target scheduling matrix, with the optimal evaluation function value, is determined from the current round's scheduling matrix.

[0076] In this embodiment, in each iteration, the evaluation function values ​​of the constrained genetic scheduling matrix and the constrained particle scheduling matrix, as well as the evaluation function values ​​of the previous rounds, are sorted and filtered. The two algorithms are optimized in parallel to avoid the latter optimization process destroying the optimization results of the former and producing excessive interaction results. The one with the better evaluation is selected as the multidimensional scheduling matrix for the next round, retaining the better scheduling schemes obtained by the two algorithms. In the end, the user's business needs are met with fewer resource blocks, and the overall resource utilization of the heterogeneous satellite network is improved.

[0077] In one embodiment, the evaluation function includes a diversity index and an adaptability index; the diversity index is determined based on the difference between each constrained scheduling matrix and the optimal scheduling matrix; the adaptability index is determined based on the normalized objective function value of each constrained scheduling matrix.

[0078] The evaluation function includes a diversity index and a fitness index. The diversity index represents the diversity of the population, while the fitness index represents the final objective function value. The difference between the constraint scheduling matrix and the optimal scheduling matrix can be calculated using methods such as normalized Hamming distance or Euclidean distance.

[0079] Specifically, since the encoding method is discrete vectors, for genetic algorithms, the diversity index is the difference between the constrained genetic scheduling matrix and the optimal constrained genetic matrix for the current round; the fitness index is the normalized objective function value of each constrained genetic scheduling matrix. For particle swarm optimization, the diversity index is the difference between the constrained particle scheduling matrix of the swarm optimization theory and the optimal constrained particle scheduling matrix for the current round; the fitness index is the normalized objective function value of each constrained particle scheduling matrix.

[0080] A higher diversity index value indicates greater diversity, while a lower adaptability index value indicates a smaller objective function value. Therefore, the formula can be: Evaluation Function = Adaptability Index - Diversity Index, with a smaller evaluation function value indicating a better evaluation. Alternatively, Evaluation Function = Diversity Index - Adaptability Index, with a larger evaluation function value indicating a better evaluation.

[0081] Optionally, both diversity and fitness indicators can be multiplied by weights. The fewer the iterations, the higher the weight of individual diversity; the more iterations, the higher the weight of fitness indicators. Assuming 100 iterations, in the early stages, such as iterations 1-50, a comprehensive evaluation function with a greater weight for diversity indicators is used; in iterations 51-100, a comprehensive evaluation function with a greater weight for fitness indicators is used.

[0082] In this embodiment, the traditional approach only uses the adaptability index as the result selection criterion, ignoring the fact that diversity leads to premature convergence and degradation of the results. By combining the diversity index with the adaptability index, the matrix population can be made more diverse, and the obtained target scheduling matrix can meet the user's business needs with fewer resource blocks, thereby improving the overall utilization rate of heterogeneous satellite network resources.

[0083] In one embodiment, the weights of the diversity and fitness metrics are determined based on the number of iterations of the algorithm; the number of iterations is negatively correlated with the weight of the diversity metrics; and the number of iterations is positively correlated with the weight of the fitness metrics.

[0084] Specifically, the weights of the diversity and fitness metrics are determined based on the number of iterations in the algorithm. The number of iterations is negatively correlated with the weight of the diversity metric, and positively correlated with the weight of the fitness metric. Therefore, in the early stages of iteration, fewer iterations result in a higher weight for the diversity metric, emphasizing individual diversity and improving global search capabilities; while in the later stages of iteration, more iterations result in a higher weight for the fitness metric, emphasizing individual fitness and improving local convergence capabilities.

[0085] The specific evaluation indicators are as follows: in For the constraint scheduling matrix The final evaluation function is selected during the selection process. and These are the normalized diversity index and the normalized fitness index, respectively. and are the weighting coefficients for diversity and adaptability indicators, respectively. and These represent the current iteration number and the final iteration number, respectively.

[0086] In this embodiment, the weights of diversity and adaptability indicators are determined based on the number of iterations of the algorithm. The number of iterations is negatively correlated with the weight of diversity indicators and positively correlated with the weight of adaptability indicators. This allows the algorithm to select individuals with lower evaluation indicators but higher individual diversity in the early stages of the algorithm to participate in iterative optimization, and to focus on individuals with higher evaluation indicators in the later stages of the algorithm. Ultimately, the algorithm can meet the user's business needs with fewer resource blocks and improve the overall utilization rate of heterogeneous satellite network resources.

[0087] In one embodiment, a method for allocating heterogeneous satellite resources for a maritime Internet of Things (IoT) scenario includes: Step (a1) obtains the objective function that minimizes the use of heterogeneous satellite resources and the corresponding constraints of the objective function; the objective function includes a multi-dimensional scheduling matrix, which represents the use of resource blocks of heterogeneous satellites by the edge master station.

[0088] Step (a2) uses a genetic algorithm to optimize the multidimensional scheduling matrix and obtain a candidate genetic scheduling matrix.

[0089] Step (a3): Based on the power constraints and the transmit power of each edge master station, select the genetic scheduling matrix that matches the transmit power between the edge master station and the satellite from the candidate genetic scheduling matrix.

[0090] Step (a4): In the genetic scheduling matrix with matching transmit power, matrix elements that exceed the latency requirements of the edge master station are set to not be occupied, matrix elements that exceed the total amount of satellite resources are set to not be occupied, and matrix elements that allow the edge master station to access more than 1 satellite in each time slot are set to not be occupied, thus obtaining the repaired genetic scheduling matrix.

[0091] Step (a5): Obtain the target data volume for each edge master station.

[0092] Step (a6): Determine the amount of allocated data for each edge master station based on the repaired genetic scheduling matrix.

[0093] Step (a7) determines the penalty value of the repaired genetic scheduling matrix based on the difference between the target data volume and the allocated data volume.

[0094] Step (a8): Determine the beam angle between each edge master station based on the edge master station corresponding to the satellite in the repaired genetic scheduling matrix.

[0095] Step (a9): Determine the penalty value of the repaired genetic scheduling matrix based on the beam angle between each edge master station and the beam isolation angle threshold.

[0096] Step (a10): Determine the constraint genetic scheduling matrix based on the penalty value.

[0097] Step (a11) uses the particle swarm optimization algorithm to optimize the multidimensional scheduling matrix and obtain the candidate particle scheduling matrix.

[0098] Step (a12) involves performing normalization on the candidate particle scheduling matrix, and then performing repair constraint processing and penalty constraint processing on the candidate particle scheduling matrix according to the constraint conditions to obtain the constrained particle scheduling matrix.

[0099] In step (a13), in each iteration, the evaluation function values ​​of the constraint genetic scheduling matrix and the constraint particle scheduling matrix for the current round are determined according to the evaluation function. The evaluation function includes a diversity index and a fitness index. The diversity index is determined based on the difference between each constraint scheduling matrix and the optimal scheduling matrix. The fitness index is determined based on the normalized objective function value of each constraint scheduling matrix. The weights of the diversity index and the fitness index are determined based on the number of iterations of the algorithm. The number of iterations is negatively correlated with the weight of the diversity index. The number of iterations is positively correlated with the weight of the fitness index.

[0100] Step (a14): Sort and filter the constraint genetic scheduling matrix and constraint particle scheduling matrix of the current round according to the evaluation function values ​​of the constraint genetic scheduling matrix and constraint particle scheduling matrix of the previous round, and obtain the scheduling matrix set of the current round.

[0101] Step (a15): When the iteration termination condition is not met, construct the multidimensional scheduling matrix for the next round based on the current round's scheduling matrix set, and return to execute the step of using a genetic algorithm to optimize the multidimensional scheduling matrix and obtain a candidate genetic scheduling matrix.

[0102] Step (a16) continues until the iteration termination condition is met. Then, the multidimensional scheduling matrix with the optimal evaluation function value is determined from the current round's scheduling matrix set, thus obtaining the target scheduling matrix. The target scheduling matrix is ​​used to allocate resource blocks of corresponding satellites to edge master stations via the maritime internet.

[0103] In this embodiment, instead of using formulas as the objective form in traditional methods, a multidimensional matrix is ​​adopted, which is more suitable for processing multidimensional data on heterogeneous satellite resource allocation and greatly improves the computational efficiency of the objective function. Genetic algorithms excel at global exploration, while particle swarm optimization excels at local development. However, using multiple optimization operators consecutively may lead to the last optimization process destroying the excellent individuals obtained by the previous optimization operator, resulting in a local optimum. By using genetic algorithms and particle swarm optimization algorithms in parallel to optimize the multidimensional scheduling matrix, the negative impact of excessive interaction can be avoided, preserving the optimal scheduling schemes obtained by each algorithm. The candidate scheduling matrix is ​​constrained according to the constraints to ensure that the obtained constrained scheduling matrix meets the business requirements. Based on the evaluation function, the target scheduling matrix is ​​determined from the constrained scheduling matrix and the constrained particle scheduling matrix, ultimately satisfying the user's business requirements with fewer resource blocks and improving the overall utilization rate of heterogeneous satellite network resources.

[0104] In one embodiment, this embodiment focuses on the resource allocation optimization problem in a heterogeneous satellite system scenario. The goal is to design an optimization scheme to maximize resource utilization efficiency and meet mission requirements, taking into account the characteristics of different types of satellites in high, medium and low orbits and the diverse latency and QoS requirements of terminals.

[0105] I. Heterogeneous Satellite Resource Allocation Model In heterogeneous satellite communication systems, multiple GEO, MEO, and LEO satellites form a multi-layered satellite network. The total number of GEO satellites is [number missing]. The total number of MEO satellites is The total number of LEO satellites is Different satellites carry different numbers of beams; GEO satellites carry... Each MEO satellite carries a beam, and each MEO satellite has a... Each MEO satellite carries a beam, and each MEO satellite has a... One beam. The total number of edge master stations transmitting data to the satellite is used. This means that each edge master station can only be served by one satellite in the same time slot. Therefore, it can be used... This represents satellite resource usage and is used as an optimization variable, representing the edge master station. In the time slot Using satellites The resource block information, with a value of 1 indicating a satellite. time slot Assigned to the edge master station The optimization objective is to minimize the total number of resource blocks used in the entire system.

[0106] In this embodiment, a multi-dimensional scheduling matrix is ​​used to uniformly encode the resource scheduling scheme. Here, the row dimension represents satellites, the column dimension represents time resource blocks, and the third dimension of the multi-dimensional scheduling matrix represents edge master stations. Elements in the matrix take values ​​of 0 or 1: 1 indicates that the corresponding resource block is occupied by the edge master station; 0 indicates that it is not occupied.

[0107] Therefore, a complete multidimensional scheduling matrix actually corresponds to a complete resource scheduling scheme.

[0108] Therefore, the objective function for heterogeneous satellite resource scheduling is as follows: Wherein, the overall objective function This means minimizing the total number of resource blocks used by a user across all time slots. Six constraints are set: This refers to the resource constraints for edge master stations, meaning that all edge master stations must meet the resource requirements, i.e., the actual amount of data transmitted must be greater than or equal to the data volume requirement. This indicates that the edge master station uses satellites. The data transfer rate of a single resource block. Indicates the edge main station (Target data volume requirements). The delay constraint condition indicates that the delay constraint of the edge master station meets the delay requirements of the edge master station. This is a constraint on the total amount of satellite resources, meaning that the total amount of satellite resources used cannot exceed the available satellite resources. The power matching constraint means that the master station's transmit power meets the satellite's minimum access power requirement. The beam spatial isolation constraint condition represents the spatial isolation constraint between beams of a single satellite. This is a single-timeslot, single-satellite access constraint, used to restrict each user to access only one satellite in a single timeslot.

[0109] II. Heuristic Algorithm Based on Multidimensional Scheduling Matrix Encoding To address the heterogeneous satellite collaborative resource scheduling problem mentioned above, this paper proposes a heuristic algorithm based on multidimensional scheduling matrix encoding. Figure 5This is a flowchart of MDMHA in one embodiment. Specifically, a multidimensional vector encoding method is used to represent the resource scheduling scheme using a multidimensional scheduling matrix. In this encoded vector, the 0-1 vector represents the association between a certain edge master station and a satellite resource block, and the matrix dimension is the total number of edge master stations. Based on the multidimensional vector method, MDMHA (Multiple Dimensional Matrix Heuristic Algorithm) can simultaneously use continuous and discrete optimization operators to update the collaborative resource scheduling scheme of multiple subgroups. After initialization, the cloud generates subgroup 1 and subgroup 2; these are input into the genetic algorithm and particle swarm optimization algorithm, respectively. Subgroup 1 is processed by the crossover and mutation operators of the genetic algorithm, and subgroup 2 is processed by the particle swarm optimization algorithm to generate individual particles and the optimal particle, thus generating a new population. If the termination condition is not met, the selection of subgroups continues. After the termination condition is met, MDMHA can generate an effective resource scheduling scheme. The solution process of MDMHA is based on a heuristic algorithm framework. Figure 3 This demonstrates the process of encoding a multidimensional scheduling matrix. Essentially, it involves encoding resource scheduling schemes into a multidimensional scheduling matrix, and then using both a genetic algorithm (GA) and a particle swarm optimization (PSO) algorithm to perform parallel iterative optimization of these schemes.

[0110] The multidimensional scheduling matrix is ​​not a new module independent of GA / PSO, but rather a "unified encoding method" used by both GA and PSO. In other words, whether in the GA population or the PSO particle swarm, each individual essentially corresponds to a resource scheduling scheme in the form of a multidimensional scheduling matrix.

[0111] 1. Initialize the GA population and PSO particle swarm optimization During the initialization phase: Each "chromosome individual" in GA is a multidimensional scheduling matrix, and each "particle" in PSO is also a multidimensional scheduling matrix.

[0112] Therefore, although the two algorithms have different optimization mechanisms, they actually optimize the same thing: a "resource scheduling scheme in the form of a multi-dimensional scheduling matrix".

[0113] 2. Parallel execution of iterative optimization using GA and PSO Subsequently, the GA subpopulation and the PSO subpopulation were optimized independently.

[0114] (1) GA subpopulation: First, calculate the individual fitness based on the objective function; then perform the selection operation; and finally generate a new multidimensional scheduling matrix of individuals through the crossover and mutation operators.

[0115] (2) PSO subpopulation: Each particle also corresponds to a multidimensional scheduling matrix; the particle state is updated based on the current particle, the individual's historical best position, and the global best position; after the update, a new multidimensional scheduling matrix scheduling scheme is generated.

[0116] 3. Execution constraint repair and penalty function calculation Since GA crossover, mutation, and PSO particle updates may generate scheduling schemes that do not meet the constraints, this paper further designs a constraint repair operator and a penalty function mechanism.

[0117] The use of the six constraints in the algorithm can be summarized into three categories: (1) The first type is the transmit power matching matrix.

[0118] The master station transmit power constraint is not simply used as a penalty, but rather during population initialization and iterative updates. That is, a transmit power matching matrix between each satellite and the edge master station is first constructed at different times; during initialization and iterative updates, only available satellites are selected based on this matching matrix, thereby ensuring a stable connection between the edge master station and the satellites.

[0119] (2) The second type is the repair operator.

[0120] Edge master station latency constraints, satellite available resource constraints, and single-timeslot single-satellite access constraints are mainly addressed through repair operators. That is, if there are positions in the candidate scheduling matrix that do not meet these constraints, the corresponding 0-1 values ​​in the matrix are modified to bring them back to a state that meets the constraints as much as possible.

[0121] (3) The third type is the soft constraint penalty function.

[0122] Edge master station transmission resource demand constraints are treated as soft constraints. When the transmission demand of an edge master station in the scheduling scheme is not met, or when transmission resources are over-allocated, the absolute value of the difference between the transmission demand and the total allocated transmission capacity is used as a penalty function. The larger the difference, the larger the penalty value, and the lower the chance that the scheduling scheme will continue to participate in iterative optimization. Satellite beam spatial isolation constraints are also treated as soft constraints. The number of matching pairs that do not meet the spatial isolation constraints is calculated, and the penalty function is positively correlated with this number.

[0123] Therefore, the constraints apply not only to the initialization and iterative update process, but also to the repair and evaluation process after the candidate scheduling scheme is generated, and are not simply superimposed on the objective function.

[0124] Therefore, both genetic algorithms and particle swarm optimization algorithms use multi-dimensional matrices as input. Taking the genetic algorithm as an example, the initial multi-dimensional scheduling matrix is ​​used to generate new candidate scheduling matrices through multi-point crossover and single-point mutation, and then constraints are applied to the candidate scheduling matrices: ① Based on the transmit power matching matrix, check whether an unaccessible satellite has been selected; ②Based on the latency constraints, the encoding matrix values ​​that exceed the latency requirements of the edge master station are set to 0; ③ Based on the constraints of available satellite resources, the allocation value exceeding the total amount of satellite resources is set to 0; ④ Based on the single-time-slot single-satellite constraint, perform a comprehensive repair of cases where the same edge master station accesses multiple satellites in the same time slot; ④ Calculate the penalty value based on transmission resource requirements and beam spatial isolation.

[0125] After the above processing, the candidate matrix after repair and evaluation is obtained.

[0126] For the particle swarm optimization (PSO) algorithm, each particle in the swarm is also a multidimensional encoding matrix. In the previous example, the scheme was updated using conventional local and global optimization operators to obtain a new particle matrix. Since the PSO algorithm may generate non-zero values ​​during the update process, values ​​in the encoding matrix are set to 1 when they are greater than 1, and set to 0 when they are less than 0. Values ​​between 0 and 1 are rounded.

[0127] Therefore, the processing flow of the particle swarm optimization algorithm can be described as follows: after performing local and global optimization updates on the particles from the previous round, a 0-1 transformation is performed to obtain the candidate particle scheduling matrix.

[0128] Subsequently, the same constraint processing is applied to the candidate particle scheduling matrix to obtain the constrained particle scheduling matrix: ① Select available satellites based on the power matching matrix; ② Set the matrix values ​​that exceed the latency requirements to 0; ③ Set the matrix value to 0 if it exceeds the total available satellite resources; ④ Match and repair the situation where multiple satellites are accessed at the same edge master station in the same time slot; ⑤ Calculate penalties for transmission demand deviations and beam isolation conflicts.

[0129] 4. Implement an improved fitness evaluation mechanism The "fitness + diversity" comprehensive evaluation mechanism proposed in this embodiment is not used only after GA / PSO has completely ended, but is continuously involved in individual selection during each iteration.

[0130] Specifically: Traditional GA / PSO typically evaluates individuals solely based on the objective function value; this embodiment further introduces an "individual diversity index" and constructs a comprehensive evaluation function through dynamic weighting.

[0131] The difference between each individual and the best individual in the population is used as the diversity evaluation index for the individual. Since the encoding method is a discrete vector, the normalized Hamming distance between the multidimensional matrix encoding of the individual and the best individual in the current population is used as the diversity evaluation index. The normalized objective function value is used as the fitness index of the individual, and the weighted sum of the two is used as the final evaluation index for the individual in the execution selection process.

[0132] in For individuals The final evaluation function is selected during the selection process. and These are the normalized diversity function and the normalized fitness function, respectively. and Let be the weight coefficients for the diversity function and the fitness function, respectively. and These represent the current iteration number and the final iteration number, respectively.

[0133] Therefore, in the early stages of iteration Larger, with greater emphasis on individual diversity; later iterations The algorithm is larger and places greater emphasis on individual fitness. In the early stages of the algorithm, individuals with lower evaluation metrics but higher individual diversity are selected for iterative optimization, focusing more on population diversity and improving global search capabilities. In the later stages of the algorithm, the algorithm focuses on individuals with higher evaluation metrics, paying more attention to individual fitness and improving local convergence capabilities.

[0134] In summary, the comprehensive evaluation function is used after the candidate scheduling matrix has completed constraint repair and penalty calculation. It is used to select individuals to be retained from newly generated individuals and the original individuals to construct the next round of the new population. In the early stages of the algorithm, the diversity weight is higher, giving individuals with greater differences a chance to be retained; in the later stages of the algorithm, the fitness weight is higher, making it easier to retain individuals that occupy fewer resource blocks and have smaller penalties.

[0135] 5. Constructing new populations In each iteration, the genetic algorithm subpopulation generates a batch of constrained genetic scheduling matrices, and the particle swarm optimization subpopulation generates a batch of constrained particle genetic scheduling matrices. These new candidate individuals, along with the original individuals, are then sorted and selected according to a comprehensive evaluation function to construct a new population of the same size as the initial population. If the termination condition is not met, the new population continues into the next iteration; if the termination condition is met, the optimal multidimensional matrix from the current new population is selected as the final resource scheduling scheme.

[0136] Therefore, the new population is not derived solely from either GA or PSO algorithms, but rather through collaborative optimization and selection by both sub-populations. The algorithm terminates when the maximum number of iterations is reached or the convergence condition is met. The final output is: "Optimal satellite resource scheduling scheme in the form of a multidimensional scheduling matrix".

[0137] In summary, the core improvements of MDMHA in this embodiment are mainly reflected in: (1) A unified encoding method for multi-dimensional scheduling matrices for discrete resource scheduling problems is proposed; (2) Design a parallel collaborative optimization mechanism for GA and PSO populations; (3) A dynamic evaluation mechanism jointly driven by fitness and population diversity is proposed; (4) Design multiple constraint repair and penalty mechanisms to improve the feasibility and stability of the solution.

[0138] III. Simulation Results The heterogeneous satellite system in the simulation experiment includes one high-orbit, one medium-orbit, and one low-orbit satellite, with 50 edge master stations. Figure 6 This is a scenario diagram illustrating the resource scheduling problem of heterogeneous satellites and maritime satellite IoT in one embodiment. Specifically, black stars and red circles represent satellite groups and edge master station groups, respectively. The population size and number of iterations in MDMHA are both 100. Figure 7 This is an example of a correlation scheme between an edge master station and a satellite at a certain moment. Figure 7 Each satellite in the system is connected to multiple edge master stations.

[0139] Figure 8 This illustrates how the number of satellite resource blocks occupied changes with the number of algorithm iterations in one embodiment. From... Figure 8 It can be seen that after the algorithm initialization, the satellite resource blocks are essentially randomly allocated, with each block occupying a relatively high number of resources. In the early stages of iteration, the number of satellite resource blocks decreases significantly. As the number of iterations increases, the number of satellite resource blocks remains relatively stable between iterations 40 and 100, indicating that the algorithm has converged. Figure 9 This is a schematic diagram illustrating the process of changing the number of satellite resource blocks in one embodiment. Figure 8 and Figure 9 The convergence trend is similar. In the later stages of the algorithm iteration, the number of satellite resource blocks occupied does not change significantly, but compared with the early stages of the iteration, the number of occupied blocks decreases significantly, and the algorithm significantly improves the utilization rate of satellite resources.

[0140] Table 1 shows the performance gain of the MDMHA algorithm compared to traditional algorithms in this embodiment. Specifically, the MDMHA algorithm is compared with traditional genetic algorithms (GA) and particle swarm optimization (PSO). The population size for GA, PSO, and MDMHA is 100, and the number of iterations for all three algorithms is 100. Tests 1-2 include 20 edge master stations and 2 satellites, tests 3-4 include 20 edge master stations and 3 satellites, and tests 5-6 include 30 edge master stations and 2 satellites. Each algorithm is run 10 times across all tests, and the average is taken.

[0141] Table 1 Statistical results of the three algorithms Table 1 shows the statistical results of the three algorithms. After adopting multidimensional vector coding and MDMHA, the satellite resource utilization rate was significantly reduced. MDMHA combines the advantages of GA and PSO, thereby enhancing the population's search capability, meeting user service needs with fewer resource blocks, and improving the overall resource utilization rate of heterogeneous satellite networks.

[0142] This embodiment develops a heterogeneous satellite network resource scheduling system, proposes a task-driven network planning model, and establishes a multi-resource collaborative scheduling model for maritime heterogeneous satellite systems, significantly improving the utilization efficiency of satellite communication resources. Furthermore, at the algorithm solution level, this embodiment further proposes a heuristic solution algorithm based on multi-dimensional matrix encoding, effectively solving the resource allocation problem of heterogeneous satellites. Simulation results show that, compared with traditional heuristic algorithms, the proposed algorithm can effectively improve the resource utilization rate of heterogeneous satellite networks by approximately 5% to 15%, avoiding resource congestion and idleness of multiple satellites.

[0143] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0144] Based on the same inventive concept, this application also provides a heterogeneous satellite resource allocation device for a maritime IoT scenario, which implements the above-described heterogeneous satellite resource allocation method for maritime IoT scenarios. The solution provided by this device is similar to the implementation described in the above method. Therefore, the specific limitations of one or more embodiments of the heterogeneous satellite resource allocation device for maritime IoT scenarios provided below can be found in the limitations of the heterogeneous satellite resource allocation method for maritime IoT scenarios described above, and will not be repeated here.

[0145] In one exemplary embodiment, a heterogeneous satellite resource allocation device for a maritime Internet of Things (IoT) scenario is provided, comprising: an acquisition module, a first optimization module, a first constraint module, a second optimization module, a second constraint module, and a target scheduling matrix determination module, wherein: The acquisition module is used to acquire the objective function that minimizes the use of heterogeneous satellite resources and the constraints corresponding to the objective function; the objective function includes a multi-dimensional scheduling matrix, which represents the use of resource blocks of heterogeneous satellites by the edge master station; The first optimization module is used to perform optimization processing on the multidimensional scheduling matrix using a genetic algorithm to obtain a candidate genetic scheduling matrix; The first constraint module is used to perform constraint processing on the candidate genetic scheduling matrix according to the constraint conditions to obtain the constrained genetic scheduling matrix. The second optimization module is used to optimize the multidimensional scheduling matrix using the particle swarm optimization algorithm to obtain a candidate particle scheduling matrix. The second constraint module is used to perform constraint processing on the candidate particle scheduling matrix according to the constraint conditions to obtain the constrained particle scheduling matrix. The target scheduling matrix determination module is used to determine the target scheduling matrix from the constrained genetic scheduling matrix and the constrained particle scheduling matrix according to the evaluation function; the target scheduling matrix is ​​used to allocate resource blocks of corresponding satellites to the edge master station via the maritime Internet.

[0146] The modules in the aforementioned heterogeneous satellite resource allocation device for maritime IoT scenarios can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0147] In one exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 10As shown, the computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores data. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communicating with external terminals via a network. When executed by the processor, the computer program implements a heterogeneous satellite resource allocation method for a maritime Internet of Things (IoT) scenario.

[0148] Those skilled in the art will understand that Figure 10 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0149] In one exemplary embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the heterogeneous satellite resource allocation methods described above for maritime Internet of Things scenarios.

[0150] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described heterogeneous satellite resource allocation methods for maritime Internet of Things scenarios.

[0151] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps of the above-described heterogeneous satellite resource allocation methods for maritime Internet of Things scenarios.

[0152] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0153] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0154] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for allocating heterogeneous satellite resources for maritime Internet of Things (IoT) scenarios, characterized in that, The method includes: Obtain the objective function that minimizes the use of heterogeneous satellite resources and the corresponding constraints of the objective function; the objective function includes a multi-dimensional scheduling matrix, which represents the use of resource blocks of heterogeneous satellites by the edge master station; A genetic algorithm is used to optimize the multidimensional scheduling matrix to obtain candidate genetic scheduling matrices; The candidate genetic scheduling matrix is ​​constrained according to the constraints to obtain the constrained genetic scheduling matrix. The multidimensional scheduling matrix is ​​optimized using a particle swarm optimization algorithm to obtain a candidate particle scheduling matrix; The candidate particle scheduling matrix is ​​constrained according to the constraints to obtain the constrained particle scheduling matrix. Based on the evaluation function, a target scheduling matrix is ​​determined from the constrained genetic scheduling matrix and the constrained particle scheduling matrix; the target scheduling matrix is ​​used to allocate resource blocks of corresponding satellites to the edge master station via the maritime Internet.

2. The method according to claim 1, characterized in that, The constraints include one or more of the following: edge master station transmission resource constraints, latency constraints, total satellite resource constraints, power matching constraints, beam spatial isolation constraints, and single-time-slot single-satellite access constraints.

3. The method according to claim 1, characterized in that, The constraints include repair constraints; The step of constraining the candidate genetic scheduling matrix according to the constraints to obtain a constrained genetic scheduling matrix includes: The matrix elements in the candidate genetic scheduling matrix are subjected to repair constraint processing according to the repair constraint conditions to obtain the repaired genetic scheduling matrix. The constrained genetic scheduling matrix is ​​determined based on the repaired genetic scheduling matrix.

4. The method according to claim 3, characterized in that, The step of performing repair constraint processing on the matrix elements in the candidate genetic scheduling matrix according to the repair constraint conditions to obtain the repaired genetic scheduling matrix includes: Based on the power constraints and the transmit power of each edge master station, a genetic scheduling matrix that matches the transmit power between the edge master station and the satellite is selected from the candidate genetic scheduling matrices; In the genetic scheduling matrix that matches the transmit power, matrix elements that exceed the latency requirements of the edge master station are set to not be occupied, matrix elements that exceed the total amount of satellite resources are set to not be occupied, and matrix elements that allow the edge master station to access more than 1 satellite in each time slot are set to not be occupied, thus obtaining the repaired genetic scheduling matrix.

5. The method according to claim 3, characterized in that, The constraints include penalty constraints, which include edge master station transmission resource constraints. The step of determining the constrained genetic scheduling matrix based on the repaired genetic scheduling matrix includes: Obtain the target data volume for each edge master station; The amount of allocated data for each edge master station is determined based on the repaired genetic scheduling matrix. The penalty value of the repaired genetic scheduling matrix is ​​determined based on the difference between the target data volume and the allocated data volume; The constraint genetic scheduling matrix is ​​determined based on the penalty value.

6. The method according to claim 3, characterized in that, The constraints include penalty constraints, which in turn include beam space isolation constraints. The step of determining the constrained genetic scheduling matrix based on the repaired genetic scheduling matrix includes: The beam angle between each edge master station is determined based on the edge master station corresponding to the satellite in the repaired genetic scheduling matrix; The penalty value of the repaired genetic scheduling matrix is ​​determined based on the beam angle and beam isolation angle threshold between each edge master station; The constraint genetic scheduling matrix is ​​determined based on the penalty value.

7. The method according to claim 1, characterized in that, The step of constraining the candidate particle scheduling matrix according to the constraints to obtain the constrained particle scheduling matrix includes: The candidate particle scheduling matrix is ​​then normalized, and then repaired and penalized according to the constraints to obtain the constrained particle scheduling matrix.

8. The method according to claim 1, characterized in that, The step of determining the target scheduling matrix from the constrained genetic scheduling matrix and the constrained particle scheduling matrix based on the evaluation function includes: In each iteration, the evaluation function values ​​of the constraint genetic scheduling matrix and the constraint particle scheduling matrix for the current round are determined according to the evaluation function. The scheduling matrix set for the current round is obtained by sorting and filtering the evaluation function values ​​of the constraint genetic scheduling matrix and constraint particle scheduling matrix of the current round, as well as the evaluation function values ​​of the constraint genetic scheduling matrix and constraint particle scheduling matrix of the previous round. When the iteration termination condition is not met, construct the multidimensional scheduling matrix for the next round based on the current round's scheduling matrix set, and return to the step of using a genetic algorithm to optimize the multidimensional scheduling matrix and obtain a candidate genetic scheduling matrix. Until the iteration termination condition is met, a multi-dimensional scheduling matrix with the optimal evaluation function value is determined from the scheduling matrix set of the current round to obtain the target scheduling matrix.

9. The method according to claim 1, characterized in that, The evaluation function includes a diversity index and an adaptability index; the diversity index is determined based on the difference between each constrained scheduling matrix and the optimal scheduling matrix; the adaptability index is determined based on the normalized objective function value of each constrained scheduling matrix.

10. The method according to claim 9, characterized in that, The weights of the diversity index and the fitness index are determined based on the number of iterations of the algorithm; the number of iterations is negatively correlated with the weight of the diversity index; and the number of iterations is positively correlated with the weight of the fitness index.

Images