Transmission task scheduling optimization method and device based on mixed learning strategy

Abstract

The embodiment of the invention discloses a transmission task scheduling optimization method and device based on a hybrid learning strategy, and relates to the field of the Internet of Things and artificial intelligence. The method comprises the following steps: constructing a task scheduling optimization model taking minimization of the total bandwidth cost of a system in a preset time period as a target function, initializing parameters of a hybrid learning strategy particle swarm optimization algorithm, including particle swarm scale, particle position and speed, iteratively optimizing a bandwidth allocation scheme according to the hybrid learning strategy particle swarm optimization algorithm, and optimizing the bandwidth allocation scheme according to the particle swarm optimization algorithm. And updating the individual historical optimal position of each particle and the global historical optimal position of the population in each iteration process, and when iteration reaches the maximum number of iterations, outputting a bandwidth allocation scheme corresponding to the global optimal position as an optimal transmission task scheduling scheme. In this way, fast and efficient allocation of network bandwidth resources is achieved.

Description

Technical Field

[0001] This invention relates to the fields of Internet of Things and artificial intelligence, and in particular to a method and apparatus for optimizing transmission task scheduling based on a hybrid learning strategy. Background Technology

[0002] In edge computing networks, efficient task scheduling is crucial for ensuring user service quality and improving resource utilization. To this end, researchers have widely adopted intelligent optimization algorithms such as genetic algorithms and particle swarm optimization (PSO), achieving significant progress in optimizing task completion time and energy efficiency. However, with the proliferation of data-intensive applications (such as video streaming and content delivery networks), network bandwidth cost has become a key factor affecting the operational efficiency of service providers. Although existing research has attempted to optimize bandwidth costs, these methods are typically based on the idealized assumption that tasks can be executed on any edge node, ignoring the complex constraints imposed by geographical distribution and Quality of Service (QoS) requirements in real-world networks. This limits the applicability of their scheduling schemes in practical scenarios.

[0003] Furthermore, although the PSO algorithm is used for task scheduling due to its advantages such as fewer parameters and faster convergence, the traditional PSO algorithm is prone to getting trapped in local optima when solving such highly constrained and nonlinear bandwidth cost optimization problems, making it difficult to generate an efficient scheduling scheme that meets both real-time requirements and minimizes costs.

[0004] Therefore, there is an urgent need for a new method for scheduling edge computing tasks that can effectively address complex real-world constraints and significantly reduce bandwidth costs. Summary of the Invention

[0005] To address the issues of limited bandwidth resources, uneven geographical distribution, and high data transmission overhead at edge nodes, this invention provides a method and apparatus for optimizing transmission task scheduling based on a hybrid learning strategy. The aim is to obtain the optimal task scheduling scheme and achieve rapid and efficient allocation of network bandwidth resources.

[0006] In a first aspect, embodiments of the present invention provide a method for optimizing transmission task scheduling based on a hybrid learning strategy, comprising:

[0007] Acquire bandwidth requirement data of multiple task nodes in the edge computing network, maximum bandwidth data of multiple edge nodes, and user service quality constraint data between task nodes and edge nodes;

[0008] Based on the bandwidth demand data, maximum carrying bandwidth data, and user service quality constraint data, a task scheduling optimization model is constructed with the objective function of minimizing the total system bandwidth cost within a preset time period.

[0009] Initialize the parameters of the hybrid learning strategy particle swarm optimization algorithm, including particle swarm size, particle position and velocity, where the position vector encoding of each particle represents a bandwidth allocation scheme;

[0010] The bandwidth allocation scheme is iteratively optimized according to the hybrid learning strategy particle swarm optimization algorithm, and the individual historical best position of each particle and the global historical best position of the population are updated in each iteration.

[0011] When the maximum number of iterations is reached, the bandwidth allocation scheme corresponding to the globally optimal position is output as the optimal transmission task scheduling scheme.

[0012] In a preferred embodiment, the step of iteratively optimizing the bandwidth allocation scheme according to the hybrid learning strategy particle swarm optimization algorithm, and updating the individual historical best position of each particle and the global historical best position of the population in each iteration, includes:

[0013] Calculate the fitness value of the scheme corresponding to the current position of each particle, wherein the fitness value represents the total system bandwidth cost calculated by the objective function;

[0014] The individual historical best position of each particle and the global historical best position of the population are updated based on the fitness value. When the current fitness value of a particle is lower than its individual historical best fitness value, the individual historical best position is updated. When the individual historical best fitness value of any particle is lower than the global historical best fitness value of the population, the global historical best position is updated.

[0015] The velocity of each particle is updated based on a hybrid learning strategy, and the position of each particle is updated based on the updated particle velocity. The hybrid learning strategy integrates an enhanced median learning strategy, a stochastic learning strategy, and an adaptive dynamic adjustment strategy.

[0016] In a preferred embodiment, the velocity of each particle is updated based on a hybrid learning strategy, and the position of each particle is updated based on the updated particle velocity, including:

[0017] If the current iteration count is less than half of the maximum iteration count, the velocity of all particles is updated using an enhanced median learning strategy.

[0018] If the current iteration count is greater than half of the maximum iteration count, for particles whose particle index is less than or equal to half of the total number of particles, the enhanced median learning strategy is used to update their velocity, and for the remaining particles, the random learning strategy is used to update their velocity.

[0019] Throughout the iteration process, an adaptive dynamic adjustment strategy is simultaneously adopted to update the inertia weights and learning factors, and the position of each particle is updated according to the updated velocity.

[0020] As a preferred implementation, the enhanced median learning strategy is implemented in the following way:

[0021] Sort all particles in the current population in ascending order according to their individual historical best fitness values;

[0022] For the i-th particle to be updated, select particles whose fitness value is better than or equal to the historical best fitness value of the i-th particle from the sorted particles that are located before it, and form the sample pool of that particle.

[0023] Select a particle at the median position from the example pool, and use its individual historical best position as the learning object to update the velocity of the i-th particle.

[0024] As a preferred implementation, the random learning strategy is implemented in the following way:

[0025] From the set of the best historical positions of all particles in the current population, a position is selected for the i-th particle to be updated using uniform random sampling.

[0026] The randomly selected position is used as the learning target, and the velocity of the i-th particle is updated by combining the parameters obtained from the adaptive dynamic adjustment strategy.

[0027] As a preferred embodiment.

[0028] In a preferred embodiment, the constraints of the task scheduling optimization model include:

[0029] The bandwidth requirements of each task node at any given moment must be fully allocated to one or more edge nodes.

[0030] The total bandwidth allocated to each edge node at any given moment must not exceed its maximum carrying bandwidth.

[0031] Task nodes can only be scheduled to edge nodes that meet their user service quality constraints, wherein the user service quality constraints are determined by the edge node level and the network latency threshold.

[0032] The bandwidth allocated to each edge node is no less than a preset proportion of the node's maximum carrying bandwidth, and the bandwidth allocated to any edge node is a non-negative real number.

[0033] In a preferred embodiment, the method further includes generating initial particle positions using a constraint-satisfying heuristic method, specifically including:

[0034] Based on user service quality constraints, determine the range of edge nodes that each task node can access;

[0035] Within the range of the edge nodes, an initial bandwidth scheme is allocated to each task node based on the bandwidth requirements of the task nodes and the remaining bandwidth that the edge nodes can bear.

[0036] If the initial bandwidth allocation scheme causes the total allocation of any edge node to exceed its maximum carrying bandwidth, it is corrected by excess bandwidth migration until all constraints are satisfied.

[0037] In a preferred embodiment, the correction via excess bandwidth migration includes:

[0038] Identify edge nodes whose current bandwidth allocation exceeds their maximum carrying bandwidth, and calculate the excess bandwidth value between the actual bandwidth allocation and the maximum carrying bandwidth of the edge node.

[0039] For task nodes that have previously allocated bandwidth to the excess edge nodes, a candidate migration node set is formed by selecting edge nodes with remaining bandwidth greater than a preset threshold from the range of selectable edge nodes that meet the user service quality constraints.

[0040] In the candidate migration node set, the node with the lowest unit bandwidth cost billing coefficient is selected as the target migration node. All excess bandwidth of the excess edge node is migrated to the target migration node, and the bandwidth allocation ratio of the original task node to the excess edge node and the target migration node is adjusted accordingly.

[0041] After the migration is completed, check whether the bandwidth allocation of the original excess edge nodes meets the maximum carrying bandwidth constraint. If the bandwidth allocation of the edge node still exceeds the maximum carrying bandwidth, repeat the above steps of identifying excess edge nodes, screening candidate migration nodes, and migrating excess bandwidth until the bandwidth allocation of all edge nodes meets the maximum carrying bandwidth constraint.

[0042] In a preferred embodiment, the total system bandwidth cost is calculated as follows:

[0043] Based on the bandwidth consumed by each edge node within the preset time period, multiply by the unit bandwidth cost billing factor determined by the user service quality level between the edge node and the task node to obtain the total bandwidth cost of each edge node.

[0044] The total bandwidth cost of the system is obtained by weighted summation of the total bandwidth costs of each edge node.

[0045] Secondly, embodiments of the present invention also provide a transmission task scheduling optimization apparatus based on a hybrid learning strategy, the apparatus comprising:

[0046] The data acquisition module is used to acquire bandwidth demand data of multiple task nodes in the edge computing network, maximum bandwidth data of multiple edge nodes, and user service quality constraint data between task nodes and edge nodes.

[0047] The task scheduling optimization model construction module is used to construct a task scheduling optimization model with the objective function of minimizing the total system bandwidth cost within a preset time period, based on the bandwidth demand data, maximum carrying bandwidth data, and user service quality constraint data.

[0048] The initialization module is used to initialize the parameters of the hybrid learning strategy particle swarm optimization algorithm, including particle swarm size, particle position and velocity. The position vector encoding of each particle represents a bandwidth allocation scheme.

[0049] The iterative optimization module is used to iteratively optimize the bandwidth allocation scheme according to the hybrid learning strategy particle swarm optimization algorithm, and update the individual historical best position of each particle and the global historical best position of the population in each iteration.

[0050] The task scheduling scheme output module is used to output the bandwidth allocation scheme corresponding to the globally optimal position when the maximum number of iterations is reached, as the optimal transmission task scheduling scheme.

[0051] Thirdly, embodiments of the present invention also provide an electronic device, the electronic device comprising:

[0052] One or more processors;

[0053] Storage device for storing one or more programs;

[0054] When the one or more programs are executed by the one or more processors, the one or more processors implement the transmission task scheduling optimization method based on hybrid learning strategy as described in any embodiment of the present invention.

[0055] Fourthly, embodiments of the present invention also provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the transmission task scheduling optimization method based on a hybrid learning strategy as described in any embodiment of the present invention.

[0056] To achieve the above objectives, the beneficial effects of the present invention are:

[0057] This invention proposes a transmission task scheduling optimization method based on a hybrid learning strategy. First, a task scheduling system model is designed based on constraints such as user node network bandwidth requirements, user Quality of Service (QoS), maximum capacity bandwidth (MCB) of edge nodes, and upper and lower limits of bandwidth allocation. Second, a hybrid learning strategy particle swarm optimization algorithm is constructed to avoid getting trapped in local optima and quickly converge to the optimal solution. Then, the proposed algorithm is iteratively optimized to achieve optimal network cost, resulting in a real-time task scheduling scheme. This method can obtain the optimal task scheduling scheme in real time based on current network conditions and the processing capacity and QoS requirements of edge nodes, enabling rapid and efficient allocation of network bandwidth resources. Attached Figure Description

[0058] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings. The drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings:

[0059] Figure 1 This is a flowchart of a transmission task scheduling optimization method based on a hybrid learning strategy provided in an embodiment of the present invention;

[0060] Figure 2 This is a schematic diagram of a task scheduling system model in an edge computing network environment provided by an embodiment of the present invention;

[0061] Figure 3 This is a schematic diagram of the enhanced median learning strategy provided in an embodiment of the present invention;

[0062] Figure 4 This is a schematic diagram of the random learning strategy provided in an embodiment of the present invention;

[0063] Figure 5 This is an adaptive linear change trend graph provided in the embodiments of the present invention;

[0064] Figure 6 This is a diagram of the scheduling algorithm model based on HLSPSO provided in an embodiment of the present invention;

[0065] Figure 7 This is a comparison chart of the maximum carrying bandwidth of homogeneous and heterogeneous edge nodes provided in the embodiments of the present invention;

[0066] Figure 8 This is a comparison chart of the optimization effects of different algorithms provided in the embodiments of the present invention;

[0067] Figure 9 This is a schematic diagram of a transmission task scheduling optimization device based on a hybrid learning strategy provided in an embodiment of the present invention;

[0068] Figure 10 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0069] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it. Furthermore, it should be noted that, for ease of description, the accompanying drawings show only the parts relevant to the present invention, and not all of the structures.

[0070] Before discussing the exemplary embodiments in more detail, it should be mentioned that some exemplary embodiments are described as processes or methods depicted as flowcharts. Although the flowcharts describe the operations (or steps) as sequential processes, many of the operations (or steps) can be performed in parallel, concurrently, or simultaneously. Furthermore, the order of the operations can be rearranged. The process can be terminated when its operation is completed, but it may also have additional steps not included in the figures. The process may correspond to a method, function, procedure, subroutine, subroutine, etc.

[0071] The acquisition, storage, use, and processing of data in this application comply with relevant national laws and regulations. It should be noted that existing industry solutions such as software, components, or models may be mentioned in the embodiments of this application. These should be considered exemplary and intended only to illustrate the feasibility of implementing the technical solution of this application, but do not imply that the applicant has already used or necessarily used such a solution.

[0072] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in sequences other than those illustrated or described herein.

[0073] like Figure 1 The diagram shows a flowchart of a transmission task scheduling optimization method 100 based on a hybrid learning strategy provided by an embodiment of the present invention. The method 100 specifically includes the following steps:

[0074] S110. Obtain bandwidth requirement data of multiple task nodes in the edge computing network, maximum bandwidth data of multiple edge nodes, and user service quality constraint data between task nodes and edge nodes.

[0075] Among them, the user service quality constraint data includes edge node level, network latency threshold, etc., which are used to limit the range of edge nodes that task nodes can access.

[0076] S120. Based on the bandwidth demand data, maximum carrying bandwidth data, and user service quality constraint data, construct a task scheduling optimization model with the objective function of minimizing the total system bandwidth cost within a preset time period.

[0077] Combination Figure 2 As can be seen, the core objective of this invention is to optimize task scheduling strategies in edge computing networks to improve the quality of service (QoS) and minimize the total system bandwidth cost. Therefore, it is crucial to establish a task scheduling optimization model with the objective function of minimizing the total system bandwidth cost within a preset time period.

[0078] Specifically, a task scheduling optimization model package is constructed with the objective function of minimizing the total system bandwidth cost within a preset time period, including the following steps:

[0079] (1) Define the objective function

[0080] The bandwidth cost of the nth edge node at timestamp t is determined by the bandwidth it consumes and the unit bandwidth cost billing factor. Assume the total bandwidth of the nth edge node at timestamp t is:

[0081] ;

[0082] Then the bandwidth of the nth edge node at different times is:

[0083] ;

[0084] Multiply the bandwidth of the nth edge node at different times by the unit bandwidth cost billing factor to obtain the bandwidth cost generated by the nth edge node:

[0085] ;

[0086] Furthermore, the total system bandwidth cost within the preset time period is the sum of the bandwidth costs of all edge nodes at all timestamps within that period. Therefore, the objective function is expressed as follows:

[0087] .

[0088] in, This represents the billing factor for different edge nodes, and defines the unit cost of a node as... .

[0089] The constraints of the above objective function include:

[0090] (Full Coverage Constraint): The bandwidth requirement of each task node at any given time must be fully allocated to one or more edge nodes;

[0091] ;

[0092] in, For the current task node, For edge nodes, For the set of edge nodes, Indicates the current moment. For the first The task node at the _ ... The timestamp is assigned to the first The specific bandwidth of each edge node. For the first The task node at the _ ... Total bandwidth requirement for each timestamp.

[0093] (Resource limit constraint): The total bandwidth allocated to each edge node at any given time must not exceed its maximum carrying bandwidth;

[0094] ;

[0095] in, For the first The maximum bandwidth that each edge node can carry. This is a set of task nodes.

[0096] (QoS constraint): Task nodes can only be scheduled to edge nodes that meet their user quality of service constraints;

[0097] ;

[0098] in, As an indicator variable, =1 indicates the first The task node is scheduled to the task at the t-th timestamp. One edge node.

[0099] (Resource minimum constraint): The amount of bandwidth allocated to each edge node shall not be less than a preset proportion of the maximum bandwidth that node can carry;

[0100] ;

[0101] in, For the first The bandwidth lower limit of each edge node is preset to a certain percentage, and .

[0102] (Non-negativity constraint): The amount of bandwidth allocated to any edge node is a non-negative real number;

[0103] and .

[0104] (Time continuity constraint): The bandwidth allocation of the same task node at adjacent timestamps must remain smooth;

[0105] .

[0106] S130. Initialize the parameters of the hybrid learning strategy particle swarm optimization algorithm, including particle swarm size, particle position and velocity, wherein the position vector encoding of each particle represents a bandwidth allocation scheme.

[0107] To compute the asymptotically optimal solution of the objective function, thereby obtaining a set of bandwidth allocation schemes X that satisfy all constraints and minimize the total bandwidth cost within time period T, this embodiment proposes a Hybrid Learning Strategy Task Real-Time Scheduling Algorithm (HLSPSO) to solve the task scheduling problem in complex edge computing environments. The specific steps include:

[0108] (1) Initialization of core parameters

[0109] Based on the task scale and real-time scheduling requirements of the edge computing network, the key parameters of the HLSPSO algorithm are initialized. These key parameters include basic particle swarm parameters (particle swarm size, maximum number of iterations, and solution space dimension), initial particle velocity, initial position parameters, and learning strategy parameters.

[0110] (2) Definition of particle coding:

[0111] First, define a scheduling scheme X:

[0112]

[0113] Here, X represents a particle in the HLSPSO algorithm. In the task scheduling optimization problem, each particle represents a possible bandwidth allocation scheme. This represents the bandwidth allocated from the m-th task node to the n-th edge node.

[0114] (3) Constraints satisfy the heuristic method for generating initial particle positions:

[0115] To avoid initial particles violating constraints (such as edge node overload or incomplete task coverage), a heuristic method is used to generate initial positions, ensuring that each particle satisfies all constraints from the moment of generation. Specifically, this includes:

[0116] Based on user service quality constraints, determine the range of edge nodes that each task node can access;

[0117] Within the edge node range, based on the bandwidth requirements of the task nodes and the remaining bandwidth that the edge nodes can bear, an initial bandwidth scheme is allocated to each task node according to the principle of prioritizing low cost.

[0118] If the initial bandwidth allocation scheme causes the total allocation of any edge node to exceed its maximum carrying bandwidth, it will be corrected by excess bandwidth migration until all constraints are satisfied.

[0119] If the initial bandwidth allocation scheme results in the total allocated bandwidth of any edge node exceeding its maximum carrying bandwidth, a correction will be made through excess bandwidth migration, including:

[0120] (3.1) Identify edge nodes whose current bandwidth allocation exceeds their maximum carrying bandwidth, and calculate the excess bandwidth value between the actual bandwidth allocation and the maximum carrying bandwidth of the edge node;

[0121] (3.2) For the task node that has previously allocated bandwidth to the excess edge node, select the edge nodes with remaining bandwidth greater than the preset threshold from the range of selectable edge nodes that meet the user service quality constraints to form a candidate migration node set;

[0122] (3.3) In the candidate migration node set, select the node with the lowest unit bandwidth cost billing coefficient as the target migration node, migrate all the excess bandwidth of the excess edge node to the target migration node, and at the same time adjust the bandwidth allocation ratio of the original task node to the excess edge node and the target migration node accordingly.

[0123] (3.4) After the migration is completed, check whether the bandwidth allocation of the original excess edge nodes meets the maximum carrying bandwidth constraint. If the bandwidth allocation of the edge node still exceeds the maximum carrying bandwidth, then repeat the above steps of identifying excess edge nodes, screening candidate migration nodes, and migrating excess bandwidth until the bandwidth allocation of all edge nodes meets the maximum carrying bandwidth constraint.

[0124] S140. The bandwidth allocation scheme is iteratively optimized according to the particle swarm optimization algorithm of the hybrid learning strategy, and the individual historical best position of each particle and the global historical best position of the population are updated in each iteration.

[0125] In some embodiments, the fitness value of the scheme corresponding to the current position of each particle is calculated, wherein the fitness value represents the total system bandwidth cost calculated by the objective function;

[0126] The individual historical best position of each particle and the global historical best position of the population are updated based on the fitness value. When the current fitness value of a particle is lower than its individual historical best fitness value, the individual historical best position is updated. When the individual historical best fitness value of any particle is lower than the global historical best fitness value of the population, the global historical best position is updated.

[0127] The velocity of each particle is updated based on a hybrid learning strategy, and the position of each particle is updated based on the updated particle velocity.

[0128] To balance the global exploration and local development capabilities of the algorithm, the hybrid learning strategy integrates an enhanced median learning strategy, a stochastic learning strategy, and an adaptive dynamic adjustment strategy.

[0129] like Figure 3 As shown, the Enhanced Median Learning Strategy (EMLS) utilizes median particles, avoiding both deviation from the overall population direction and excessive conservatism. In task scheduling, EMLS balances the impact of different allocation schemes, preventing a single extreme case from overly influencing the overall scheduling plan. By selecting the median of the historical best positions of all particles in the population, the algorithm can make more robust decisions when choosing solutions, thus avoiding getting trapped in local optima. This is particularly important when network load fluctuates significantly, as the strategy ensures that the scheduling plan remains stable under different network conditions. Each particle (i.e., a possible scheduling scheme) is sorted in descending order of fitness value, and those particles considered superior or equivalent constitute the example pool. Each particle in the population has a unique example pool. From the example pool, the particle at the median position is selected as the optimal individual, Pbest.

[0130] To optimize the learning process of the scheduling scheme, the velocity update equation was precisely improved:

[0131] ;

[0132] ;

[0133] in, This is the deviation amount. Indicates the first One particle (a possible scheduling scheme) The optimal solution obtained by the search, Indicates the scheduling scheme index. This represents a uniformly random number with values ​​ranging from [0,1]. It is a one-dimensional random function used to enhance the randomness of the search. This represents a constant parameter used to control the degree to which the guiding direction affects particle velocity updates. To guide the direction, Adjusting trends for historical purposes This indicates the updated speed.

[0134] Furthermore, to enhance the adaptability and robustness of the PSO algorithm, this invention introduces a PSO variant (RLSPSO) incorporating a stochastic learning strategy, such as... Figure 4 As shown, the RLSPSO algorithm improves the diversity and exploration ability of the population by allowing particles to randomly select the optimal positions of other particles for learning. This random selection strategy breaks the traditional convergence mode towards the global optimum or individual optimum, providing particles with more opportunities to explore different regions of the solution space, thereby reducing the risk of premature convergence to a local optimum.

[0135] The RLSPSO algorithm updates the velocity of each particle as follows:

[0136] ;

[0137]

[0138] in, From The group is randomly selected, and compared to the traditional PSO algorithm, RLSPSO only requires manually setting one parameter. .

[0139] The stochastic learning strategy helps achieve a more balanced trade-off between global and local search. Some particles may choose to learn from particles closer to the global optimum, thus enhancing their local exploitation capabilities, while others may choose particles further away to strengthen their global exploration capabilities. This strategy essentially improves the robustness of the PSO algorithm, enabling it to operate effectively in a wide range of optimization problems. By incorporating stochastic selection, the algorithm exhibits better adaptability to diverse optimization scenarios, thereby improving convergence speed and solution accuracy.

[0140] The standard PSO algorithm defines a fixed inertia weight. To coordinate the global and local search capabilities of the PSO algorithm, it is usually defined as follows: .if A value that is relatively large and close to 1 is beneficial for global optimization. If the inertia weight is too high and close to 0, it is beneficial for local optimization. Choose an appropriate inertia weight. This helps the PSO algorithm balance its search capabilities:

[0141] ;

[0142] in, , , It is the maximum number of iterations. This is the current iteration number. The adaptive decreases with increasing iterations, thus ensuring the HLSPSO model converges to a local minimum. Similarly, the individual learning factor... and Make dynamic adjustments:

[0143] ;

[0144] / G;

[0145] Preferably, the change in inertia weight Change curve and individual learning factor and The change curve is as follows Figure 5 As shown, the inertia weight changes with the increase of the number of iterations. It shows a linear downward trend. It also shows a linear downward trend. It shows a linear upward trend. along with The decrease leads to an increase. Dynamically reducing inertia weights and individual learning factors. It helps to enhance exploration in the early stages of the algorithm and strengthen development in the later stages, enabling us to find the near-optimal bandwidth cost in task scheduling problems more quickly and accurately.

[0146] By combining enhanced median learning, stochastic learning, and dynamic adjustment strategies, HLSPSO demonstrates significant advantages in edge computing task scheduling. It enhances the algorithm's adaptability and global optimization capabilities, enabling it to more effectively optimize task scheduling in complex and dynamic edge computing environments.

[0147] Combination such as Figure 6 The flowchart shown below illustrates the scheduling algorithm model based on HLSPSO. The specific steps are as follows:

[0148] Step 1, Initialization Process: Initialize particle swarm parameters, including swarm size, particle dimensions, number of iterations, inertia weight, learning factor, and maximum particle velocity. Initialize the velocity and position of each particle, setting the initial optimal individual position (…). ) is set as the initial position, and the global optimal position is determined among all particles. ).

[0149] Step 2, Iterative process: Calculate the fitness function value for each particle, which represents the total bandwidth cost; the lower the fitness value, the better the scheduling scheme.

[0150] Step 3, Velocity and Position Update Process: Update the historical best fitness function value and position for each particle, as well as the historical best global fitness function value and position for the entire swarm. Update the velocity of each particle using a hybrid learning strategy. The velocity of each particle is updated according to the hybrid learning strategy and converted into velocity:

[0151]

[0152] in, , . , .

[0153] Location update:

[0154] ;

[0155] in, This represents the speed scaling factor.

[0156] Step 4: Determine whether to end the iteration: If the maximum number of iterations has not been reached, return to step 2 to continue iterating; otherwise, output the globally optimal cost and the corresponding scheduling scheme and end the iteration. The final output reflects the most efficient scheme for allocating bandwidth-intensive tasks among edge nodes to ensure bandwidth usage and cost minimization.

[0157] S150. When the iteration reaches the maximum number of iterations, output the bandwidth allocation scheme corresponding to the globally optimal position as the optimal transmission task scheduling scheme.

[0158] In some embodiments, when the maximum number of iterations G is reached, the bandwidth allocation scheme corresponding to the globally optimal position of the population is the optimal solution, and this scheme is output as the final transmission task scheduling scheme. This scheme satisfies all constraints (full demand coverage, resource limits, QoS matching, etc.) and has the lowest total bandwidth cost within the time period T.

[0159] The embodiments of the present invention achieve the following beneficial effects:

[0160] To address the challenges of limited bandwidth resources, uneven geographical distribution, and high data transmission overhead at edge nodes, this invention designs an intelligent optimization algorithm, HLSPSO, based on a hybrid learning strategy to minimize network bandwidth costs. First, a task scheduling system model and mathematical model are designed based on constraints such as user node network bandwidth requirements, user Quality of Service (QoS), maximum capacity bandwidth (MCB) of edge nodes, and upper and lower limits of bandwidth allocation. Second, a hybrid learning strategy particle swarm optimization algorithm is constructed to avoid getting trapped in local optima and to quickly converge to the optimal solution. Then, the proposed algorithm is iteratively optimized to achieve optimal network costs, resulting in a real-time task scheduling scheme. Finally, experiments are conducted using real bandwidth requirements, predicted bandwidth requirements, QoS, and MCB data to verify the feasibility and robustness of the algorithm. It can obtain the optimal task scheduling scheme in real time based on current network conditions, the processing capacity of edge nodes, and QoS requirements, significantly reducing costs and achieving rapid and efficient allocation of network bandwidth resources.

[0161] Example 2

[0162] This embodiment aims to verify the effectiveness and superiority of the proposed hybrid learning strategy-based transmission task scheduling optimization method in transmission task scheduling optimization through detailed comparative experiments.

[0163] Hardware environment: The experiments were conducted on a computer with an Intel(R) Core(TM) i9 10900X CPU @ 3.70 GHz. All experiments were performed on the same computer.

[0164] Core parameters: The number of fixed edge nodes and other parameters are compared to evaluate the performance of the proposed algorithm under different task scales. The parameter settings for the task scheduling experiment based on HLSPSO are shown in Table 3-2.

[0165] Table 3-2

[0166]

[0167] Vmax is set to 200 to limit the maximum speed of the particle search, preventing particles from moving too fast and missing potential optimal solutions. This helps maintain the stability of the algorithm and ensures that particles have sufficient opportunity to thoroughly explore potential optimal regions. The maximum speed setting effectively balances global and local search capabilities, preventing particles from ignoring important search areas due to excessive speed.

[0168] like Figure 7 As shown, MCB-EN-IS represents the maximum bandwidth of an edge node under ideal conditions. MCB-HEN-RS represents the maximum bandwidth of a heterogeneous edge node in a real-world scenario. The applicability to both heterogeneous and homogeneous nodes demonstrates the adaptability of the algorithm proposed in this invention.

[0169] Extensive experiments were conducted to evaluate the performance impact of various factors, including changes in key parameters, different numbers of tasks, and the maximum bandwidth of different edge nodes, on the proposed algorithm. Evaluation metrics included: convergence speed, total bandwidth, and fitness function (i.e., minimizing bandwidth cost).

[0170] To avoid experimental randomness caused by the initial randomness of particles in the model and the different operating states of hardware devices, all simulation experiments were repeated 10 times and the average value was taken as the final result.

[0171] To explore whether the proposed strategy can improve the PSO algorithm in the task scheduling bandwidth optimization problem, ablation experiments were conducted using a method of retaining only a single optimization strategy to investigate the contribution of each strategy in HLSPSO. EMLSPSO is defined as a particle swarm optimization algorithm that retains only the enhanced median learning strategy; RLSPSO is defined as a particle swarm optimization algorithm that retains only the stochastic learning strategy; and DSPSO is defined as a particle swarm optimization algorithm that retains only the dynamically adjusted strategy. The bandwidth cost optimization results of the original PSO, the three algorithms with independent optimization strategies, and the PSO with the hybrid learning strategy are shown in Table 3-3.

[0172] Table 3-3

[0173]

[0174] As shown in Table 3-3, the single strategies exhibit differences in network bandwidth cost and convergence speed. The original PSO has a bandwidth cost of 897,981 and a convergence iteration count of 27 generations. After introducing the enhanced median learning strategy, the bandwidth cost of EMLSPSO decreases to 897,729, a cost reduction of 252, and the convergence iteration count is shortened to 13 generations, indicating that this strategy can effectively improve optimization accuracy and accelerate convergence speed. Based on this, the cost of the dynamically adjusted strategy DSPSO is 897,764. Although its convergence speed is slightly slower than EMLSPSO, it is still better than the original PSO, indicating that it has the ability to adjust the search process.

[0175] Furthermore, while the stochastic learning strategy RLSPSO has a slightly lower cost than the original PSO, it requires 21 iterations to converge. Compared to the reinforced median learning strategy and the dynamic adjustment strategy, its improvement in optimization performance and speed is limited, and it may even introduce perturbations that affect search stability. Combining the advantages of the three strategies, HLSPSO achieves the optimal result, reducing bandwidth cost to 897,693, a reduction of 288 compared to the original PSO, and also reducing the number of convergence iterations to 16, demonstrating excellent global search capabilities and robustness. In summary, the reinforced median learning strategy plays the most crucial role in improving performance, the dynamic adjustment strategy has some collaborative optimization capabilities, while the stochastic strategy has weak applicability. HLSPSO, through the combination of these three strategies, effectively balances optimization performance and search efficiency, validating its application value in complex task scheduling scenarios.

[0176] Comparison with baseline algorithm experimental results:

[0177] To further verify the robustness and adaptability of the proposed method, another set of experiments was conducted. Peak traffic usage times were selected as a time period, and the bandwidth requirements of 200 task nodes were used as input for scheduling. The results are as follows: Figure 9 As shown.

[0178] As shown in Tables 3-4, the HLSPSO proposed in this paper achieves better bandwidth and cost compared to other representative algorithms.

[0179] Table 3-4

[0180]

[0181] In terms of objective function optimization performance, HLSPSO outperforms ACLPSO, TLSPSO, PSO, and SELPSO. HLSPSO also outperforms the other baseline algorithms in terms of convergence speed. Since the enhanced median learning strategy can accelerate convergence, the proposed algorithm initially uses only the median learning strategy for optimization. However, to avoid entering local optima later, a stochastic learning strategy is added, ultimately resulting in a better objective function value.

[0182] Compared to other algorithms, the proposed method achieves the lowest bandwidth cost. The superiority of the proposed method is evident in both the average and standard deviation of ten experiments. Compared to traditional PSO, the bandwidth is reduced by 279.5 Mbps. Compared to the optimal bandwidth of representative algorithms SELPSO, TLSPSO, and ACLPSO, the bandwidth reduction is 2528.7 Mbps, 103.4 Mbps, and 279.5 Mbps, respectively.

[0183] In summary, through experimental verification using real bandwidth demand, predicted bandwidth demand, QoS, MCB, and other data, the transmission task scheduling optimization method based on the hybrid learning strategy proposed in this invention can obtain the optimal task scheduling scheme in real time according to the current network conditions, the processing capabilities of edge nodes, and QoS requirements, thereby achieving rapid and efficient allocation of network bandwidth resources.

[0184] Example 3

[0185] Figure 9 This is a schematic diagram of a transmission task scheduling optimization device based on a hybrid learning strategy provided in Embodiment 3 of the present invention. Figure 9 As shown, the system includes:

[0186] The data acquisition module 910 is used to acquire bandwidth demand data of multiple task nodes in the edge computing network, maximum carrying bandwidth data of multiple edge nodes, and user service quality constraint data between task nodes and edge nodes.

[0187] The task scheduling optimization model construction module 920 is used to construct a task scheduling optimization model with the objective function of minimizing the total system bandwidth cost within a preset time period, based on the bandwidth demand data, maximum carrying bandwidth data, and user service quality constraint data.

[0188] Initialization module 930 is used to initialize the parameters of the hybrid learning strategy particle swarm optimization algorithm, including particle swarm size, particle position and velocity, wherein the position vector encoding of each particle represents a bandwidth allocation scheme.

[0189] The iterative optimization module 940 is used to iteratively optimize the bandwidth allocation scheme according to the hybrid learning strategy particle swarm optimization algorithm, and update the individual historical best position of each particle and the global historical best position of the population in each iteration.

[0190] The task scheduling scheme output module 950 is used to output the bandwidth allocation scheme corresponding to the globally optimal position when the iteration reaches the maximum number of iterations, as the optimal transmission task scheduling scheme.

[0191] The transmission task scheduling optimization device based on hybrid learning strategy provided in the embodiments of the present invention can execute the transmission task scheduling optimization method based on hybrid learning strategy provided in any of the embodiments of the present invention, and has the corresponding functions and beneficial effects of executing the method. For details, please refer to the relevant operations of the transmission task scheduling optimization method based on hybrid learning strategy in the foregoing embodiments.

[0192] Example 4

[0193] Figure 10 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. The electronic device 10 is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices (such as helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0194] like Figure 10As shown, the electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory (ROM) 12 or a random access memory (RAM) 13, communicatively connected to the at least one processor 11. The memory stores computer programs executable by the at least one processor. The processor 11 can perform various appropriate actions and processes based on the computer program stored in the ROM 12 or loaded into the RAM 13 from storage unit 18. The RAM 13 can also store various programs and data required for the operation of the electronic device 10. The processor 11, ROM 12, and RAM 13 are interconnected via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0195] Multiple components in electronic device 10 are connected to I / O interface 15, including: input unit 16, such as keyboard, mouse, etc.; output unit 17, such as various types of displays, speakers, etc.; storage unit 18, such as disk, optical disk, etc.; and communication unit 19, such as network card, modem, wireless transceiver, etc. Communication unit 19 allows electronic device 10 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0196] Processor 11 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. Processor 11 performs the various methods and processes described above, such as a transport task scheduling optimization method based on a hybrid learning strategy.

[0197] In some embodiments, the transmission task scheduling optimization method based on a hybrid learning strategy can be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program can be loaded and / or installed on the electronic device 10 via ROM 12 and / or communication unit 19. When the computer program is loaded into RAM 13 and executed by processor 11, one or more steps of the transmission task scheduling optimization based on a hybrid learning strategy described above can be performed. Alternatively, in other embodiments, processor 11 can be configured to perform the transmission task scheduling optimization method based on a hybrid learning strategy by any other suitable means (e.g., by means of firmware).

[0198] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0199] Computer programs used to implement the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0200] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0201] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0202] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or middleware components (e.g., application servers), or frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.

[0203] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system to address the shortcomings of traditional physical hosts and VPS services, such as high management difficulty and weak business scalability.

[0204] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.

[0205] The above embodiments are merely illustrative examples and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. A method for optimizing transmission task scheduling based on a hybrid learning strategy, characterized in that, The method includes: Acquire bandwidth requirement data of multiple task nodes in the edge computing network, maximum bandwidth data of multiple edge nodes, and user service quality constraint data between task nodes and edge nodes; Based on the bandwidth demand data, maximum carrying bandwidth data, and user service quality constraint data, a task scheduling optimization model is constructed with the objective function of minimizing the total system bandwidth cost within a preset time period. Initialize the parameters of the hybrid learning strategy particle swarm optimization algorithm, including particle swarm size, particle position and velocity, where the position vector encoding of each particle represents a bandwidth allocation scheme; The bandwidth allocation scheme is iteratively optimized according to the hybrid learning strategy particle swarm optimization algorithm, and the individual historical best position of each particle and the global historical best position of the population are updated in each iteration. When the maximum number of iterations is reached, the bandwidth allocation scheme corresponding to the globally optimal position is output as the optimal transmission task scheduling scheme.

2. The method according to claim 1, characterized in that, The step of iteratively optimizing the bandwidth allocation scheme according to the hybrid learning strategy particle swarm optimization algorithm, updating the individual historical best position of each particle and the global historical best position of the population in each iteration, includes: Calculate the fitness value of the scheme corresponding to the current position of each particle, wherein the fitness value represents the total system bandwidth cost calculated by the objective function; The individual historical best position of each particle and the global historical best position of the population are updated based on the fitness value. When the current fitness value of a particle is lower than its individual historical best fitness value, the individual historical best position is updated. When the individual historical best fitness value of any particle is lower than the global historical best fitness value of the population, the global historical best position is updated. The velocity of each particle is updated based on a hybrid learning strategy, and the position of each particle is updated based on the updated particle velocity. The hybrid learning strategy integrates an enhanced median learning strategy, a stochastic learning strategy, and an adaptive dynamic adjustment strategy.

3. The method according to claim 2, characterized in that, The velocity of each particle is updated based on a hybrid learning strategy, and the position of each particle is updated based on the updated particle velocity, including: If the current iteration count is less than half of the maximum iteration count, the velocity of all particles is updated using an enhanced median learning strategy. If the current iteration count is greater than half of the maximum iteration count, for particles whose particle index is less than or equal to half of the total number of particles, the enhanced median learning strategy is used to update their velocity, and for the remaining particles, the random learning strategy is used to update their velocity. Throughout the iteration process, an adaptive dynamic adjustment strategy is simultaneously adopted to update the inertia weights and learning factors, and the position of each particle is updated according to the updated velocity.

4. The method according to claim 3, characterized in that, The enhanced median learning strategy is implemented in the following ways: Sort all particles in the current population in ascending order according to their individual historical best fitness values; For the i-th particle to be updated, select particles whose fitness value is better than or equal to the historical best fitness value of the i-th particle from the sorted particles that are located before it, and form the sample pool of that particle. Select a particle at the median position from the example pool, and use its individual historical best position as the learning object to update the velocity of the i-th particle.

5. The method according to claim 3, characterized in that, The random learning strategy is implemented in the following way: From the set of the best historical positions of all particles in the current population, a position is selected for the i-th particle to be updated using uniform random sampling. The randomly selected position is used as the learning target, and the velocity of the i-th particle is updated by combining the parameters obtained from the adaptive dynamic adjustment strategy.

6. The method according to claim 1, characterized in that, The constraints of the task scheduling optimization model include: The bandwidth requirements of each task node at any given moment must be fully allocated to one or more edge nodes. The total bandwidth allocated to each edge node at any given moment must not exceed its maximum carrying bandwidth. Task nodes can only be scheduled to edge nodes that meet their user service quality constraints, wherein the user service quality constraints are determined by the edge node level and the network latency threshold. The bandwidth allocated to each edge node is no less than a preset proportion of the node's maximum carrying bandwidth, and the bandwidth allocated to any edge node is a non-negative real number.

7. The method according to claim 1, characterized in that, The method further includes generating initial particle positions using a constraint-satisfying heuristic method, specifically including: Based on user service quality constraints, determine the range of edge nodes that each task node can access; Within the range of the edge nodes, an initial bandwidth scheme is allocated to each task node based on the bandwidth requirements of the task nodes and the remaining bandwidth that the edge nodes can bear. If the initial bandwidth allocation scheme causes the total allocation of any edge node to exceed its maximum carrying bandwidth, it is corrected by excess bandwidth migration until all constraints are satisfied.

8. The method according to claim 7, characterized in that, The correction through excess bandwidth migration includes: Identify edge nodes whose current bandwidth allocation exceeds their maximum carrying bandwidth, and calculate the excess bandwidth value between the actual bandwidth allocation and the maximum carrying bandwidth of the edge node. For task nodes that have previously allocated bandwidth to the excess edge nodes, a candidate migration node set is formed by selecting edge nodes with remaining bandwidth greater than a preset threshold from the range of selectable edge nodes that meet the user service quality constraints. In the candidate migration node set, the node with the lowest unit bandwidth cost billing coefficient is selected as the target migration node. All excess bandwidth of the excess edge node is migrated to the target migration node, and the bandwidth allocation ratio of the original task node to the excess edge node and the target migration node is adjusted accordingly. After the migration is completed, check whether the bandwidth allocation of the original excess edge nodes meets the maximum carrying bandwidth constraint. If the bandwidth allocation of the edge node still exceeds the maximum carrying bandwidth, repeat the above steps of identifying excess edge nodes, screening candidate migration nodes, and migrating excess bandwidth until the bandwidth allocation of all edge nodes meets the maximum carrying bandwidth constraint.

9. The method according to claim 1, characterized in that, The total bandwidth cost of the system is calculated as follows: Based on the bandwidth consumed by each edge node within the preset time period, multiply by the unit bandwidth cost billing factor determined by the user service quality level between the edge node and the task node to obtain the total bandwidth cost of each edge node. The total bandwidth cost of the system is obtained by weighted summation of the total bandwidth costs of each edge node.

10. A transmission task scheduling optimization device based on a hybrid learning strategy, characterized in that, The device includes: The data acquisition module is used to acquire bandwidth demand data of multiple task nodes in the edge computing network, maximum bandwidth data of multiple edge nodes, and user service quality constraint data between task nodes and edge nodes. The task scheduling optimization model construction module is used to construct a task scheduling optimization model with the objective function of minimizing the total system bandwidth cost within a preset time period, based on the bandwidth demand data, maximum carrying bandwidth data, and user service quality constraint data. The initialization module is used to initialize the parameters of the hybrid learning strategy particle swarm optimization algorithm, including particle swarm size, particle position and velocity. The position vector encoding of each particle represents a bandwidth allocation scheme. The iterative optimization module is used to iteratively optimize the bandwidth allocation scheme according to the hybrid learning strategy particle swarm optimization algorithm, and update the individual historical best position of each particle and the global historical best position of the population in each iteration. The task scheduling scheme output module is used to output the bandwidth allocation scheme corresponding to the globally optimal position when the maximum number of iterations is reached, as the optimal transmission task scheduling scheme.

Images