A task evolution-oriented distributed resource dynamic configuration method

Abstract

This invention provides a distributed resource dynamic allocation method oriented towards task evolution, comprising the following steps: S1, constructing a task demand evolution model and recursively calculating the probability distribution of task demand events at each stage based on conditional probability; S2, receiving the probability distribution of task demand events at each stage, constructing a distributed resource allocation model oriented towards task evolution, the distributed resource allocation model taking maximizing the resource system's processing capacity in response to task demands as its primary objective and balancing the task demands at each stage as its secondary objective, establishing a multi-objective nonlinear programming model; S3, inputting the multi-objective nonlinear programming model into a heuristic solution algorithm, using a segmented symbolic encoding multi-strategy genetic algorithm for solution, and outputting the optimal resource allocation scheme. This invention, by combining with multi-stage conditional probability recursive calculation, achieves a quantitative description of the stage-wise evolution law of task demands, providing accurate probabilistic input for dynamic resource allocation.

Description

Technical Field

[0001] This invention relates to the field of resource allocation technology, and more particularly to a distributed dynamic resource allocation method oriented towards task evolution. Background Technology

[0002] In the field of distributed resource management and scheduling, the resource system consists of several different types of resource units. These resource units are distributed across multiple candidate sites or nodes and need to be dynamically configured according to the constantly changing phased task requirements. Task requirements exhibit obvious phased evolution characteristics; that is, as the process progresses, the task type, the number of objectives, and the priority will change. The resource system must have adaptive adjustment capabilities to continuously meet the task requirements of each phase. Typical application scenarios include emergency resource scheduling, multi-phase production task allocation, and distributed computing resource management. In these scenarios, there are functional complementarity or collaborative constraints between resource units, and there is a complex mapping relationship between task requirements and resource capabilities.

[0003] To address the aforementioned resource allocation problem, existing technologies primarily focus on three directions. The first direction involves qualitative research into processing capacity requirements analysis and resource system design processes, such as establishing a capacity requirement decomposition model based on event activity flows, or constructing an integrated design framework from requirement input to resource deployment. The second direction involves establishing mathematical models for resource target allocation problems in specific domains, such as resource allocation models based on fuzzy optimization techniques and target allocation models that consider the advantages of both sides. Most of these models are static and do not consider the phased changes in task requirements. The third direction involves comparing and optimizing the effects of multiple resource allocation schemes, such as scheme optimization methods based on data envelopment analysis. In addition, some studies employ heuristic algorithms such as genetic algorithms to solve resource allocation problems.

[0004] However, existing technologies still have significant shortcomings in handling dynamic resource allocation problems where task requirements evolve in stages. Most models assume that task requirements are fixed and ignore the dynamic evolution of requirement types during actual tasks, making it difficult for allocation schemes to adapt to multi-stage changes. Existing models are mostly single-stage static allocations, unable to perform coordinated optimization of resource capabilities across multiple task stages, easily leading to problems such as excessive resource consumption in the early stages or insufficient capabilities in the later stages. Traditional genetic algorithms, when dealing with combinatorial optimization problems involving multiple resource types and multiple sites, suffer from long encoding lengths, large search spaces, and difficulties in satisfying constraints. These algorithms are prone to getting trapped in local optima, and their solution efficiency and quality fail to meet practical requirements. Therefore, there is an urgent need for a method that can describe the evolution of task requirements, construct multi-stage dynamic allocation models, and solve them efficiently. Summary of the Invention

[0005] In view of this, the purpose of this invention is to propose a distributed resource dynamic configuration method oriented towards task evolution, so as to solve the technical problem that the existing technology ignores the dynamic evolution law of demand types in actual task processes.

[0006] The technical means employed in this invention are as follows:

[0007] A method for dynamic allocation of distributed resources oriented towards task evolution includes the following steps: S1. Construct a task requirement evolution model, divide the task timeline into N stages, define the task requirement event in stage i and the corresponding task requirement type, set the probability of the initial task requirement event, and set the conditional probability of the task requirement type evolving from stage i to stage i+1; recursively calculate the probability distribution of task requirement events in each stage based on the conditional probability. S2 receives the probability distribution of task requirement events at each stage from the output of S1, constructs a distributed resource allocation model oriented towards task evolution, takes maximizing the resource system processing capacity to respond to task requirements as the first objective and balancing the task requirements at each stage as the second objective, and establishes a multi-objective nonlinear programming model. S3. Input the multi-objective nonlinear programming model constructed in S2 into the heuristic solution algorithm, and use the segmented symbolic encoding multi-strategy genetic algorithm to solve it, and output the optimal resource allocation scheme.

[0008] Furthermore, the probability distribution of task requirement events at each stage is calculated recursively based on conditional probability, specifically including: The formula for calculating the conditional probability after the first evolution of task requirements is:

[0009] The formula for calculating the probability of the occurrence of the task requirement event after the first evolution of the task requirement is:

[0010] The formula for calculating the conditional probability after the evolution of the second task requirement is:

[0011] The formula for calculating the probability of the occurrence of the task requirement event after the second evolution of task requirements is:

[0012] The formula for calculating the conditional probability after the Nth task requirement evolution is:

[0013] The formula for calculating the probability of the occurrence of the task requirement event after the Nth evolution of task requirements is:

[0014] in, Indicates the first The task requirements and events of the phase. Indicates the first The types of tasks required at each stage , ; Indicates the number of stages in the evolution of task requirements; Indicates the total number of task requirement types; Indicates the first Phase Task Requirements Events The probability of its occurrence; Indicates the first Phase Task Requirements Events Under the conditions of occurrence, the first Phase Task Requirements Events The conditional probability of occurrence; The task requirement type is indicated by the first stage Evolved into the first stage The conditional probability; Indicates the task requirement event in the initial stage. The probability of its occurrence, i.e. ; Indicates the first Summing all possible task requirement types for each stage.

[0015] Furthermore, in S2, the multi-objective nonlinear programming model is as follows:

[0016] in, The value of the first objective function. This is the value of the second objective function; For task direction indexing; For mission direction Number of task objectives; For mission direction The overall response variable; Index for task phases; Total number of task phases; For the first The types of tasks required at each stage; For mission direction In the Phase task requirements The decision variable for whether or not a response is obtained; For the first Phase Task Requirements Events The probability of its occurrence; and Indexed by task requirement type; and For indexing resource units or alternative sites; In order to respond to the mission direction Task requirement types A collection of resource units; Single-function resource unit Is it used to respond to task requirement type? Decision variables; As a multi-functional resource unit Is it used to respond to task requirement type? Decision variables; For single-function resource units, respond to task requirement types Resource consumption at that time; For multi-functional resource units to respond to task requirements types Resource consumption at that time; This sets the total upper limit for resource allocation. For mission direction Task requirement types The decision variable for whether or not a response is obtained; The first objective function is:

[0017] The first objective function is used to maximize the scale of the overall task demand. The second objective function is:

[0018] The second objective function is used to maximize the multi-stage probability-weighted response under the condition of task requirement evolution; The main constraints in a multi-objective nonlinear programming model fall into five categories: The first type of constraint is the task requirement coverage constraint:

[0019] The first type of constraint is used to limit the direction of the task. Task requirement types Only when there is a single-function resource unit or a multi-function resource unit capable of responding, Only then can it be 1; The second type of constraint is the resource consumption constraint:

[0020] The second type of constraint is used to limit the total resource allocation consumption from not exceeding the upper limit of the total resource allocation consumption. ; The third type of constraint is the global response constraint:

[0021] The third type of constraint is used to limit only the task direction. When all kinds of task requirements are responded to, Only then can it be 1; The fourth type of constraint is the resource allocation uniqueness constraint:

[0022] The fourth type of constraint is used to limit each resource unit or alternative site to being configured at most once in the same configuration scheme; The fifth type of constraint is the phased task response constraint:

[0023] The fifth type of constraint is used to describe the probability of occurrence of task-required events. Basic response variables With staged response variables The relationship between them; The final variable value constraints are used to limit... Class variables and All class variables are 0-1 decision variables.

[0024] Furthermore, before constructing the S2 multi-objective nonlinear programming model, a single-objective programming model is constructed, with the following formula:

[0025] The single-objective programming model is used for comparison and verification with the multi-objective nonlinear programming model.

[0026] Furthermore, in S3, the segmented symbol encoding multi-strategy genetic algorithm specifically includes the following sub-steps: S31. An initial population is generated using a segmented symbolic encoding strategy. The chromosome is divided into multiple segments, each segment corresponding to a resource type configuration scheme. Each gene bit represents the deployment flag of the resource for a specific task at a specific site. The segmented symbolic encoding strategy reduces the encoding length from the cubic order of magnitude of traditional binary encoding to the quadratic order of magnitude. S32. Apply the quantity discrimination operator to each chromosome generated in S31, check and repair the resource quantity constraints, so that the deployment quantity of each type of resource in each chromosome does not exceed the maximum available quantity of that type of resource; S33. The chromosome population that meets the quantity constraints output by S32 is used as the initial parent population and enters the evolutionary cycle.

[0027] Furthermore, the evolutionary cycle described in S33 includes the following sub-steps: S331. Calculate the overall fitness of each chromosome in the current population. The overall fitness is determined by the product of the basic fitness and the soft constraint penalty factor. The basic fitness includes the weighted sum of the first objective function value and the second objective function value. The soft constraint penalty factor is calculated based on whether the total consumption exceeds the budget limit. S332. Based on the overall fitness calculated in S331, a dynamic elite retention strategy is adopted to select a number of individuals with the highest fitness and directly retain them to the next generation of the population. S333. For the remaining individuals, the roulette wheel selection method is used to select parent individuals, and the selection probability of the roulette wheel selection is proportional to the overall fitness. S334. Perform multi-segment multi-point crossover operation on the parent individuals selected in S333 with a preset crossover probability. The multi-segment multi-point crossover only performs gene segment exchange at the resource type boundary. After the exchange, apply the quantity discrimination operator to the offspring chromosomes again for constraint repair. S335. Perform mutation operations on the offspring chromosomes generated in S334 with an adaptive mutation rate. The adaptive mutation rate decreases non-linearly with the number of generations. The mutation operations include three types: adding deployments, deleting deployments, and changing deployments. S336. Merge the elite individuals retained in S332 with the offspring individuals produced in S335 to form a new generation of population; S337. Periodically inject randomly generated new individuals into the new generation of the population and replace the same number of individuals with the lowest fitness to maintain population diversity; S338. Repeat S331 to S337 until the preset maximum number of generations or fitness convergence condition is reached.

[0028] Furthermore, the formula for calculating the adaptive variability rate in S334 is as follows:

[0029] in, For the first The adaptive mutation rate of the generation; The initial mutation rate; This is the coefficient for the decrease in the rate of variation; The current generation number; The maximum number of generations; It is a natural exponential function.

[0030] The minimum value of the adaptive variability rate is not lower than a preset lower limit, that is:

[0031] in, This is the preset minimum variation rate.

[0032] Furthermore, the calculation method for the soft constraint penalty factor in S331 is as follows: When total consumption is less than or equal to the budget limit, the penalty factor is greater than 1 and increases with the increase of the remaining budget percentage; when total consumption is greater than the budget limit, the penalty factor is less than 1 and decreases with the increase of the overspending percentage.

[0033] Furthermore, in S32, the quantity discrimination operator performs the following operations: For each resource type, count the actual number of resources deployed in the current chromosome; if the actual number of deployments exceeds the maximum limit for that resource type, randomly select the excess deployment gene positions and set their values ​​to zero, indicating that the deployment is canceled.

[0034] Compared with the prior art, the present invention has the following advantages: The task requirement evolution model provided by this invention, by combining it with multi-stage conditional probability recursive calculation, achieves a quantitative description of the stage-by-stage evolution law of task requirements, and provides accurate probabilistic input for dynamic resource allocation.

[0035] The multi-objective nonlinear programming model for task evolution provided by this invention achieves a coordinated balance of task requirements across multiple stages by combining it with probability weighting of task requirement events at each stage and dual-objective optimization, thus avoiding resource mismatch problems caused by single-stage static configuration.

[0036] The segmented symbol encoding strategy provided by this invention, by combining it with resource type boundary segmentation and symbolic gene expression, reduces the encoding complexity from the cubic order of magnitude to the quadratic order of magnitude, thus significantly compressing the search space.

[0037] The quantity discrimination operator provided by this invention, by combining it with real-time constraint checking and repair after chromosome generation and crossover mutation, ensures that all resource allocation schemes strictly meet the hard constraints on the quantity of resources, thereby improving the feasibility of the solution.

[0038] The multi-segment, multi-point crossover and adaptive mutation strategy provided by this invention, by combining the selection of crossover points at resource type boundaries with a mutation rate that decays nonlinearly with the number of generations, achieves a dynamic balance between global exploration and local development, effectively avoiding premature convergence. Attached Figure Description

[0039] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0040] Figure 1 This is a flowchart of the method of the present invention.

[0041] Figure 2 This is a model diagram illustrating the evolution of the task requirements of this invention.

[0042] Figure 3 This is the S3 flowchart of the present invention. Detailed Implementation

[0043] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0044] 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 orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0045] like Figure 1 As shown, this invention provides a distributed dynamic resource allocation method oriented towards task evolution. This method describes the probabilistic transition process of multi-stage task requirements by constructing a task requirement evolution model. Based on this, a multi-objective nonlinear programming model is established, and a piecewise symbolic encoding multi-strategy genetic algorithm is used for efficient solution. Finally, the optimal resource allocation scheme that satisfies the task requirements at each stage is output. Specifically, it includes the following steps: S1. Construct a task requirement evolution model, divide the task timeline into N stages, define the task requirement event in stage i and the corresponding task requirement type, set the probability of the initial task requirement event, and set the conditional probability of the task requirement type evolving from stage i to stage i+1; recursively calculate the probability distribution of task requirement events in each stage based on the conditional probability. S2 receives the probability distribution of task requirement events at each stage from the output of S1, constructs a distributed resource allocation model oriented towards task evolution, takes maximizing the resource system processing capacity to respond to task requirements as the first objective and balancing the task requirements at each stage as the second objective, and establishes a multi-objective nonlinear programming model. S3. Input the multi-objective nonlinear programming model constructed in S2 into the heuristic solution algorithm, and use the segmented symbolic encoding multi-strategy genetic algorithm to solve it, and output the optimal resource allocation scheme.

[0046] S31. An initial population is generated using a segmented symbolic encoding strategy. The chromosome is divided into multiple segments, each segment corresponding to a resource type configuration scheme. Each gene bit represents the deployment flag of the resource for a specific task at a specific site. The segmented symbolic encoding strategy reduces the encoding length from the cubic order of magnitude of traditional binary encoding to the quadratic order of magnitude. S32. Apply the quantity discrimination operator to each chromosome generated in S31, check and repair the resource quantity constraints, so that the deployment quantity of each type of resource in each chromosome does not exceed the maximum available quantity of that type of resource; For each resource type, count the actual number of resources deployed in the current chromosome; if the actual number of deployments exceeds the maximum limit for that resource type, randomly select the excess deployment gene positions and set their values ​​to zero, indicating that the deployment is canceled.

[0047] S33. The chromosome population that meets the quantity constraints output by S32 is used as the initial parent population and enters the evolutionary cycle.

[0048] S331. Calculate the overall fitness of each chromosome in the current population. The overall fitness is determined by the product of the basic fitness and the soft constraint penalty factor. The basic fitness includes the weighted sum of the first objective function value and the second objective function value. The soft constraint penalty factor is calculated based on whether the total consumption exceeds the budget limit. The calculation method for the soft constraint penalty factor is as follows: When total consumption is less than or equal to the budget limit, the penalty factor is greater than 1 and increases with the increase of the remaining budget percentage; when total consumption is greater than the budget limit, the penalty factor is less than 1 and decreases with the increase of the overspending percentage.

[0049] S332. Based on the overall fitness calculated in S331, a dynamic elite retention strategy is adopted to select a number of individuals with the highest fitness and directly retain them to the next generation of the population. S333. For the remaining individuals, the roulette wheel selection method is used to select parent individuals, and the selection probability of the roulette wheel selection is proportional to the overall fitness. S334. Perform multi-segment multi-point crossover operation on the parent individuals selected in S333 with a preset crossover probability. The multi-segment multi-point crossover only performs gene segment exchange at the resource type boundary. After the exchange, apply the quantity discrimination operator to the offspring chromosomes again for constraint repair. S335. Perform mutation operations on the offspring chromosomes generated in S334 with an adaptive mutation rate. The adaptive mutation rate decreases non-linearly with the number of generations. The mutation operations include three types: adding deployments, deleting deployments, and changing deployments. S336. Merge the elite individuals retained in S332 with the offspring individuals produced in S335 to form a new generation of population; S337. Periodically inject randomly generated new individuals into the new generation of the population and replace the same number of individuals with the lowest fitness to maintain population diversity; S338. Repeat S331 to S337 until the preset maximum number of generations or fitness convergence condition is reached.

[0050] The specific steps for S1 are as follows: In the resource allocation and management for completing specified tasks, task requirements are constantly changing and evolve according to different task stages. To describe this evolution in detail, this patent divides the entire timeline of different tasks into... Different stages, such as Figure 2 As shown in the diagram, the phased process can be understood as follows: the task requirements of each phase are the evolution of the task requirements of the previous phase.

[0051] Assumptions in the diagram Indicates the first Task-required events during the task phase, among which Indicates the first Task requirement types at each task phase. Assume the probability of the initial task requirement occurring is defined as... From the first Phase Task Requirement Types Evolved to the Phase Task Requirement Types The conditional probability is .

[0052] like Figure 2As shown, the evolution of task requirements can be visually represented by a directed graph, where each node represents a task requirement event, and the directed edges between task requirement nodes represent the probability of one task requirement state evolving into another. This probabilistic evolution of task requirements can be clearly described using the following series of formulas.

[0053] The conditional probability and the probability of the occurrence of the task requirement event after the first evolution of the task requirement are expressed as follows: (1) (2) The conditional probability and the probability of the occurrence of the task requirement event after the second evolution of the task requirement are expressed as follows: (3) (4) And so on, until the... The conditional probability and the probability of occurrence of the task requirement event after the evolution of the sub-task requirement are respectively expressed as follows: (5) (6) In resource allocation activities oriented towards specified tasks, the efficiency of the resource system's processing capacity depends not only on the type of task requirements but also on the actual distance between the task objective and the resource unit, the priority level of the task objective, and the functional type and configuration of the resource unit. Different task objective stages and different task requirements may require different resource system processing capabilities to correspond to them. Especially in the context of multi-functional cooperation, the requirements may need to be responded to by resource units with different processing capability types or levels. To ensure the rationality and fairness of resource allocation, it is crucial to evaluate the probability distribution of all possible task requirement events. Therefore, incorporating the probability of task requirement events into the model in this invention is also an important consideration.

[0054] The specific steps for S2 are as follows: To ensure timely and effective response to task objectives of varying priorities when completing assigned tasks, it is essential to consider how to configure resource systems across different functional types. This resource system configuration involves network collaboration among alternative task sites and potential task requirements. This section first defines relevant configuration parameters and decision variables, and then establishes resource system configuration models across different functional types and resource system configuration models oriented towards task evolution.

[0055] To simplify the calculation process, this invention uses Euclidean distance as a linear approximation to represent the resource unit behaviors of different priority levels of task objectives and their corresponding responses. Two assumptions are made: first, due to economic effects, the total resource unit usage cost for multi-functional resource units is lower than or equal to the total resource unit usage cost for single-functional resource units; second, the processing capacity of resource units configured at task sites meets the task requirements of the task objectives entering that task site. Relevant configuration parameters and their meanings are explained in Table 1.

[0056] Table 1. Explanation of Relevant Configuration Parameters and Their Meanings

[0057] To respond to specified task requirements, resources across different functional types are configured with the aim of maximizing task requirements with minimal resource allocation. Therefore, a single-objective programming model for resource allocation across different functional types is constructed with the objective of maximizing the resource processing capacity to respond to task requirements.

[0058] (7) The single-objective function of this mathematical programming model satisfies the task requirements of all task objectives. The first constraint ensures the type of task requirement. The task objective can be responded to by multifunctional resource units or single-function resource units, and also reflects the coordinated configuration of multifunctional resource units or single-function resource units; the second constraint sets an upper limit on the total consumption of resource system resource allocation for the task; the third constraint states that the overall demand is considered to be satisfied only when all specific demands of the task objective are met by the responses of resource units; the fourth constraint stipulates that only one resource unit can be configured for each alternative task target; the last three constraints are about the definition of model decision variables.

[0059] Taking into account the evolution of task requirements, the above model introduces the probability of requirement events. To better respond to changing demands, in addition to maximizing the resource processing capacity of the resource system to meet task requirements, we also balance the task requirements at each stage with equal importance, and construct a resource allocation model for resource systems across different functional types that is oriented towards task evolution, as follows.

[0060] (8) in, The value of the first objective function. This is the value of the second objective function; For task direction indexing; For mission direction Number of task objectives; For mission direction The overall response variable; Index for task phases; Total number of task phases; For the first The types of tasks required at each stage; For mission direction In the Phase task requirements The decision variable for whether or not a response is obtained; For the first Phase Task Requirements Events The probability of its occurrence; and Indexed by task requirement type; and For indexing resource units or alternative sites; In order to respond to the mission direction Task requirement types A collection of resource units; Single-function resource unit Is it used to respond to task requirement type? Decision variables; As a multi-functional resource unit Is it used to respond to task requirement type? Decision variables; For single-function resource units, respond to task requirement types Resource consumption at that time; For multi-functional resource units to respond to task requirements types Resource consumption at that time; This sets the total upper limit for resource allocation. For mission direction Task requirement types The decision variable for whether or not a response is obtained; This is a multi-objective nonlinear programming model, and the relative importance of the two objective functions needs to be considered when solving it. Compared with the aforementioned single-objective programming model, the newly added fifth constraint sets that the task objective and task requirements must be met in each task stage. In other words, the task objective and task requirement event must occur and there must be a resource unit that can provide the response for the task objective and task requirements to be considered met. The newly added final constraint is also about the definition of the model decision variables, setting the standard for the requirements in each task stage.

[0061] The first objective function is:

[0062] The first objective function is used to maximize the scale of the overall task demand. The second objective function is:

[0063] The second objective function is used to maximize the multi-stage probability-weighted response under the condition of task requirement evolution; The main constraints in a multi-objective nonlinear programming model fall into five categories: The first type of constraint is the task requirement coverage constraint:

[0064] The first type of constraint is used to limit the direction of the task. Task requirement types Only when there is a single-function resource unit or a multi-function resource unit capable of responding, Only then can it be 1; The second type of constraint is the resource consumption constraint:

[0065] The second type of constraint is used to limit the total resource allocation consumption from not exceeding the upper limit of the total resource allocation consumption. ; The third type of constraint is the global response constraint:

[0066] The third type of constraint is used to limit only the task direction. When all kinds of task requirements are responded to, Only then can it be 1; The fourth type of constraint is the resource allocation uniqueness constraint:

[0067] The fourth type of constraint is used to limit each resource unit or alternative site to being configured at most once in the same configuration scheme; The fifth type of constraint is the phased task response constraint:

[0068] The fifth type of constraint is used to describe the probability of occurrence of task-required events. Basic response variables With staged response variables The relationship between them; The final variable value constraints are used to limit... Class variables and All class variables are 0-1 decision variables.

[0069] like Figure 3 As shown, S3 specifically includes the following steps: PSCMSGA (Segmented Symbolic Encoding Multi-Strategy Genetic Algorithm) is a genetic algorithm framework designed based on traditional genetic algorithms to address the special complexity of resource allocation problems in resource systems. It comprehensively adopts segmented symbolic encoding strategy, quantity discrimination operator, multi-segment multi-point crossover mechanism, and adaptive mutation strategy.

[0070] 1. Segmented Symbolic Encoding Strategy: The chromosome is divided into multiple segments, each corresponding to a resource type configuration scheme. Symbolic encoding is used instead of binary encoding to reduce the encoding length. Each gene bit represents the deployment of resources for a specific task at a specific site. The traditional binary encoding length is R × J × H, while the PSCMSGA encoding length is ∑∑ Max_r^h, reducing the encoding complexity from O(n³) to O(n²).

[0071] 2. Quantity discrimination operator: A constraint checking mechanism is introduced during the encoding stage, and a quantity scheme vector is defined. This ensures that resource quantity constraints are met when chromosomes are generated.

[0072] 3. Multi-segment, multi-point crossover mechanism: Crossover occurs at resource type boundaries, rather than at random locations, to maintain the structural integrity of the solution. After crossover, constraints are automatically repaired to maintain the integrity of superior gene segments. The proportion of feasible solutions increases after crossover, accelerating the spread of superior genes.

[0073] 4. Adaptive mutation strategy: The mutation rate is dynamically adjusted according to the evolution stage, and various mutation operations such as adding, deleting, and changing deployment are performed. After mutation, the constraints are forced to be met, balancing exploration and development capabilities, avoiding premature convergence, and improving global search capabilities.

[0074] 5. Dual constraint handling mechanism: A penalty function is used, and a repair mechanism is used for hard constraints (resource quantity, processing capacity). Different types of constraints are handled in layers to ensure that all solutions satisfy the hard constraints, guide the search to the budget feasible region, and improve the practicality of the algorithm.

[0075] 6. Elite retention and diversity maintenance: Dynamic elite retention ratio, roulette wheel selection based on fitness ranking, periodic injection of random individuals to maintain diversity, prevent loss of superior genes, maintain population diversity, and avoid premature convergence.

[0076] Based on the above six components, the specific steps of the PSCMSGA algorithm are as follows: Step 1: Initialize parameters; Set the population size N, maximum number of generations G, crossover probability Pc, initial mutation rate Pm0, elite retention ratio e, random injection ratio r, etc.

[0077] Set adaptive mutation parameters: mutation rate reduction coefficient α, minimum mutation rate Pm_min.

[0078] Step 2: Initialize the population; For each individual in the population, a segmented symbolic encoding strategy is used to generate chromosomes: For each resource type r (there are R types in total): Randomly generate the number of deployments d_r (between 0 and the number of resources of this type).

[0079] Randomly select d_r sites (from J sites).

[0080] For each selected site, a task type is randomly assigned (randomly selected from the tasks that the resource can perform).

[0081] The quantity discrimination operator is applied to ensure that each individual satisfies the resource quantity constraint (this step has already been satisfied during the generation process, because this patent is generated by quantity).

[0082] Calculate the fitness of each individual (considering the objective function and budget constraint penalties).

[0083] Step 3: Evolutionary Cycle; For each generation g (from 1 to G): 1. Select operation; Ne individuals (excluding elite individuals) are selected from the current population using the roulette wheel selection method.

[0084] At the same time, the e individuals with the highest fitness from the current population are selected as elites and directly preserved for the next generation.

[0085] 2. Cross-operation; For the selected Ne individuals, pair them up randomly (if the number is odd, add one more randomly), forming a total of (Ne) / 2 pairs.

[0086] For each pair of parent individuals, crossover is performed with probability Pc: Randomly select several resource type boundary points (for example, randomly select k points from R-1 boundary points, where k is usually 1 or 2).

[0087] Swap the entire gene sequence between these boundary points between the two parent individuals (i.e., swap the configuration vectors of the corresponding resource types).

[0088] The two offspring individuals after crossover are repaired using a quantity discrimination operator to ensure that the resource quantity constraint is met.

[0089] If no crossover occurs, the offspring replicates the parent.

[0090] 3. Mutation operation; Calculate the adaptive mutation rate of the current generation: Pm = Pm0 exp(-α (g / G), and Pm is not lower than Pm_min.

[0091] For the offspring individuals generated by crossover (a total of Ne individuals), mutation is performed with probability Pm: Randomly select the mutation type: add deployment, delete deployment, or change deployment.

[0092] Perform the selected mutation operation (see the mutation strategy above for details).

[0093] Apply quantity discrimination operators to ensure that resource quantity constraints are met.

[0094] Note: Elite individuals do not participate in crossover and mutation and are directly retained.

[0095] 4. Formation of a new generation of populations; The elite individuals (e individuals) and the offspring individuals (Ne individuals) produced through crossover mutation are merged to form a new generation of population.

[0096] 5. Maintain diversity in operations; RN new individuals are randomly generated (using the initialization method in step 2), and the rN individuals with the worst fitness in the new generation population are replaced (Note: Elite individuals may be replaced, but this patent usually protects elite individuals from being replaced, so in actual operation, only the worst individual among non-elite individuals is replaced).

[0097] 6. Fitness calculation; Calculate the fitness of each individual in the new generation of the population (recalculation is required for newly generated individuals and individuals that have changed; the fitness of individuals that have not changed remains unchanged).

[0098] Step 4: Terminate the inspection; If the maximum number of generations G is reached, evolution stops, and the individual with the highest fitness is output as the optimal solution; otherwise, return to step 3.

[0099] Step 5: Output the optimal solution; Output the chromosome code of the globally optimal solution (i.e., the individual with the highest fitness throughout the entire evolutionary process), and decode it to obtain the specific resource deployment plan.

[0100] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for dynamic allocation of distributed resources oriented towards task evolution, characterized in that, Includes the following steps: S1. Construct a task requirement evolution model, divide the task timeline into N stages, define the task requirement event in stage i and the corresponding task requirement type, set the probability of the initial task requirement event, and set the conditional probability of the task requirement type evolving from stage i to stage i+1; recursively calculate the probability distribution of task requirement events in each stage based on the conditional probability. S2 receives the probability distribution of task requirement events at each stage from the output of S1, constructs a distributed resource allocation model oriented towards task evolution, takes maximizing the resource system processing capacity to respond to task requirements as the first objective and balancing the task requirements at each stage as the second objective, and establishes a multi-objective nonlinear programming model. S3. Input the multi-objective nonlinear programming model constructed in S2 into the heuristic solution algorithm, and use the segmented symbolic encoding multi-strategy genetic algorithm to solve it, and output the optimal resource allocation scheme.

2. The distributed resource dynamic allocation method oriented towards task evolution according to claim 1, characterized in that, The probability distribution of task requirement events at each stage is calculated recursively based on conditional probability, specifically including: The formula for calculating the conditional probability after the first evolution of task requirements is: The formula for calculating the probability of the occurrence of the task requirement event after the first evolution of the task requirement is: The formula for calculating the conditional probability after the evolution of the second task requirement is: The formula for calculating the probability of the occurrence of the task requirement event after the second evolution of task requirements is: The formula for calculating the conditional probability after the Nth task requirement evolution is: The formula for calculating the probability of the occurrence of the task requirement event after the Nth evolution of task requirements is: in, Indicates the first The task requirements and events of the phase. Indicates the first The types of tasks required at each stage , ; Indicates the number of stages in the evolution of task requirements; Indicates the total number of task requirement types; Indicates the first Phase Task Requirements Events The probability of its occurrence; Indicates the first Phase Task Requirements Events Under the conditions of occurrence, the first Phase Task Requirements Events The conditional probability of occurrence; The task requirement type is indicated by the first stage Evolved into the first stage The conditional probability; Indicates the task requirement event in the initial stage. The probability of its occurrence, i.e. ; Indicates the first Summing all possible task requirement types for each stage.

3. The distributed resource dynamic allocation method oriented towards task evolution according to claim 1, characterized in that, In S2, the multi-objective nonlinear programming model is as follows: in, The value of the first objective function. This is the value of the second objective function; For task direction indexing; For mission direction Number of task objectives; For mission direction The overall response variable; Index for task phases; Total number of task phases; For the first The types of tasks required at each stage; For mission direction In the Phase Task Requirements The decision variable for whether or not a response is obtained; For the first Phase Task Requirements Events The probability of its occurrence; and Indexed by task requirement type; and For indexing resource units or alternative sites; In order to respond to the mission direction Task requirement types A collection of resource units; Single-function resource unit Is it used to respond to task requirement type? Decision variables; As a multi-functional resource unit Is it used to respond to task requirement type? Decision variables; For single-function resource units, respond to task requirement types Resource consumption at that time; For multi-functional resource units, respond to task requirements types Resource consumption at that time; This sets the total upper limit for resource allocation. For mission direction Task requirement types The decision variable for whether or not a response is obtained; The first objective function is: The first objective function is used to maximize the scale of the overall task demand. The second objective function is: The second objective function is used to maximize the multi-stage probability-weighted response under the condition of task requirement evolution; The main constraints in a multi-objective nonlinear programming model fall into five categories: The first type of constraint is the task requirement coverage constraint: The first type of constraint is used to limit the direction of the task. Task requirement types Only when there is a single-function resource unit or a multi-function resource unit capable of responding, Only then can it be 1; The second type of constraint is the resource consumption constraint: The second type of constraint is used to limit the total resource allocation consumption from not exceeding the upper limit of the total resource allocation consumption. ; The third type of constraint is the global response constraint: The third type of constraint is used to limit only the task direction. When all kinds of task requirements are responded to, Only then can it be 1; The fourth type of constraint is the resource allocation uniqueness constraint: The fourth type of constraint is used to limit each resource unit or alternative site to being configured at most once in the same configuration scheme; The fifth type of constraint is the phased task response constraint: The fifth type of constraint is used to describe the probability of occurrence of task-required events. Basic response variables With staged response variables The relationship between them; The final variable value constraints are used to limit... Class variables and All class variables are 0-1 decision variables.

4. The distributed resource dynamic allocation method oriented towards task evolution according to claim 1, characterized in that, Before constructing the S2 multi-objective nonlinear programming model, a single-objective programming model is constructed, as shown in the following formula: The single-objective programming model is used for comparison and verification with the multi-objective nonlinear programming model.

5. The distributed resource dynamic allocation method oriented towards task evolution according to claim 1, characterized in that, In S3, the segmented symbol encoding multi-strategy genetic algorithm specifically includes the following sub-steps: S31. An initial population is generated using a segmented symbolic encoding strategy. The chromosome is divided into multiple segments, each segment corresponding to a resource type configuration scheme. Each gene bit represents the deployment flag of the resource for a specific task at a specific site. The segmented symbolic encoding strategy reduces the encoding length from the cubic order of magnitude of traditional binary encoding to the quadratic order of magnitude. S32. Apply the quantity discrimination operator to each chromosome generated in S31, check and repair the resource quantity constraints, so that the deployment quantity of each type of resource in each chromosome does not exceed the maximum available quantity of that type of resource; S33. The chromosome population that meets the quantity constraints output by S32 is used as the initial parent population and enters the evolutionary cycle.

6. The distributed resource dynamic allocation method oriented towards task evolution according to claim 5, characterized in that, The evolutionary cycle described in S33 includes the following sub-steps: S331. Calculate the overall fitness of each chromosome in the current population. The overall fitness is determined by the product of the basic fitness and the soft constraint penalty factor. The basic fitness includes the weighted sum of the first objective function value and the second objective function value. The soft constraint penalty factor is calculated based on whether the total consumption exceeds the budget limit. S332. Based on the overall fitness calculated in S331, a dynamic elite retention strategy is adopted to select a number of individuals with the highest fitness and directly retain them to the next generation of the population. S333. For the remaining individuals, the roulette wheel selection method is used to select parent individuals, and the selection probability of the roulette wheel selection is proportional to the overall fitness. S334. Perform multi-segment multi-point crossover operation on the parent individuals selected in S333 with a preset crossover probability. The multi-segment multi-point crossover only performs gene segment exchange at the resource type boundary. After the exchange, apply the quantity discrimination operator to the offspring chromosomes again for constraint repair. S335. Perform mutation operations on the offspring chromosomes generated in S334 with an adaptive mutation rate. The adaptive mutation rate decreases non-linearly with the number of generations. The mutation operations include three types: adding deployments, deleting deployments, and changing deployments. S336. Merge the elite individuals retained in S332 with the offspring individuals generated in S335 to form a new generation of population; S337. Periodically inject randomly generated new individuals into the new generation of the population and replace the same number of individuals with the lowest fitness to maintain population diversity; S338. Repeat S331 to S337 until the preset maximum number of generations or fitness convergence condition is reached.

7. The distributed resource dynamic allocation method oriented towards task evolution according to claim 6, characterized in that, The formula for calculating the adaptive mutation rate in S334 is: in, For the first The adaptive mutation rate of the generation; The initial mutation rate; This is the coefficient for the decrease in the rate of variation; The current generation number; The maximum number of generations; It is a natural exponential function; The minimum value of the adaptive variability rate is not lower than a preset lower limit, that is: in, This is the preset minimum variation rate.

8. The distributed resource dynamic allocation method oriented towards task evolution according to claim 1, characterized in that, The calculation method for the soft constraint penalty factor in S331 is as follows: When total consumption is less than or equal to the budget limit, the penalty factor is greater than 1 and increases with the increase of the remaining budget percentage; when total consumption is greater than the budget limit, the penalty factor is less than 1 and decreases with the increase of the overspending percentage.

9. The distributed resource dynamic allocation method oriented towards task evolution according to claim 1, characterized in that, In S32, the quantity discrimination operator performs the following operations: For each resource type, count the actual number of resources deployed in the current chromosome; if the actual number of deployments exceeds the maximum limit for that resource type, randomly select the excess deployment gene positions and set their values ​​to zero, indicating that the deployment is canceled.

Images