A mission planning and control system based on digital twin

Abstract

The present invention relates to the technical field of task planning and management, and discloses a task planning and control system based on digital twins, comprising: a project information collection module, for collecting project information of M implementation projects; a unit information collection module, for collecting unit information of N implementation units; a graph network data construction module, for constructing graph network data based on the project information of the M implementation projects and the unit information of the N implementation units; an allocation scheme generation module, for constructing an objective function according to a trained prediction model, and for generating an allocation scheme through an Arctic puffin optimization algorithm; the present invention constructs the project information of the implementation project and the unit information of the implementation unit into graph network data, performs information aggregation on the graph network data through a prediction model to obtain an estimated completion time and an estimated cost of the project, and constructs an objective function based on the prediction model, with the goal of minimizing the completion time and cost, and completes the allocation of the implementation project through the Arctic puffin optimization algorithm.

Description

Technical Field

[0001] The present invention relates to the technical field of mission planning and management, and more specifically, to a mission planning and control system based on digital twins. Background Art

[0002] Task planning management plays a core role in software engineering, manufacturing industry, service industry and other fields. When faced with multiple tasks being assigned to multiple teams or resources at the same time, reasonable task planning is particularly important. Existing task planning management systems usually rely on manual decision-making, that is, project managers or managers complete the task allocation based on their own work experience. However, this method is affected by human subjective factors and may only guarantee the completion time of the task, but cannot guarantee the maximization of benefits. In addition, when the number of tasks and teams is very large, manual decision-making is inefficient, resulting in delays in generating allocation plans and affecting the progress of tasks.

[0003] Therefore, a task planning and management system is urgently needed to solve the above technical problems. Summary of the invention

[0004] The present invention provides a task planning and control system based on digital twins, which solves the technical problem that traditional task planning and management systems in related technologies usually rely on manual decision-making, are affected by human subjective factors, and may not be able to ensure maximum benefits.

[0005] The present invention provides a task planning and control system based on digital twins, comprising:

[0006] A project information collection module, which is used to collect project information of M implementation projects;

[0007] Where M is a custom parameter;

[0008] A unit information collection module, which is used to collect unit information of N implementation units;

[0009] Where N is a custom parameter;

[0010] A graph network data construction module, which is used to construct graph network data based on project information of M implementation projects and unit information of N implementation units;

[0011] Graph network data includes: vertices, initial features of vertices, and edges between vertices;

[0012] The vertices include: a first vertex and a second vertex, the first vertex establishes data mapping with the implementation project, and the second vertex establishes data mapping with the implementation unit; the initial feature of the first vertex is represented by vectorizing the project information of the implementation project mapped by the first vertex; the initial feature of the second vertex is represented by vectorizing the unit information of the implementation unit mapped by the second vertex;

[0013] An edge is constructed between the first vertex and the second vertex to indicate that the implementation item mapped by the first vertex is allocated to the implementation unit mapped by the second vertex;

[0014] The implementation project mapped by the data of the same first vertex can only be assigned to one implementation unit mapped by the data of the second vertex at the same time;

[0015] The implementation unit of the data mapping of the same second vertex can only be assigned Q implementation projects of the data mapping of the first vertex at most at the same time, where Q is a custom parameter.

[0016] A prediction model building module, which is used to build and train a prediction model;

[0017] The input of the prediction model is graph network data, and the output values ​​represent the estimated completion time and estimated cost of the project respectively;

[0018] An allocation plan generation module is used to construct an objective function based on the trained prediction model and generate an allocation plan through the Arctic Puffin Optimization Algorithm;

[0019] The allocation plan execution module is used to allocate M implementation projects to N implementation units according to the allocation plan.

[0020] Furthermore, the initial feature of the first vertex is expressed as:

[0021] ;

[0022] in They respectively represent the project implementation area, project cost budget, project completion days, implementation personnel budget, implementation equipment budget and estimated implementation difficulty of the implementation project. The estimated implementation difficulty includes: high, medium and low, and is represented by real number coding.

[0023] Furthermore, the initial feature of the second vertex is expressed as:

[0024] ;

[0025] in They respectively represent the number of implementation personnel, the number of implementation equipment, the annual turnover, the number of historically completed implementation projects, the number of historical implementation accidents and the construction qualification level of the implementation unit. The construction qualification levels include: special grade, first grade, second grade and third grade, and are represented by real number codes.

[0026] Further, the prediction model includes: a first hidden layer, a second hidden layer, an extraction layer, a first classifier, and a second classifier;

[0027] The first hidden layer inputs the initial features of the vertex and outputs the updated features;

[0028] The second hidden layer inputs the graph network data and outputs the graph update matrix;

[0029] The graph update matrix includes M row vectors, each row vector corresponding to the comprehensive features of the edges between the M first vertices and the second vertices;

[0030] The extraction layer is used to extract M row vectors of the graph update matrix and input them into the first classifier and the second classifier;

[0031] The classification space of the first classifier represents the estimated completion time of the project;

[0032] The classification space of the second classifier represents the estimated cost of the project.

[0033] Furthermore, the calculation formula of the first hidden layer is as follows:

[0034] ;

[0035] Where H represents the updated features of the first hidden output, init represents the initial features of the vertices input to the first hidden layer, represents the first weight parameter, represents the first bias parameter, T represents the transposition operation, and softmax represents the softmax activation function.

[0036] Furthermore, the calculation formula of the second hidden layer includes:

[0037] ;

[0038] Where 1≤m≤M, 1≤n≤N, represents the comprehensive characteristics of the edge between the mth first vertex and the nth second vertex, represents the updated features of the mth first vertex, represents the updated features of the nth second vertex, represents the second weight parameter, Represents the second bias parameter, concat represents the concatenation operation, and sigmoid represents the sigmoid activation function.

[0039] Furthermore, sample labels of training samples used to train the prediction model are obtained by taking the average value of results simultaneously labeled by multiple managers, and a cross entropy loss function is specified as the loss function of the prediction model during the training process of the prediction model.

[0040] Further, the allocation scheme is generated by the Arctic Puffin Optimization Algorithm, including the following steps:

[0041] Step S201, randomly generate an initialization population with U individuals that meets the constraint conditions, each individual in the initialization population is represented by a vector code, and define the current iteration number t as 0 and the maximum iteration number T as V, where U and V are both custom parameters;

[0042] The constraints include: the same implementation project can only be assigned to one implementation unit at a time; the same implementation unit can only be assigned to Q implementation projects at most at the same time;

[0043] The length of the vector code of an individual is M, and the mth dimension value of the vector code represents the number of the implementation unit to which the mth implementation project is assigned, where 1≤m≤M;

[0044] Step S202, constructing an objective function according to the trained prediction model;

[0045] The objective function is calculated as follows:

[0046] ;

[0047] Where Fit represents the fitness value of the individual, It indicates the estimated completion time of the mth implementation project assigned to the implementation unit. represents the estimated cost of the mth implementation project allocated to the implementation unit, and denote the first weight coefficient and the second weight coefficient respectively, and All are custom parameters, and their sum value is 1;

[0048] Step S203, generating a behavior conversion factor according to the current number of iterations t and the maximum number of iterations T, and judging that the behavior conversion factor is greater than a preset threshold, then proceeding to step S204, otherwise proceeding to step S205, wherein the preset threshold is a custom parameter;

[0049] The calculation formula of behavior conversion factor Behavior is as follows:

[0050] ;

[0051] Where rand represents a random number ranging from 0 to 1, excluding 0 and 1;

[0052] Step S204, updating the vector codes of all individuals in the initialized population through an air flight strategy;

[0053] Aerial flight strategies include: aerial search strategy, raid predation strategy and merging strategy;

[0054] The aerial search strategy is used to update the vector codes of all individuals in the initialization population to the first vector code;

[0055] The raid predation strategy is used to update the first vector encoding of all individuals in the initialization population to the second vector encoding;

[0056] The merging strategy is used to merge the first vector codes and the second vector codes of all individuals in the initialization population, calculate the fitness value of each individual in the initialization population through the objective function, and sort them in descending order of fitness values ​​to obtain the first U individuals as the initialization population;

[0057] Step S205, updating the vector codes of all individuals in the initialized population through an underwater foraging strategy;

[0058] Underwater foraging strategies include: collective foraging strategy, enhanced search strategy, predator avoidance strategy and merging strategy;

[0059] The collective foraging strategy is used to update the vector encoding of all individuals in the initialization population to the third vector encoding;

[0060] The enhanced search strategy is used to update the third vector encoding of all individuals in the initialization population to the fourth vector encoding;

[0061] The predator avoidance strategy is used to update the vector codes of all individuals in the initialization population to the fifth vector code;

[0062] The merging strategy is used to merge the third vector code, the fourth vector code and the fifth vector code of all individuals in the initialization population, calculate the fitness value of each individual in the initialization population through the objective function, and sort them in descending order of fitness value to obtain the first U individuals as the initialization population;

[0063] Step S206, the current iteration number t is increased by 1. When it is determined that the current iteration number t is greater than or equal to the maximum iteration number T, the individual with the largest fitness value in the initialized population is used as the allocation scheme, otherwise, the process returns to step S203 to continue execution.

[0064] Furthermore, the calculation formula of the air search strategy includes:

[0065] ;

[0066] ;

[0067] Among them, 1≤i≤U, represents the vector encoding of the i-th individual with the current iteration number t, Represents the vector encoding of a random individual with the current iteration number t, represents the first vector code of the i-th individual with the current iteration number t+1, represents the random number generated by Levy flight, D represents the number of dimensions of the individual vector encoding, D is assigned the value M, R represents the air flight factor, rand represents a random number ranging from 0 to 1, excluding 0 and 1, and round represents rounding the value in the brackets, that is, if the value in the brackets is less than 0.5, it is assigned to 0, otherwise it is assigned to 1. Represents random numbers that conform to the standard normal distribution;

[0068] The calculation formula for the predation strategy includes:

[0069] ;

[0070] ;

[0071] in represents the first vector code of the i-th individual with the current iteration number t, It represents the second vector encoding of the i-th individual with the current iteration number t+1, S represents the predation factor, and tanh represents the hyperbolic tangent activation function.

[0072] Furthermore, the calculation formula of the collective foraging strategy is as follows:

[0073] ;

[0074] Where 1≤i≤U, r1, r2 and r3 are all random positive integers between 1 and U-1, excluding i. , and They represent the vector codes of individuals r1, r2 and r3 at the current iteration number t, respectively. represents the third vector code of the i-th individual with the current iteration number t+1, F represents the cooperation factor, and F is assigned a value of 0.5. represents the random number generated by Levy flight, D represents the number of dimensions of the individual vector encoding, D is assigned a value of M, and rand represents a random number ranging from 0 to 1;

[0075] The calculation formula for strengthening the search strategy includes:

[0076] ;

[0077] ;

[0078] in represents the fourth vector code of the i-th individual with the current iteration number t+1, and f represents the search factor;

[0079] The calculation formula for the predator avoidance strategy is as follows:

[0080] ;

[0081] in The fifth vector code of the i-th individual with the current iteration number t+1 is, Represents a random number uniformly distributed between 0 and 1.

[0082] The beneficial effects of the present invention are as follows: the present invention constructs the project information of the implementation project and the unit information of the implementation unit into graph network data, aggregates the graph network data through a prediction model to obtain the estimated project completion time and the estimated project cost, and constructs an objective function based on the prediction model, with the goal of minimizing the completion time and cost, and completes the allocation of implementation projects through the Arctic Puffin optimization algorithm. BRIEF DESCRIPTION OF THE DRAWINGS

[0083] Figure 1 is a schematic diagram of a digital twin-based task planning and control system of the present invention;

[0084] Figure 2 is a flow chart of generating an allocation scheme through an Arctic Puffin Optimization Algorithm of the present invention;

[0085] Figure 3 It is a comparison chart of the test results of the present invention.

[0086] In the figure: project information collection module 101, unit information collection module 102, graph network data construction module 103, prediction model construction module 104, allocation plan generation module 105, allocation plan execution module 106. DETAILED DESCRIPTION

[0087] The subject matter described herein will now be discussed with reference to example embodiments. It should be understood that the discussion of these embodiments is only to enable those skilled in the art to better understand and implement the subject matter described herein, and the functions and arrangements of the elements discussed may be changed without departing from the scope of protection of the contents of this specification. Each example may omit, replace or add various processes or components as needed. In addition, the features described relative to some examples may also be combined in other examples.

[0088] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in one or more embodiments of the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The words "first", "second" and similar words used in one or more embodiments of the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0089] like Figure 1 to Figure 3 As shown, a task planning and control system based on digital twins includes:

[0090] A project information collection module 101 is used to collect project information of M implementation projects;

[0091] The unit information collection module 102 is used to collect unit information of N implementation units;

[0092] A graph network data construction module 103, which is used to construct graph network data based on project information of M implementation projects and unit information of N implementation units;

[0093] A prediction model building module 104, which is used to build and train a prediction model;

[0094] The input of the prediction model is graph network data, and the output values ​​represent the estimated completion time and estimated cost of the project respectively;

[0095] An allocation plan generation module 105 is used to construct an objective function according to the trained prediction model and generate an allocation plan through an Arctic Puffin optimization algorithm;

[0096] The allocation plan execution module 106 is used to allocate the M implementation projects to the N implementation units according to the allocation plan.

[0097] In one embodiment of the present invention, both M and N are custom parameters, M is set according to the actual number of implementation projects, and N is set according to the actual number of implementation units.

[0098] In one embodiment of the present invention, the graph network data includes: vertices, initial features of the vertices, and edges between the vertices;

[0099] Vertices include: a first vertex and a second vertex;

[0100] The first vertex establishes data mapping with the implementation project;

[0101] The second vertex establishes data mapping with the implementation unit;

[0102] The initial feature of the first vertex is represented by vectorizing the project information of the implementation project to which the first vertex is data mapped;

[0103] The initial feature of the second vertex is represented by vectorizing the unit information of the implementation unit to which the second vertex data is mapped;

[0104] An edge is constructed between the first vertex and the second vertex to indicate that the implementation item mapped by the first vertex is allocated to the implementation unit mapped by the second vertex;

[0105] The implementation project mapped by the data of the same first vertex can only be assigned to one implementation unit mapped by the data of the second vertex at the same time;

[0106] The implementation unit of the data mapping of the same second vertex can only be assigned Q implementation items of the data mapping of the first vertex at most at the same time, where Q is a custom parameter. Preferably, Q is set to 3.

[0107] In one embodiment of the present invention, the initial feature of the first vertex is expressed as:

[0108] ;

[0109] in They respectively represent the project implementation area, project cost budget, project completion days, implementation personnel budget, implementation equipment budget and estimated implementation difficulty of the implementation project. The estimated implementation difficulty includes: high, medium and low, and is represented by real number coding.

[0110] For example, an estimated implementation difficulty of high is represented by a real number 1.1, and an estimated implementation difficulty of medium is represented by a real number 1.2.

[0111] In one embodiment of the present invention, the initial feature of the second vertex is expressed as:

[0112] ;

[0113] in They respectively represent the number of implementation personnel, the number of implementation equipment, the annual turnover, the number of historically completed implementation projects, the number of historical implementation accidents and the construction qualification level of the implementation unit. The construction qualification levels include: special grade, first grade, second grade and third grade, and are represented by real number codes.

[0114] For example, a special grade of construction qualification is represented by the real number 2.1, and a first grade of construction qualification is represented by the real number 2.2.

[0115] In one embodiment of the present invention, the prediction model includes: a first hidden layer, a second hidden layer, an extraction layer, a first classifier, and a second classifier;

[0116] The first hidden layer inputs the initial features of the vertex and outputs the updated features;

[0117] That is, the initial features of all vertices of the graph network data are converted into updated feature representations through the first hidden layer;

[0118] The second hidden layer inputs the graph network data and outputs the graph update matrix;

[0119] The graph update matrix includes M row vectors, each row vector corresponding to the comprehensive features of the edges between the M first vertices and the second vertices;

[0120] The extraction layer is used to extract M row vectors of the graph update matrix and input them into the first classifier and the second classifier;

[0121] The classification space of the first classifier represents the estimated completion time of the project;

[0122] The classification space of the second classifier represents the estimated cost of the project.

[0123] In one embodiment of the present invention, the calculation formula of the first hidden layer is as follows:

[0124] ;

[0125] Where H represents the updated features of the first hidden output, init represents the initial features of the vertices input to the first hidden layer, represents the first weight parameter, represents the first bias parameter, T represents the transposition operation, and softmax represents the softmax activation function.

[0126] In one embodiment of the present invention, the calculation formula of the second hidden layer includes:

[0127] ;

[0128] Where 1≤m≤M, 1≤n≤N, represents the comprehensive characteristics of the edge between the mth first vertex and the nth second vertex, represents the updated features of the mth first vertex, represents the updated features of the nth second vertex, represents the second weight parameter, Represents the second bias parameter, concat represents the concatenation operation, and sigmoid represents the sigmoid activation function.

[0129] In one embodiment of the present invention, sample labels of training samples used to train a prediction model are obtained by taking the average value of results simultaneously labeled by multiple managers, and a cross entropy loss function is specified as the loss function of the prediction model during the training process of the prediction model.

[0130] It should be noted that the weight parameters and bias parameters of the prediction model are updated through back propagation and gradient descent algorithms. The training of the prediction model is a conventional technical means and will not be elaborated here.

[0131] In one embodiment of the present invention, the first hidden layer can be pre-trained before the prediction model is trained. During the pre-training of the first hidden layer, the project cost budget of the implementation project mapped by the data of the first vertex is masked, and the updated features of the first vertex of the first hidden output are input to the third classifier, and the classification space of the third classifier represents the project cost budget of the implementation project; during the pre-training of the first hidden layer, the annual turnover of the implementation unit mapped by the data of the second vertex is masked, and the updated features of the second vertex of the first hidden output are input to the fourth classifier, and the classification space of the fourth classifier represents the annual turnover of the implementation unit.

[0132] It should be noted that pre-training the first hidden layer of the prediction model can improve the ability of the first hidden layer to extract features, thereby improving the prediction accuracy of the prediction model.

[0133] In one embodiment of the present invention, Figure 2 As shown, the allocation scheme is generated by the Arctic Puffin Optimization Algorithm, including the following steps:

[0134] Step S201, randomly generate an initialization population with U individuals that meets the constraint conditions, each individual in the initialization population is represented by a vector code, and define the current iteration number t as 0 and the maximum iteration number T as V, where U and V are both custom parameters. Preferably, U is set to 20 and V is set to 100;

[0135] The constraints include: the same implementation project can only be assigned to one implementation unit at a time; the same implementation unit can only be assigned to Q implementation projects at most at the same time;

[0136] The length of the vector code of an individual is M, and the mth dimension value of the vector code represents the number of the implementation unit to which the mth implementation project is assigned, where 1≤m≤M;

[0137] For example, if the value of the first dimension of the vector encoding of an individual is 3, it means that the first implementation project is assigned to the third implementation unit;

[0138] Step S202, constructing an objective function according to the trained prediction model;

[0139] The objective function is calculated as follows:

[0140] ;

[0141] Where Fit represents the fitness value of the individual, It indicates the estimated completion time of the mth implementation project assigned to the implementation unit. represents the estimated cost of the mth implementation project allocated to the implementation unit, and denote the first weight coefficient and the second weight coefficient respectively, and All are custom parameters, and their sum value is 1;

[0142] Step S203, generating a behavior conversion factor according to the current number of iterations t and the maximum number of iterations T, and judging that the behavior conversion factor is greater than a preset threshold, then proceeding to step S204, otherwise proceeding to step S205, wherein the preset threshold is a custom parameter, preferably, the preset threshold is set to 0.5;

[0143] The calculation formula of behavior conversion factor Behavior is as follows:

[0144] ;

[0145] Where rand represents a random number ranging from 0 to 1, excluding 0 and 1;

[0146] Step S204, updating the vector codes of all individuals in the initialized population through an air flight strategy;

[0147] Step S205, updating the vector codes of all individuals in the initialized population through an underwater foraging strategy;

[0148] Step S206, the current iteration number t is increased by 1. When it is determined that the current iteration number t is greater than or equal to the maximum iteration number T, the individual with the largest fitness value in the initialized population is used as the allocation scheme, otherwise, the process returns to step S203 to continue execution.

[0149] In one embodiment of the present invention, the aerial flight strategy includes: an aerial search strategy, a surprise predation strategy, and a merging strategy;

[0150] The aerial search strategy is used to update the vector codes of all individuals in the initialization population to the first vector code;

[0151] The raid predation strategy is used to update the first vector encoding of all individuals in the initialization population to the second vector encoding;

[0152] The merging strategy is used to merge the first vector codes and the second vector codes of all individuals in the initialization population, calculate the fitness value of each individual in the initialization population through the objective function, and sort them in descending order of fitness values ​​to obtain the first U individuals as the initialization population.

[0153] In one embodiment of the present invention, the underwater foraging strategy includes: collective foraging strategy, enhanced search strategy, predator avoidance strategy and merging strategy;

[0154] The collective foraging strategy is used to update the vector encoding of all individuals in the initialization population to the third vector encoding;

[0155] The enhanced search strategy is used to update the third vector encoding of all individuals in the initialization population to the fourth vector encoding;

[0156] The predator avoidance strategy is used to update the vector codes of all individuals in the initialization population to the fifth vector code;

[0157] The merging strategy is used to merge the third vector code, the fourth vector code and the fifth vector code of all individuals in the initialization population, calculate the fitness value of each individual in the initialization population through the objective function, and sort them in descending order of fitness value to obtain the first U individuals as the initialization population.

[0158] In one embodiment of the present invention, the calculation formula of the air search strategy includes:

[0159] ;

[0160] ;

[0161] Among them, 1≤i≤U, represents the vector encoding of the i-th individual with the current iteration number t, Represents the vector encoding of a random individual with the current iteration number t, represents the first vector code of the i-th individual with the current iteration number t+1, represents the random number generated by Levy flight, D represents the number of dimensions of the individual vector encoding, D is assigned the value M, R represents the air flight factor, rand represents a random number ranging from 0 to 1, excluding 0 and 1, and round represents rounding the value in the brackets, that is, if the value in the brackets is less than 0.5, it is assigned to 0, otherwise it is assigned to 1. Represents random numbers that conform to the standard normal distribution;

[0162] The calculation formula for the predation strategy includes:

[0163] ;

[0164] ;

[0165] in represents the first vector code of the i-th individual with the current iteration number t, It represents the second vector encoding of the i-th individual with the current iteration number t+1, S represents the predation factor, and tanh represents the hyperbolic tangent activation function.

[0166] It should be noted that Levy flight is a random walk model in which the probability distribution of step length follows the Levy stable distribution, which can help the Arctic Puffin optimization algorithm jump out of the local optimal solution and explore a wider solution space, thereby increasing the possibility of finding the global optimal solution.

[0167] In one embodiment of the present invention, the calculation formula of the collective foraging strategy is as follows:

[0168] ;

[0169] Where 1≤i≤U, r1, r2 and r3 are all random positive integers between 1 and U-1, excluding i. , and They represent the vector codes of individuals r1, r2 and r3 at the current iteration number t, respectively. represents the third vector code of the i-th individual with the current iteration number t+1, F represents the cooperation factor, and F is assigned a value of 0.5. represents the random number generated by Levy flight, D represents the number of dimensions of the individual vector encoding, D is assigned a value of M, and rand represents a random number ranging from 0 to 1;

[0170] The calculation formula for strengthening the search strategy includes:

[0171] ;

[0172] ;

[0173] in represents the fourth vector code of the i-th individual with the current iteration number t+1, and f represents the search factor;

[0174] The calculation formula for the predator avoidance strategy is as follows:

[0175] ;

[0176] in The fifth vector code of the i-th individual with the current iteration number t+1 is, Represents a random number uniformly distributed between 0 and 1.

[0177] In one embodiment of the present invention, Figure 3 As shown, 10 groups of test data are executed respectively by genetic algorithm, particle swarm algorithm and Arctic puffin optimization algorithm provided by the present invention. Each group of test data is to distribute 100 implementation projects to 40 implementation units at the same time. It is found through observation that the speed of finding the optimal solution of the present invention is significantly faster than that of genetic algorithm and particle swarm algorithm.

[0178] It should be noted that the present invention takes the allocation of multiple implementation projects to multiple implementation units at the same time as an example, but is not limited to this example. Similarly, according to the idea provided by the present invention, task planning of allocating multiple tasks to multiple resources at the same time can be completed. For example, multiple tasks are allocated to multiple servers at the same time. For this distributed task scheduling scenario, the task allocation is usually completed through a load balancer. However, the default rules of the load balancer are usually polling algorithms, weighted polling, etc., which cannot complete the optimal allocation of tasks according to the volume of tasks and the load status of servers, resulting in a long time for tasks to request servers. Therefore, the idea of ​​the present invention is used to construct the volume information of tasks and the load status information of servers into graph network data, and the objective function is to minimize the response time of task requests to servers. The task allocation is completed through an optimization algorithm. Therefore, the idea of ​​the present invention has a certain universality.

[0179] The above describes an embodiment of the present embodiment, but the present embodiment is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present embodiment, ordinary technicians in this field can also make many forms, all of which are within the protection of the present embodiment.

Claims

1. A task planning and control system based on digital twins, characterized in that: include: A project information collection module, which is used to collect project information of M implementation projects; Where M is a custom parameter; A unit information collection module, which is used to collect unit information of N implementation units; Where N is a custom parameter; A graph network data construction module, which is used to construct graph network data based on project information of M implementation projects and unit information of N implementation units; Graph network data includes: vertices, initial features of vertices, and edges between vertices; The vertices include: a first vertex and a second vertex, the first vertex establishes data mapping with the implementation project, and the second vertex establishes data mapping with the implementation unit; the initial feature of the first vertex is represented by vectorizing the project information of the implementation project mapped by the first vertex; the initial feature of the second vertex is represented by vectorizing the unit information of the implementation unit mapped by the second vertex; An edge is constructed between the first vertex and the second vertex to indicate that the implementation item mapped by the first vertex is allocated to the implementation unit mapped by the second vertex; The implementation project mapped by the data of the same first vertex can only be assigned to one implementation unit mapped by the data of the second vertex at the same time; The implementation unit of the data mapping of the same second vertex can only be assigned Q implementation projects of the data mapping of the first vertex at most at the same time, where Q is a custom parameter; A prediction model building module, which is used to build and train a prediction model; The input of the prediction model is graph network data, and the output values ​​represent the estimated completion time and estimated cost of the project respectively; An allocation plan generation module is used to construct an objective function based on the trained prediction model and generate an allocation plan through the Arctic Puffin Optimization Algorithm; The allocation plan execution module is used to allocate M implementation projects to N implementation units according to the allocation plan.

2. A digital twin-based task planning and control system according to claim 1, characterized in that: The initial feature representation of the first vertex is: ； in They respectively represent the project implementation area, project cost budget, project completion days, implementation personnel budget, implementation equipment budget and estimated implementation difficulty of the implementation project. The estimated implementation difficulty includes: high, medium and low, and is represented by real number coding.

3. According to the digital twin-based task planning and control system of claim 1, it is characterized in that: The initial feature representation of the second vertex is: ； in They respectively represent the number of implementation personnel, the number of implementation equipment, the annual turnover, the number of historically completed implementation projects, the number of historical implementation accidents and the construction qualification level of the implementation unit. The construction qualification levels include: special grade, first grade, second grade and third grade, and are represented by real number codes.

4. The digital twin-based task planning and control system according to claim 1, characterized in that: The prediction model includes: a first hidden layer, a second hidden layer, an extraction layer, a first classifier and a second classifier; The first hidden layer inputs the initial features of the vertex and outputs the updated features; The second hidden layer inputs the graph network data and outputs the graph update matrix; The graph update matrix includes M row vectors, each row vector corresponding to the comprehensive features of the edges between the M first vertices and the second vertices; The extraction layer is used to extract M row vectors of the graph update matrix and input them into the first classifier and the second classifier; The classification space of the first classifier represents the estimated completion time of the project; The classification space of the second classifier represents the estimated cost of the project.

5. A digital twin-based task planning and control system according to claim 4, characterized in that: The calculation formula for the first hidden layer is as follows: ； Where H represents the updated features of the first hidden output, init represents the initial features of the vertices input to the first hidden layer, represents the first weight parameter, represents the first bias parameter, T represents the transposition operation, and softmax represents the softmax activation function.

6. A digital twin-based mission planning and control system according to claim 4, characterized in that: The calculation formula for the second hidden layer includes: ； Where 1≤m≤M, 1≤n≤N, represents the comprehensive characteristics of the edge between the mth first vertex and the nth second vertex, represents the updated features of the mth first vertex, represents the updated features of the nth second vertex, represents the second weight parameter, Represents the second bias parameter, concat represents the concatenation operation, and sigmoid represents the sigmoid activation function.

7. The digital twin-based mission planning and control system according to claim 1, characterized in that: The sample labels of the training samples used to train the prediction model are obtained by taking the average of the results simultaneously labeled by multiple managers. During the training process of the prediction model, the cross entropy loss function is specified as the loss function of the prediction model.

8. The digital twin-based mission planning and control system according to claim 1, characterized in that: The allocation scheme is generated by the Arctic Puffin Optimization Algorithm, which includes the following steps: Step S201, randomly generate an initialization population with U individuals that meets the constraint conditions, each individual in the initialization population is represented by a vector code, and define the current iteration number t as 0 and the maximum iteration number T as V, where U and V are both custom parameters; The constraints include: the same implementation project can only be assigned to one implementation unit at a time; the same implementation unit can only be assigned to Q implementation projects at most at the same time; The length of the vector code of an individual is M, and the mth dimension value of the vector code represents the number of the implementation unit to which the mth implementation project is assigned, where 1≤m≤M; Step S202, constructing an objective function according to the trained prediction model; The objective function is calculated as follows: ； Where Fit represents the fitness value of the individual, It indicates the estimated completion time of the mth implementation project assigned to the implementation unit. represents the estimated cost of the mth implementation project allocated to the implementation unit, and denote the first weight coefficient and the second weight coefficient respectively, and All are custom parameters, and their sum value is 1; Step S203, generating a behavior conversion factor according to the current number of iterations t and the maximum number of iterations T, and judging that the behavior conversion factor is greater than a preset threshold, then proceeding to step S204, otherwise proceeding to step S205, wherein the preset threshold is a custom parameter; The calculation formula of behavior conversion factor Behavior is as follows: ； Where rand represents a random number ranging from 0 to 1, excluding 0 and 1; Step S204, updating the vector codes of all individuals in the initialized population through an air flight strategy; Aerial flight strategies include: aerial search strategy, raid predation strategy and merging strategy; The aerial search strategy is used to update the vector codes of all individuals in the initialization population to the first vector code; The raid predation strategy is used to update the first vector encoding of all individuals in the initialization population to the second vector encoding; The merging strategy is used to merge the first vector codes and the second vector codes of all individuals in the initialization population, calculate the fitness value of each individual in the initialization population through the objective function, and sort them in descending order of fitness values ​​to obtain the first U individuals as the initialization population; Step S205, updating the vector codes of all individuals in the initialized population through an underwater foraging strategy; Underwater foraging strategies include: collective foraging strategy, enhanced search strategy, predator avoidance strategy and merging strategy; The collective foraging strategy is used to update the vector encoding of all individuals in the initialization population to the third vector encoding; The enhanced search strategy is used to update the third vector encoding of all individuals in the initialization population to the fourth vector encoding; The predator avoidance strategy is used to update the vector codes of all individuals in the initialization population to the fifth vector code; The merging strategy is used to merge the third vector code, the fourth vector code and the fifth vector code of all individuals in the initialization population, calculate the fitness value of each individual in the initialization population through the objective function, and sort them in descending order of fitness value to obtain the first U individuals as the initialization population; Step S206, the current iteration number t is increased by 1. When it is determined that the current iteration number t is greater than or equal to the maximum iteration number T, the individual with the largest fitness value in the initialized population is used as the allocation scheme, otherwise, the process returns to step S203 to continue execution.

9. A digital twin-based mission planning and control system according to claim 8, characterized in that: The calculation formula for the air search strategy includes: ； ； Among them, 1≤i≤U, represents the vector encoding of the i-th individual with the current iteration number t, Represents the vector encoding of a random individual with the current iteration number t, represents the first vector code of the i-th individual with the current iteration number t+1, represents the random number generated by Levy flight, D represents the number of dimensions of the individual vector encoding, D is assigned to M, R represents the air flight factor, rand represents a random number ranging from 0 to 1, excluding 0 and 1, and round represents rounding the value in the brackets, that is, if the value in the brackets is less than 0.5, it is assigned to 0, otherwise it is assigned to 1. Represents random numbers that conform to the standard normal distribution; The calculation formula for the predation strategy includes: ； ； in represents the first vector code of the i-th individual with the current iteration number t, It represents the second vector encoding of the i-th individual with the current iteration number t+1, S represents the predation factor, and tanh represents the hyperbolic tangent activation function.

10. The digital twin-based mission planning and control system according to claim 8, characterized in that: The calculation formula of the collective foraging strategy is as follows: ； Where 1≤i≤U, r1, r2 and r3 are all random positive integers between 1 and U-1, excluding i. , and They represent the vector codes of individuals r1, r2 and r3 at the current iteration number t, respectively. represents the third vector code of the i-th individual with the current iteration number t+1, F represents the cooperation factor, and F is assigned a value of 0.

5. represents the random number generated by Levy flight, D represents the number of dimensions of the vector encoding of the individual, D is assigned the value of M, and rand represents a random number ranging from 0 to 1; The calculation formula for strengthening the search strategy includes: ； ； in represents the fourth vector code of the i-th individual with the current iteration number t+1, and f represents the search factor; The calculation formula for the predator avoidance strategy is as follows: ； in The fifth vector code of the i-th individual with the current iteration number t+1 is, Represents a random number uniformly distributed between 0 and 1.

Images