Multi-level system element automatic mapping method

Abstract

The invention discloses a multi-level system element automatic mapping method, which comprises the following steps of: setting a fitness function by calculating total task execution cost and total task execution efficiency, optimizing a system element mapping relation by using a real-value particle swarm optimization algorithm, and introducing a Pareto optimal theory to select an optimal solution so as to realize comprehensive consideration of multiple factors. According to the method, solving of a multi-objective optimization problem is effectively achieved, system units are reasonably distributed for system tasks, automatic mapping of system elements is achieved, system design requirements are met, and the method has important application value.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of system design, and particularly relates to a multi-level system element automatic mapping method. BACKGROUND

[0002] In the complex system design process, a key problem is to reasonably allocate system units to achieve system tasks. For a specific task, the type and quantity of units required to complete the task are determined by a reasonable and scientific method, and the system task unit demand mapping is achieved. Through formulaic modeling, a suitable allocation strategy of system units is found to achieve the completion of system tasks, and the automatic mapping of system elements is achieved.

[0003] The automatic mapping of system elements is a typical multi-objective optimization problem. Li Yingzi et al. consider the task-person-resource matching degree, establish a scheduling model, and design a double-layered single-parent genetic algorithm to solve the model. Guo Chenghao et al. establish a task-resource matching optimization model and use genetic radiation to obtain the optimal resource combination. In the existing research results of automatic mapping of system elements, most of them consider a single fitness function and do not comprehensively consider multiple optimization objectives. The Pareto optimal theory and the particle swarm optimization algorithm can effectively solve the multi-objective problem, have the characteristics of simple structure and fast convergence speed, and are widely used in solving nonlinear optimization problems. SUMMARY

[0004] The purpose of the present application is to provide a multi-level system element automatic mapping method, which solves the problem of automatically matching tasks and units in the system under the premise of meeting the task capability demand, taking execution cost and execution efficiency as optimization objectives.

[0005] The technical solution of the present application is as follows: the system is composed of three elements of task, capability and unit. The task is described by three attributes of task dimension, task type and task degree, and the unit is described by four attributes of unit type, target, unit index and unit value.

[0006] A multi-level system element mapping method, the steps are as follows:

[0007] Step 1: Determine the capability type of task mapping by task dimension and task type, and determine the capability value of task mapping by task degree, to generate a task capability demand matrix .

[0008] Step 2: Determine the capability type that can be realized by the unit by unit type and target, and determine the capability value that can be realized by the unit by unit index, to generate a unit capability attribute matrix At the same time, a unit value matrix is formed by unit value .

[0009] Step 3, a multi-objective particle swarm optimization algorithm is established, the multi-objective particle swarm optimization algorithm combines the Pareto optimal theory, the real value particle swarm optimization algorithm is used for optimizing system task unit matching, multi-objective particle swarm optimization algorithm parameters are set, particle position matrix and particle velocity matrix are randomly initialized, and the last item of the particle position matrix is a penalty term.

[0010] Step 4, the distribution of the unit is determined according to the particle position matrix, and a unit distribution matrix is generated .

[0011] Step 5, in combination with the task capability demand matrix , the unit capability attribute matrix and the unit value matrix , according to the unit distribution matrix , the completion of each task is counted, and the total cost and total efficiency of task execution are calculated, and an optimization model of system task and system unit mapping is established.

[0012] Step 6, the penalty term of the particle position matrix is set according to the task completion degree.

[0013] Step 7, in combination with the total cost and total efficiency of task execution, the system cost-effectiveness ratio is set.

[0014] Step 8, the global optimal solution of the particle is selected based on the Pareto optimal theory, and the specific steps are as follows:

[0015] Step 8.1, the current particle is compared with the particles in the archive, and the Pareto optimal solution is selected to constitute the archive;

[0016] Step 8.2, according to the reduction criterion, part of the particles in the archive are reduced, and the size of the archive is constrained;

[0017] Step 8.3, the adaptive grid method is applied, and the global optimal solution is selected in the archive according to the congestion degree;

[0018] Step 9, the particle position and particle velocity are updated based on the real value particle swarm algorithm;

[0019] Step 10, whether the end condition is met, that is, whether the upper limit of the iteration number is reached, if not, return to step 4; otherwise, terminate the optimization, and obtain the optimal archive, that is, the optimal Pareto front;

[0020] Step 11, according to the system cost-effectiveness ratio, the optimal particle is selected from the optimal archive, the optimal feasible solution of the system task and system unit mapping optimization model is found, and the automatic mapping of the task and the unit in the system is realized.

[0021] Compared with the prior art, the present application has the following advantages:

[0022] (1) Real-valued particle swarm optimization algorithm is used to realize automatic mapping between tasks and units in the system.

[0023] (2) The Pareto optimal theory is introduced to effectively realize solution of the multi-objective optimization problem, so that automatic mapping of system elements is realized with lower total task execution cost and higher total task execution efficiency. BRIEF DESCRIPTION OF DRAWINGS

[0024] Figure 1 A flow chart of the automatic mapping method of the multi-level system elements is provided. DETAILED DESCRIPTION

[0025] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0026] The technical solutions of the various embodiments of the present application can be combined with each other, but must be based on the fact that they can be realized by those skilled in the art. When the combination of technical solutions appears to be contradictory or unachievable, it should be considered that the combination of technical solutions does not exist and is not within the scope of protection claimed by the present application.

[0027] The specific embodiments, technical difficulties and points of the present application will be further introduced below with reference to the design examples.

[0028] The automatic mapping method of the multi-level system elements of the present application uses task completion degree and unit maximum number as constraints, and uses execution cost and execution efficiency as optimization targets, and uses a multi-objective particle swarm optimization algorithm to optimize the mapping relationship between system tasks and units.

[0029] The construction of the system (S) can be represented in the form of three elements of tasks (M), capabilities (A) and units (C), i.e. The task is described by three attributes of task dimension, task type and task degree, and the unit is described by four attributes of unit type, target, unit index and unit value, as shown in the following table:

[0030]

[0031] The number of tasks in the system is N, the total number of capability types is P, the number of unit types is K, the maximum number of the kth unit is , and the maximum total number of units is . With system capacity as the intermediate variable, the system task and system unit are mapped and allocated to realize automatic mapping of system elements. The method flow chart is shown in Figure 1 . The specific steps are as follows:

[0032] Step 1, determine the type of task mapping capability from the task dimension and task type, determine the value of task mapping capability from the task degree, and generate the task capability demand matrix , , is a natural number set. The task capability demand matrix describes the demand value of the task on the capability in the system, and there are cases where a certain task needs multiple capabilities.

[0033] Step 2, determine the type of capability that the unit can achieve from the unit type and the target, determine the value of the capability that the unit can achieve from the unit index, and generate the unit capability attribute matrix , . The unit capability attribute matrix describes the capability value that each unit in the system can achieve, and there are cases where a certain unit can achieve multiple capabilities. Form a unit value matrix from the unit value , The qth element in the matrix represents the unit value of the qth unit.

[0034] Step 3, establish a multi-objective particle swarm optimization algorithm, and use the real-valued particle swarm optimization algorithm to optimize the matching of system tasks and units. Set the parameters of the multi-objective particle swarm optimization algorithm, the number of particle swarms is , the number of iterations is , the size of the archive is , the number of optimization objectives is 2, and the number of grids in each dimension is . Randomly generate a particle position matrix and a corresponding particle velocity matrix , is a real number set, the initial value is , the maximum particle velocity is , the value range of the first Q columns of the particle position matrix is , and the Q+1th column of the particle position matrix is a penalty item. The value of the Q+1th column is 0 or 1, when the task completion degree 100, ; when the task completion degree <100, .

[0035] Step 4, determine the allocation of the unit according to the particle position matrix , i.e., determine the execution unit of the task, and generate the unit allocation matrix Specifically as follows:

[0036] According to the particle position matrix The lth element in the first Q elements The value range of the lth unit corresponding to the home task: if The lth unit does not participate in the execution of the task, ; if The lth unit executes the first task, ; if The lth unit executes the second task, ; if The lth unit executes the Nth task, .

[0037] Step 5, combining the task capability requirement matrix , the unit capability attribute matrix and the unit value matrix , according to the unit allocation matrix , the completion degree of each task in the system is calculated and the total cost of task execution and the efficiency of task execution are calculated, as follows:

[0038] Step 5.1, calculate the completion degree of each task. Form the task completion matrix from the unit capability attribute matrix and the unit allocation matrix , that is, for the nth task, find the unit set allocated to the nth task in , and according to , the aggregate capability of the unit set is calculated into . Compare and by row to generate the task completion degree statistical vector , for the nth task, if satisfies , , otherwise ;

[0039] Step 5.2, calculate the total cost of task execution. According to the unit participation in the task in , generate the unit condition matrix , if the lth unit participates in the execution of the task, , otherwise . Combine to calculate the total cost of task execution TC:

[0040] ,

[0041] If the task completion degree statistical vector There is a term with 0, i.e. there is a task not completed, then the total cost of task execution increases the penalty term That is

[0042] ,

[0043] Wherein, ;

[0044] Step 5.3, calculate the total efficiency of task execution. For the nth task, its task execution efficiency Can be expressed as,

[0045] ,

[0046] Wherein, The pth item ability is the exertion efficiency, as shown in the following formula:

[0047] ,

[0048] The total efficiency of task execution F can be expressed as:

[0049] ;

[0050] Step 5.4, build the optimization model of system task and system unit mapping. Considering the total cost of task execution and the total efficiency of task execution as the optimization target, the optimization model of system task and system unit mapping can be expressed as:

[0051] ,

[0052] Step 6, according to the task completion degree statistical vector The particle position matrix Is set as follows:

[0053] ,

[0054] Step 7, combined with the total cost of task execution and the total efficiency of task execution, set the system cost-effectiveness ratio As shown in the following formula:

[0055] ,

[0056] The system cost-effectiveness ratio is used as a single fitness function to update the local optimal solution of the particle That is, the smaller the system cost-effectiveness ratio of the current particle, the better the particle.

[0057] Step 8, select the global optimal solution based on the Pareto optimal theory, as follows:

[0058] Step 8.1, compare the current particle with the particles in the archive according to the total cost TC and the total efficiency F of task execution as the fitness function, and select the Pareto optimal solution to join the archive;

[0059] Step 8.2, constrain the size of the archive to ensure that the number of particles in the archive does not exceed the limited value, and sort the particles in the archive from large to small according to the system cost-effectiveness ratio CER, if the total number of particles in the current archive exceeds the limited value by Q, then delete the Q particles with larger CER before sorting, so as to maintain a stable archive size;

[0060] Step 8.3, based on the adaptive grid, divide the archive target space of the multi-objective optimization problem into sub-regions of equal size according to the grid number Select the region with the smallest congestion, i.e. the sub-region with the fewest particles, and randomly select a particle from it as the leader, i.e. the global optimal solution .

[0061] Step 9, update the particle position and particle velocity based on the real-valued particle swarm optimization algorithm, as follows:

[0062] Update the particle position X and its corresponding particle velocity matrix using the real-valued particle swarm optimization algorithm, the particle position and particle velocity update formula of the real-valued particle swarm optimization algorithm is as follows:

[0063] ,

[0064] ,

[0065] where w、 and are real numbers, rand is a random number belonging to the interval [0,1], and the values of learning factors and are usually 2.05. The inertia weight w decreases from 0.9 to 0.4 with the number of iterations. is the task allocation matrix of the i-th particle in the t-th iteration, i.e. the particle position of the real-valued particle swarm optimization algorithm, is the particle velocity of the i-th particle in the t-th iteration, represents the local optimal solution in the t-th iteration, represents the global optimal solution in the t-th iteration. is the particle position of the i-th particle in the t+1-th iteration, i.e. the updated particle position, is the particle velocity of the i-th particle in the t+1-th iteration, i.e. the updated particle velocity, where .

[0066] When the updated position of the particle is not in the constraint range, boundary processing is needed to ensure that the particle position is in the constraint range, and common boundary processing includes absorbing boundary and the like.

[0067] Step 10, judging whether the ending condition is met, that is, whether the upper limit G of the iteration number is reached, if the optimization is terminated and the optimal archive is obtained, that is, the optimal Pareto front, if not, returning to step 4.

[0068] Step 11, according to the selection criterion, selecting the particle with the minimum CER in the optimal archive as the optimal particle to obtain the optimal solution of the particle position , and further obtaining the optimal unit allocation matrix , so as to determine the optimal mapping relationship between the system task and the unit, and realize the solution of the mapping optimization model of the system task and the system unit.

[0069] In summary, the application innovatively proposes a multi-level system element automatic mapping method, the method optimizes the mapping relationship between the system task and the unit based on the multi-objective particle swarm optimization algorithm, optimizes the particle position and speed based on the real value particle swarm optimization algorithm according to the Pareto optimal theory, obtains the optimal particle position, and further obtains the optimal unit allocation matrix. The application comprehensively considers the influence of multiple factors on the optimal solution, realizes the mapping of the system task and the unit with lower total task execution cost and higher total task execution efficiency under the requirements of meeting the task completion degree and the unit quantity limit.

Claims

1. A method for automatic mapping of elements in a multi-level system, characterized in that: The system is characterized by three elements: task, capability, and unit. The attributes of task and unit are analyzed to derive the task-capability mapping relationship and the unit-capability mapping relationship.

2. The method for automatic mapping of multi-level system elements according to claim 1, characterized in that, The specific steps are as follows: Step 1: Determine the task-capability mapping relationship based on task dimensions, task type, and task severity, and generate a task capability requirement matrix; Step 2: Determine the unit-capability mapping relationship and cost based on unit type, target, unit indicators, and unit value, and generate the unit capability attribute matrix and unit value matrix; Step 3: Set the parameters of the multi-objective particle swarm optimization algorithm, randomly generate the particle position matrix and particle velocity matrix, and the last item of the particle position matrix is ​​the penalty term; Step 4: Determine the allocation of units based on the particle position matrix and generate a unit allocation matrix; Step 5: Combining the task capability requirement matrix, unit capability attribute matrix, and unit value matrix, and based on the unit allocation matrix, statistically analyze the completion status of each task, calculate the total task execution cost and total task execution efficiency, and establish a system task and system unit mapping optimization model. Step 6: Set the particle position matrix penalty item based on the task completion rate; Step 7: Update the local optimum of the particle by setting a single fitness function; Step 8: Update the global optimal solution of the particle based on Pareto optimality theory; Step 9: Update particle position and particle velocity; Step 10: Determine if the upper limit of the number of iterations has been reached. If so, obtain the optimal Pareto front; otherwise, return to step 4. Step 11: Based on a single fitness function, select the optimal particle from the optimal archive, find the optimal feasible solution of the system task and system unit mapping optimization model, and realize the automatic mapping of tasks and units in the system.

3. The automatic mapping method for multi-level system elements according to claim 2, characterized in that: In step 3, the multi-objective particle swarm optimization algorithm combines Pareto optimality theory and uses real-valued particle swarm optimization algorithm to optimize the matching of system task units, and updates particle positions and particle velocities through real-valued particle swarm optimization algorithm.

4. The automatic mapping method for multi-level system elements according to claim 2, characterized in that: In step 5, the total task execution cost and total task execution efficiency are the optimization objectives, while the task completion rate and the total number of units are the constraints.

5. The automatic mapping method for multi-level system elements according to claim 2, characterized in that: In step 7, the system cost-effectiveness ratio is used as the single fitness function.

6. The automatic mapping method for multi-level system elements according to claim 2, characterized in that: In step 8, the total task execution cost and the total task execution efficiency are the two fitness functions of Pareto optimality theory.

7. The method for automatic mapping of multi-level system elements according to claim 6, characterized in that, In step 8, the global optimal solution for the particle is selected based on Pareto optimality theory, as follows: Step 8.1: Compare the current particle with the particles in the archive, and select the Pareto optimal solution to form the archive; Step 8.2: Delete some particles from the archive according to the deletion criteria to constrain the size of the archive; Step 8.3: Apply the adaptive grid method to select the globally optimal solution in the archive based on the degree of crowding.

Images