Satellite planning and scheduling method and device based on improved differential evolution algorithm

Abstract

The invention discloses a satellite planning and scheduling method and equipment based on an improved differential evolution algorithm, and relates to the field of satellite planning and scheduling, and the satellite planning and scheduling method based on the improved differential evolution algorithm mainly comprises the steps: obtaining a feasible solution and a parent of emergency task planning through a greedy method; selecting, mutating, crossing and selecting the parent according to the initial population fitness value, preventing the algorithm from falling into local optimum by using an orthogonal strategy, generating a better individual by using a crossover and mutation operator which successfully generates the better individual, linearly reducing the scale of the current population, updating the current population, and obtaining the optimal population. And iterating a final optimized emergency task planning sequence. By implementing the satellite planning and scheduling method and equipment based on the improved differential evolution algorithm provided by the invention, emergency task planning can be carried out on satellites in a high resource conflict degree scene.

Description

Technical Field

[0001] The present invention relates to the field of satellite planning and scheduling, and more specifically, to a satellite planning and scheduling method and device based on an improved differential evolution algorithm. Background Art

[0002] With the vigorous development of earth observation satellites by countries around the world, more and more earth observation satellites have flown into space to serve humanity. However, earth observation satellites are ultimately an important high-tech strategic resource that embodies the world's cutting-edge technology. On the one hand, their launch costs are high and their lifetimes are limited. On the other hand, as the application fields of earth observation satellites become increasingly extensive, the whole society's dependence on earth observation satellites is getting higher and higher. Inevitably, there will be more and more demands for earth observation. Therefore, satellite resources are still extremely precious compared to tasks. Therefore, planning tasks to maximize the utilization of earth observation satellite resources and give full play to their value is undoubtedly a very necessary and meaningful issue.

[0003] Satellite mission planning is a typical combinatorial optimization problem that needs to consider a large number of constraints. Satellites are not always visible to tasks, but there are specific time windows, and each observation of a task must be completed within this time window. Therefore, when performing satellite mission planning, not only the time window needs to be allocated, but also the start time of the observation within this time window needs to be allocated. For such problems, the current common practice is to perform advance planning, that is, to assume that all tasks are known, which is conventional mission planning. However, with the continuous development of satellite technology and the increasing dependence of humans on satellites, users have also put forward requirements for imaging satellites to respond quickly and observe emergency tasks, such as sudden natural disasters like volcanic eruptions and earthquakes. Satellites also need to be ready to quickly respond to these sudden tasks for observation on the basis of conventional tasks, which is emergency mission planning.

[0004] Sudden tasks have the characteristics of suddenness, high timeliness, and spatio-temporal concentration. Suddenness means that the arrival of the task is unpredictable. High timeliness indicates that the satellite needs to quickly plan for emergency tasks and observe them as soon as possible. Spatio-temporal concentration shows that the time window of sudden tasks is limited and the possibility of conflict with other tasks is relatively high. These characteristics lead to the contention of satellite resources for emergency tasks with other emergency tasks and conventional tasks at certain moments, which is a strong resource contention conflict problem. At present, in the research on emergency mission planning problems, optimization algorithms are mainly used, but there is still the problem of long solution time. Therefore, how to reasonably allocate limited satellite resources to all observation tasks in combination with the current resource status of satellites and task constraints to efficiently meet the requirements of each task is an urgently solved problem with very important practical significance.

[0005] In summary, in order for imaging satellites to be widely used in various fields and meet the observation needs of emergency tasks, this embodiment conducts research on the emergency mission planning problem of imaging satellites.

[0006] So far, scholars at home and abroad have conducted a lot of research from the aspects of problem models and solution algorithms, elaborated, classified, and analyzed various models in detail, and proposed relative algorithms to solve the models. These works provide theoretical guidance for emergency mission planning. At present, in the research of models, the constraint satisfaction model is the main model used in the research of emergency mission planning. In the research of algorithms, the emergency mission planning problem in simple scenarios can be solved using deterministic algorithms. However, in more complex scenarios, heuristic algorithms are mainly used to solve the problem. At the same time, with the gradual expansion of the application fields of machine learning research, for example, reinforcement learning has been applied in the field of emergency mission planning. Currently, the mainstream research direction of emergency mission planning is still the intelligent optimization algorithm direction. However, the learning ability of machine learning for historical information and the prediction of the future based on this are worthy of reference for the intelligent optimization algorithm in terms of parameter control and adjustment.

[0007] At present, most of the research on emergency mission planning focuses on the emergency mission planning of a single satellite, and relatively little research has been done on the emergency mission planning of multiple satellites. The research on the problem model of mission planning mainly selects some of the constraints existing in practical problems. The timeliness of the emergency mission planning problem is a very important characteristic. Therefore, research should continue to be carried out in terms of algorithms to continuously improve the solution performance. Summary of the Invention

[0008] The purpose of the present invention is to provide a satellite planning and scheduling method and device based on an improved differential evolution algorithm, which can perform emergency mission planning for satellites in high resource conflict scenarios.

[0009] The present invention provides a satellite planning and scheduling method based on an improved differential evolution algorithm, including the following steps: S1: Determine the satellite planning and scheduling problem parameters according to the satellite mission scenario; S2: Obtain a feasible solution for the emergency mission planning using the greedy method according to the satellite planning and scheduling problem parameters; S3: Obtain a set of parent mission planning sequences based on the existing emergency mission planning results, existing regular mission planning results, and the feasible solution of the emergency mission planning; S4: Calculate the fitness values according to the set of parent mission planning sequences to obtain the initial population fitness values; Select the set of parent mission planning sequences according to the initial population fitness values to obtain a new set of parent mission planning sequences; S5: Perform a hybrid crossover and mutation operation according to the new set of parent mission planning sequences to obtain mutant individuals; S6: Perform a crossover operation on the mutant individuals and the parent individuals to obtain new offspring individuals; S7: Perform a selection operation on the new offspring individuals to obtain an updated set of parent mission planning sequences; S8: Calculate the population diversity according to the updated set of parent mission planning sequences; When the population diversity is greater than the preset diversity threshold, use the current population as the next generation population; When the population diversity is not greater than the preset diversity threshold, generate better individuals using the orthogonal strategy to obtain the next generation population; S9: According to the next generation population, record the crossover and mutation operators that successfully generate better individuals, use the crossover and mutation operators that successfully generate better individuals to generate better individuals, and linearly reduce the current population size to update the current population; S10: When the number of iterations is less than the maximum number of iterations, return to step S4 to update the parent planning sequence and increment the number of iterations by 1; When the number of iterations is not less than the maximum number of iterations, use the latest parent planning sequence as the final optimized emergency mission planning sequence.

[0010] Further, the above satellite planning and scheduling problem parameters include the simulation period, satellite resource set, regular mission set, emergency mission set, and maximum number of iterations.

[0011] Further, the above fitness calculation is as follows:

[0012] , ,

[0013] , , , , , , , , Among them, is the optimization evaluation objective based on task benefits, is the i th task executed in the j th satellite's k th time window, which is a Boolean decision variable, represents the number of regular tasks, represents the number of emergency tasks, represents the number of satellites, represents the number of time windows that satellite is visible to task , represents the weight of task , represents the degree of influence on the original task sequence, is the optimization evaluation objective based on minimum perturbation, represents the degree of constraint violation, represents the tolerance parameter of the equality constraint, represents the i th inequality constraint, represents the j th equality constraint, q represents the number of inequality constraints, m represents the number of equality constraints, represents the total degree of constraint violation of an individual, represents the fitness function, represents the objective function of the constraint satisfaction model constructed for the emergency task planning problem, is the penalty coefficient, represents the i th task's j th satellite's k th available time window, represents the set of time windows with potential resource contention conflicts with time window , represents the set of time windows of satellite for task , represents the start time of time window , represents the end time of time window , represents the th small time slice corresponding to the further division of l , Indicates the time slice There is a set of time windows with potential resource contention conflicts, Indicates the time window The start time of, Indicates the time window The end time of, Indicates the satellite The execution conversion duration between two adjacent tasks on, And Of, Indicates the task The shortest continuous observation time required by, Represents the time slice The number of tasks with potential resource contention conflicts, Indicates considering the time slice The total weight of tasks with potential resource contention conflicts, Indicates the satellite For the task The set of time windows of, Indicates the i Degree of flexibility in arranging the th task, Indicates the th task considering the number of tasks with resource contention conflicts i The value of the evaluation function of, Indicates the th task of the weights of tasks with resource contention conflicts i The value of the evaluation function of.

[0014] Furthermore, the above-mentioned hybrid crossover and mutation operations are performed, such as the formula: , , , , , Among them, Is an individual randomly selected from the th generation population in the evolutionary process, t Is an individual randomly selected from the top 100*P% best individuals in the th generation population in the evolutionary process, Is the optimal individual archived in the th generation population in the evolutionary process, t t Are the fitness functions of the individual Respectively, Of the individual, Is the direction control function. For ease of description, Is also defined as the direction control function. The parameter wildcards Respectively represent ; is a random number between 0 and 1, is the threshold value, is the new individual generated by mutation operation of other individuals in the current population, is the scaling factor, is the t th generation population randomly selected from the evolutionary process.

[0015] Furthermore, the above preset diversity threshold is 0.5.

[0016] Furthermore, the above calculation of population diversity is as follows: , wherein, is the population diversity index value, is the optimal individual in the current population, is the individual with medium fitness ranking in the current population, is the dimension of the individual, and respectively represent the upper and lower bounds of the j th dimension in the search space.

[0017] The present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the above satellite planning and scheduling method based on the improved differential evolution algorithm are implemented.

[0018] The present invention also provides a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, the steps of the above satellite planning and scheduling method based on the improved differential evolution algorithm are implemented.

[0019] The present invention also provides a computer program product, including a computer program. When the computer program is executed by a processor, the steps of the above satellite planning and scheduling method based on the improved differential evolution algorithm are implemented.

[0020] Implementing the satellite planning and scheduling method and device based on the improved differential evolution algorithm provided by the present invention has the following beneficial effects: Aiming at the constraint optimization problem of emergency mission planning with mixed integer and real variables, this invention utilizes the characteristics of the efficient population-based heuristic global search algorithm - the differential evolution algorithm, which has a simple structure, is easy to implement, converges quickly and has strong robustness, and its strong solving performance in solving large-space nonlinear problems, and can better meet the high timeliness requirements of emergency missions. The Constraint Differential Evolution Algorithm based on Population Information and Orthogonal Strategy (PIOS-CDE / G) is adopted for multi-satellite emergency mission planning, which can solve the constraint optimization problem and conduct emergency mission planning in scenarios with a high degree of resource conflict. In addition, considering the limitation that the differential evolution algorithm, as a typical unconstrained algorithm, cannot be directly applied to the emergency mission planning problem, this invention designs a processing strategy based on penalty function and resource contention conflict elimination for the individuals that violate the constraints generated during the algorithm solving process, and processes the infeasible solutions that violate the constraints. To further improve the performance of the algorithm, this invention designs an initialization strategy, a hybrid mutation strategy and a local search strategy based on the L-SHADE (Success-History based Adaptive Differential Evolution with Linear Population Size Reduction) algorithm. To avoid the problem that the initial population distribution is uneven and the represented solutions are poor caused by random initialization, a greedy algorithm based on resource contention elimination is designed, and the feasible solutions obtained by the algorithm are used as part of the individuals in the initial population to help the algorithm improve the convergence speed. To avoid the randomness in the algorithm search process, a hybrid mutation strategy based on population information is designed, and the population information in the evolution process is used to guide the search direction of the algorithm. To prevent the algorithm from falling into local optimum, an orthogonal-based local search strategy is designed, and potential better solution individuals are found for the population through orthogonal experimental design to help the algorithm achieve "jumping out of the pit". Brief Description of the Drawings

[0021] The present invention will be further described below in conjunction with the drawings and embodiments. In the drawings: Figure 1 is the flowchart of the satellite planning and scheduling method based on the improved differential evolution algorithm provided by the present invention; Figure 2 is the schematic diagram of the algorithm flow of the satellite planning and scheduling method based on the improved differential evolution algorithm provided by the present invention; Figure 3 is the schematic diagram of the details of the optimized mission planning sequence provided by the present invention; Figure 4 It is the structural block diagram of the computer device provided by the present invention. Specific Embodiments

[0022] For a clearer understanding of the technical features, objectives, and effects of the present invention, the specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0023] Figure 1 The schematic diagram of the satellite planning and scheduling method based on the improved differential evolution algorithm in this embodiment is shown. In this embodiment, the satellite planning and scheduling method based on the improved differential evolution algorithm includes the following steps: S1: Determine the satellite planning and scheduling problem parameters according to the satellite mission scenario; In an exemplary embodiment, the satellite planning and scheduling problem parameters include the simulation period, satellite resource set, regular task set, emergency task set, and maximum number of iterations; As an exemplary embodiment, in step S1, determine the simulation period of the current scenario , the satellite resource set in the scenario , the regular task set in the scenario , the emergency task set to be planned , the maximum number of iterations ; S2: Obtain a feasible solution for the emergency task planning according to the satellite planning and scheduling problem parameters by using the greedy algorithm; As an exemplary embodiment, in step S2, use the greedy algorithm based on task priority and task arrival deadline to solve a feasible solution for the emergency task planning; S3: Obtain the set of parent task planning sequences according to the existing emergency task planning results, existing regular task planning results, and the feasible solution of the emergency task planning; As an exemplary embodiment, in step S3, let the number of iterations ; Take emergency task planning results and regular task planning results as a task planning sequence, randomly generate task sequences, and use them together with the feasible solution obtained by the greedy algorithm in step S2 as the parents. Denote the set of parent task planning sequences as Z, specifically as follows: , Wherein, is the number of existing emergency task planning results, is the number of existing regular task planning results, and H is the number of task sequences in the set of parent task planning sequences; Denote the \(i\)-th individual of the current population, which stores the \(i\)-th set of emergency task planning sequences, as follows: , where, 、 、…、 are the observation satellite resources corresponding to task sequences 1, 2, …, respectively, 、 、…、 are the candidate observation window numbers on the observation satellite resources corresponding to task sequences 1, 2, …, respectively, 、 、…、 are the arranged observation start times for task sequences 1, 2, …, respectively; that is, the satellite resource sequence and the observation window sequence indicate in which visible time window of which satellite for the current task in the scenario the observation is carried out, encoded with integers, and the observation time sequence \(t\) is encoded with real numbers to represent the position of the start time of this task in the selected time window; S4: Calculate the fitness values based on the set of parent task planning sequences to obtain the fitness values of the initial population; select the set of parent task planning sequences according to the fitness values of the initial population to obtain a new set of parent task planning sequences; In an exemplary embodiment, the fitness calculation is as follows:

[0024] , ,

[0025] , , , , , , , , where, is the optimization evaluation objective based on task benefits, is thei A Boolean decision variable for a task to be executed in the j th satellite's k th time window, indicating the number of regular tasks, indicating the number of emergency tasks, indicating the number of satellites, indicating the satellite for the task number of visible time windows, indicating the task weight, indicating the degree of influence on the original task sequence, indicating the optimization evaluation objective based on minimum perturbation, indicating the degree of constraint violation of the constraint, representing the tolerance parameter of the equality constraint, representing the i th inequality constraint, representing the j th equality constraint, q indicating the number of inequality constraints, m indicating the number of equality constraints, indicating the total degree of constraint violation of an individual, indicating the fitness function, representing the objective function of the constraint satisfaction model constructed for the emergency task planning problem, is the penalty coefficient, indicating the i th task's j th satellite's k th available time window, indicating the set of time windows with potential resource contention conflicts with time window , indicating the satellite for the task set of time windows, indicating the time window start time, indicating the time window end time, indicating the after further division, the corresponding l th small time slice, indicating the set of time windows with potential resource contention conflicts with time slice , indicating the time window start time, indicating the time window end time, Denote the satellite The execution conversion duration between two adjacent tasks and on the satellite, Denote the task The shortest continuous observation time required Represent the number of tasks that have potential resource contention conflicts with the time slice on the satellite, Denote considering the time slice The total weight of tasks with potential resource contention conflicts Denote the satellite For the task The time window set Denote the i Flexibility of the arranged [[th]] task Denote the evaluation function value of the [[th]] task considering the number of tasks with resource contention conflicts i where the number of tasks is considered Denote the evaluation function value of the [[th]] task considering the weights of tasks with resource contention conflicts i where the weights of tasks are considered

[0026] It should be noted that in step S4, according to the initial population fitness value, the planning sequence corresponding to this individual is judged by the or index. Among them, the two indexes can be selected according to the problem requirements. After obtaining the judgment result, the task represented by the gene position with the highest index value is eliminated, and then the conflict degree of the tasks represented by other gene positions is updated again and then eliminated again. This process is repeated until this individual no longer violates the constraints, so as to obtain a new set of parent task planning sequences; As an exemplary embodiment, in step S4, based on the characteristics of the emergency task planning problem, for the set of parent task planning sequences constructed in step S3, for each emergency task planning sequence, an optimization evaluation objective based on task benefits and an optimization evaluation objective based on minimum perturbation are respectively established according to the characteristics of the emergency task planning problem. The former objective evaluates the comprehensive task benefits based on the emergency task response time and task benefits; the specific calculation method is the sum of the comprehensive benefits of the executable tasks in the task sequence after planning is completed;

[0027] It should be noted that in addition to emergency scenarios such as earthquakes and floods, emergency tasks can also be generalized to their scenarios. In actual scenarios, there will also be some non-emergency and suddenly arriving tasks; although such tasks do not have the characteristics of high timeliness and spatio-temporal concentration, their arrival time is still unpredictable, so they cannot be planned in advance; in actual scenarios, the priority of such tasks is basically the same as that of regular tasks. Therefore, when planning emergency tasks for such tasks, in addition to considering the comprehensive benefits of the tasks, the goal of minimizing the interference caused by emergency tasks to the original task sequence can also be considered; in this embodiment, is used to represent the degree of influence on the original task sequence; generally speaking, the tasks in the original task sequence are disturbed in two cases. One is that the execution time window of the task is reallocated, and the other is that the task is deleted from the execution sequence. These two cases undoubtedly have different degrees of influence on the task and need to be distinguished. In this embodiment, the influence degree of the former case is set to 1, and the value of the influence degree of the latter case is 2. Therefore, a new optimization evaluation target is obtained, that is, minimum perturbation: , Since emergency task planning belongs to a constrained optimization problem, and the differential evolution algorithm itself cannot handle constraints; therefore, after determining the above optimization index, it is also necessary to handle the constraints. This embodiment designs two methods for constraint handling; The latter target evaluates based on the operation of the emergency task on the original task sequence; for the parent planning sequence in step S3, an optimization evaluation target is selected according to the type of emergency task to calculate the fitness value, and the solutions that violate the constraints are resolved using the constraint handling strategy during the calculation process; Specifically, in step S4, the infeasible solution resolution strategy for violating constraints is as follows: (1) Using the penalty function method for constraint handling: Before describing the penalty function method, it is necessary to first quantitatively describe the degree of constraint violation: , Among them, represents the inequality constraint, represents the equality constraint, i and j respectively represent the i-th inequality constraint or equality constraint, represents the tolerance parameter of the equality constraint (in this embodiment, it is set to a specific value of 0.0001). Therefore, the total constraint violation degree of an individual can be expressed by the formula:

[0028] The penalty function method is one of the most widely used constraint handling methods at present. Its basic principle is to convert the constrained optimization problem into an unconstrained optimization problem by adding a penalty term based on the total constraint violation degree to the objective function. Its definition is as follows, where represents the objective function of the constraint satisfaction model constructed for the emergency task planning problem in the previous text: , where, is the penalty coefficient. In the penalty function method, the relationship between the objective function and the constraint conditions is mainly balanced by the penalty coefficient. In this embodiment, it is modified by dynamically adjusting it with the increase of the iteration times. Its main purpose is to set a smaller penalty coefficient at the initial stage of iteration, so that the algorithm can search the infeasible solution region to a certain extent to promote population diversity, and set a larger value in the later stage to make the solution concentrate in the feasible region to increase the convergence speed; Design an infeasible solution resolution strategy based on the conflict degree for constraint handling. In the preprocessing stage of emergency task planning, the visible time windows of all tasks have been calculated. Therefore, based on this, the distribution characteristics of the visible time windows in the current problem scenario can be determined, and then the resource contention conflict degree of each time window can be analyzed. The process is as follows: For any available time window in a scenario, the potential resource contention conflict time window set of can be obtained by judging whether this window coincides with the available time windows of other tasks: , After obtaining the potential resource contention conflict time window set of , in order to evaluate the conflict degree more precisely, is further divided into multiple small time windows , where depends on the step size used for further dividing the time window; , Based on , the conflict degree index of is defined as follows, representing the number of tasks that potentially conflict with : , In emergency mission planning, generally, the benefit value of an emergency mission is greater than that of a regular mission. At the same time, the impact degree on other time windows varies when different time slices are selected. Therefore, in order to more reasonably describe the conflict degree in the emergency mission planning problem, a conflict degree index considering task weights is defined. ; , After obtaining the conflict degree index of the task, another issue needs to be concerned. According to the relevant theory of satellite-ground coverage calculation, different tasks have different numbers of visible time windows. It is possible that there is a task with only one time window with a very high conflict degree. In this case, if this task is to be planned, it can only be arranged in this time window, rather than being resolved only according to the conflict degree index. Therefore, on the basis of the conflict degree, this embodiment also pays attention to the flexibility of task arrangement: , It reflects the difficulty level of the task that can be executed. The larger it is, the easier the task is to be planned. Therefore, when resolving conflicts for infeasible solutions, a task with a lower task benefit, a higher conflict degree, and a higher flexibility of being arranged should be resolved first. To sum up, this embodiment obtains a complete evaluation index. and : , , Each population individual in the differential evolution algorithm represents a solution, that is, a planning sequence. When there are individuals violating constraints during the evolution process, the planning sequence corresponding to this individual can be or evaluated according to the index (the two indexes can be selected according to the problem requirements). After obtaining the evaluation result, the task represented by the gene position with the highest index value is eliminated, and then the conflict degree of the tasks represented by other gene positions is updated again and then eliminated again, repeating until this individual no longer violates the constraints; S5: According to the new set of parent task planning sequences, perform a hybrid crossover and mutation operation to obtain mutant individuals; In an exemplary embodiment, the hybrid crossover and mutation operation is as shown in the formula: , , , , , Among them, is an individual randomly selected from the t -th generation population in the evolutionary process, is an individual randomly selected from the top 100×P% best individuals in the t -th generation population in the evolutionary process, is the optimal individual archived in the t -th generation population in the evolutionary process, are the fitness functions of individuals respectively, is the direction control function. For the convenience of description, is also defined as the direction control function. The parameter wildcards represent respectively; is a random number between 0 and 1, is the threshold, is a new individual generated by the mutation operation of other individuals in the current population, is the scaling factor, is an individual randomly selected from the t -th generation population in the evolutionary process.

[0029] As an exemplary embodiment, in step S5, a mutation operation in the PIOS-CDE / G algorithm is performed on the parent task planning sequence calculated in step S4. A hybrid crossover and mutation strategy is designed for the characteristics of the emergency task planning problem. The purpose is to adopt different mutation strategies for different individuals in the population, improve the search efficiency of the algorithm, and use the existing information exchange between individuals to find better results. The specific implementation of the hybrid mutation strategy is as follows: The purpose of the hybrid mutation strategy is to adopt different mutation strategies for different individuals in the population, improve the search efficiency of the algorithm, and use the existing information exchange between individuals to find better results. It mainly includes: DE / rand / 1 strategy, DE / best / 1 strategy, DE / current-to-pbest / 1; (1) The DE / rand / 1 strategy randomly selects three individuals , and from the population, and . The difference between and the other two individuals is multiplied by the mutation operator F and then summed to obtain the mutant individual: , (2) The DE / best / 1 strategy is a variant of DE / rand / 1, which selects the best individual along with two other individuals randomly selected from the current population and participate in the mutation operation: , (3) The meaning of the DE / current-to-pbest / 1 strategy is to select an individual from the external archive and randomly select an individual from the best 100*P% individuals in the current population Then randomly select an individual from the combined set of the external archive and the current population : , In these mutation strategies, generally only the overall information of the population is utilized, and all the information in the mutation process is not fully applied. Therefore, in order to further improve the optimization ability of the algorithm, differential information is designed to be used to control the search direction. The direction control function: , The direction control function is determined by the relationship between its variables and can be expressed as: , The above formula indicates that when or , the fitness value of individual is better than , which means that the moving direction represented by the differential vector is more optimal and may make better. Therefore, the direction of remains unchanged at this time; when the fitness value of is worse than , or these three cases, which means that the search direction of the population may point to a worse area in the solution space. Therefore, at this time, the search direction should move in the opposite direction, which is more likely to make better; when this situation the fitness value performs the worst, while the fitness value of is better than , although at this time Searching in the opposite direction can improve the diversity of the algorithm. Therefore, this strategy sets a threshold Q, which dynamically changes with the increase of the number of iterations, so as to achieve the effect of maintaining a certain diversity of the algorithm in the early stage of the algorithm and increasing the convergence speed in the later stage of the algorithm. On the basis of the direction control function, this embodiment optimizes the above three mutation strategies, combines the direction control function with the mutation strategy to increase the convergence speed of the algorithm: , , , S6: Perform a crossover operation on the mutated individual and the parent individual to obtain a new offspring individual; As an exemplary embodiment, in step S6, perform a crossover operation on the mutated individual obtained in step S5 and the parent individual to obtain a new offspring individual. Its main purpose is to perform a wider search in the solution space and improve the global search ability of the algorithm; S7: Perform a selection operation on the new offspring individual to obtain an updated set of parent task planning sequences; As an exemplary embodiment, in step S7, perform a selection operation on the offspring individual obtained in step S6, and intercept H task planning sequences as the new parent set. The purpose is to promote the update of the task planning sequence by the information interaction between the offspring populations generated by two different operations, compare the advantages and disadvantages of different task planning sequences, and select the optimal H task planning sequences to obtain a better task planning sequence; S8: Calculate the population diversity according to the updated set of parent task planning sequences; when the population diversity is greater than the preset diversity threshold, use the current population as the next generation population; when the population diversity is not greater than the preset diversity threshold, use the orthogonal strategy to generate better individuals to obtain the next generation population; In an exemplary embodiment, the preset diversity threshold is 0.5; In an exemplary embodiment, the calculation of the population diversity is as follows: , where, is the population diversity index value, is the optimal individual in the current population, is the individual with medium fitness ranking in the current population, is the dimension of the individual, and respectively represent the upper and lower bounds of the j th dimension in the search space.

[0030] As an exemplary embodiment, in step S8, a population diversity estimation mechanism based on Euclidean distance is used to determine whether the differential evolution algorithm has fallen into stagnation or premature convergence. When this is the case, the diversity of the current population is good, and the next step can be directly carried out. While when this is the case, the diversity of the current population is poor, and the population may fall into a local optimum. At this time, an orthogonal strategy is used to help the algorithm "jump out of the pit" to obtain a better solution. The specific implementation of the global optimization strategy based on orthogonality is as follows: Use a population diversity estimation mechanism based on Euclidean distance to determine whether the differential evolution algorithm has fallen into stagnation or premature convergence: , According to existing research, when this is the case, the diversity of the current population is good, and the next step can be directly carried out; while when this is the case, the diversity of the current population is poor, and the population may fall into a local optimum. Therefore, to solve this problem, a new strategy needs to be introduced to generate new individuals to help the algorithm jump out of the current area. To achieve this goal, the premise and key are to generate better solutions based on the existing information on the basis of the current population. However, in conventional experiments, if the dimension of the problem is D and there are N different values for each dimension, where N represents the number of population individuals, it is not difficult to find that there are a total of combinations, and the number of them grows exponentially. This means that the cost required to obtain a new and better individual by this method is extremely high. Based on this, this embodiment introduces an orthogonal strategy to generate better solutions when stagnation or premature convergence problems occur during the evolution process. The main idea of this strategy is to use the method of orthogonal experimental design to sample a part of representative individuals and conduct multi-dimensional and multi-value experiments. Multi-dimensional represents the problem dimension, and multi-value means that there are multiple possible values for each dimension. Representative individuals are obtained from the entire population according to the orthogonal table for experiments, so as to achieve the method of obtaining results equivalent to a large number of experiments with the least number of experiments. Among them, the orthogonal table is a pre-defined table of mutually orthogonal, represented by M, where M represents the combined array: , After having the orthogonal table, it can be used to determine the best individual. First, use the orthogonal table to select the corresponding combination, and conduct factor analysis on it to determine the influence of each value on each dimension, so as to determine the best value for each dimension. The calculation formula for the degree of influence is: , where is the degree of influence of the value q on the k-th dimension, is the number of combinations, is the fitness function of m, is whether the value of the k-th dimension of the m-th combination is q; In this embodiment, based on the above example, 2 individuals are selected to perform the orthogonal process, that is, the values (1, 2) and the dimension is the individual's own dimension D to construct orthogonal individuals, where the two selected individuals are the optimal individual and a random individual among the top 100p% of the population; S9: According to the next-generation population, record the crossover and mutation operators that successfully generate better individuals, use the crossover and mutation operators that successfully generate better individuals to generate better individuals, and linearly reduce the current population size to update the current population; As an exemplary embodiment, in step S9, record the crossover and mutation operators that successfully generate better individuals, the purpose of which is to further generate better individuals using these operators later, and then linearly reduce the current population size, the purpose of which is to balance the exploration and exploitation capabilities of the algorithm and improve the convergence performance and search efficiency of the algorithm; S10: When the number of iterations is less than the maximum number of iterations, return to step S4, update the parent planning sequence, and increment the number of iterations by 1; when the number of iterations is not less than the maximum number of iterations, use the latest parent planning sequence as the final optimized emergency task planning sequence; As an exemplary embodiment, in step S10, if the number of iterations T is less than the maximum number of iterations T max , return to steps S4 to S9, update the parent planning sequence, and the number of iterations T is incremented by 1; otherwise, output the latest set of parent planning sequences as the finally obtained optimized emergency task planning sequence; the purpose is to continuously repeat steps S4 to S9, update the parent planning sequence, and output the latest planning sequence as the finally obtained optimized task planning sequence of the present invention.

[0031] In some embodiments, the above satellite planning and scheduling method based on the improved differential evolution algorithm can also be implemented in the following manner. Figure 2 is the algorithm flow diagram of the satellite planning and scheduling method based on the improved differential evolution algorithm provided by the present invention; as an exemplary embodiment, the satellite planning and scheduling method based on the improved differential evolution algorithm includes the following steps: Step 1: Determine the satellite resource set in the scenario , the set of regular tasks in the scenario , the set of emergency tasks to be planned , the maximum number of iterations , the start and end times of the current scenario simulation period are from January 1, 2023 to January 2, 2023; Step 2: Use the greedy algorithm based on task priority and task arrival deadline to solve the feasible solution of the emergency task planning; Step 3: Let the number of iterations ; Consider the planning results of 50 emergency task plans and the planning results of 150 regular tasks as a task planning sequence. Randomly generate 97 task sequences, which together with the feasible solution obtained by the greedy algorithm in Step 2 are used as the parent generation. The total quantity in the parent generation is represented by to represent the set of parent planning sequences, specifically as follows: , where represents the i-th individual in the current population, storing the i-th group of emergency task planning sequences, which is represented as follows: , where and represent in which satellite of the scene and which visible time window of this task the current task is observed; The time point uses the real number coding method to represent the start time of this task in the start time of the selected time window; To demonstrate the detailed calculation process of the present invention, assume now:

[0032] Step 4: Based on the characteristics of the emergency task planning problem, for each emergency task planning sequence in the set of parent task planning sequences constructed in Step 3, establish an optimization evaluation objective based on task benefits. The objective evaluates the comprehensive task benefits based on the emergency task response time and task benefits. The specific calculation method is the sum of the comprehensive benefits of the executable tasks in the task sequence after planning;

[0033] In addition, to make the display more comprehensive, this example also calculates the objective of minimizing perturbation:

[0034] Since the emergency task planning belongs to a constrained optimization problem, and the differential evolution algorithm itself cannot handle constraints; Next, perform constraint processing on the individuals in the population. For the sake of complete display, this embodiment still uses two methods to perform constraint processing respectively: (1) Penalty function method: Before describing the penalty function method, it is necessary to first quantitatively describe the degree of violation of the constraints: , In the above formula represents the inequality constraint, Represents an equality constraint. Here, i and j respectively represent the number of the inequality constraint or equality constraint. Represents the tolerance parameter of the equality constraint. Therefore, the total constraint violation degree of the individuals in this embodiment can be expressed as:

[0035] The basic principle of the penalty function method is to add a penalty term based on the total constraint violation degree to the objective function to convert the constrained optimization problem into an unconstrained optimization problem:

[0036] (2) Resolution of infeasible solutions based on the conflict degree: For the available time window of this embodiment , the potential resource contention conflict time window set of can be obtained by judging whether this window coincides with the available time windows of other tasks:

[0037] For , it is further divided into multiple small time windows , and the conflict time window set is further obtained;

[0038] Based on , the conflict degree index of is defined as follows, representing the number of tasks that have potential conflicts with ;

[0039] In emergency task planning, in order to more reasonably describe the conflict degree in the emergency task planning problem, the conflict degree index considering task weights is defined:

[0040] In addition, it is possible that there is a task with only one time window with a very high conflict degree. Then, if this task is to be planned, it can only be arranged in this time window, rather than being resolved only according to the conflict degree index. Therefore, based on the conflict degree, this embodiment also pays attention to the flexibility of task arrangement, as follows:

[0041] reflects the difficulty level of the task that can be executed, The larger it is, the easier the task is to be planned; therefore, when resolving conflicts for infeasible solutions, a task with a lower task benefit, a higher conflict degree, and a higher flexibility in being arranged should be resolved first. In summary, the complete evaluation index is obtained in this embodiment and :

[0042]

[0043] Evaluate the individuals in the population or the evaluation of the index. After obtaining the evaluation result, eliminate the task represented by the gene position with the highest index value, then re-update the conflict degree of the tasks represented by other gene positions and eliminate them again. Repeat this process until the individual no longer violates the constraints; Step 5: Perform the mutation operation in the PIOS-CDE / G algorithm on the parent task planning sequence calculated in Step 4. A hybrid crossover mutation strategy is designed for the characteristics of the emergency task planning problem, aiming to adopt different mutation strategies for different individuals in the population, improve the search efficiency of the algorithm, and use the existing information exchange between individuals to find better results; mainly including: DE / rand / 1 strategy, DE / best / 1 strategy, DE / current-to-pbest / 1 strategy; In order to further improve the optimization ability of the algorithm, the differential information is designed to control the search direction, and the direction control function: , The direction control function is determined by the relationship between its variables and can be expressed as: , Based on the direction control function, this embodiment optimizes the several mutation strategies mentioned above, combines the direction control function with the mutation strategies to increase the convergence speed of the algorithm: , , , Step 6: Perform the crossover operation on the mutated individuals and the parent individuals obtained in Step 5 to obtain new offspring individuals, whose main purpose is to perform a wider search in the solution space and improve the global search ability of the algorithm; Step 7: Perform a selection operation on the offspring individuals obtained in Step 6, and intercept H task planning sequences as the new parent set. The purpose is to promote the update of the task planning sequence by the information interaction between the offspring populations generated by two different operations, compare the advantages and disadvantages of different task planning sequences, and select the optimal H task planning sequences from them to obtain a better task planning sequence; Step 8: Use the population diversity estimation mechanism based on Euclidean distance to determine whether the differential evolution algorithm is stuck or converges prematurely. Still taking this example for specific demonstration:

[0044] At this time , if the diversity of the current population is poor, the population may fall into a local optimum. Therefore, a new strategy needs to be introduced to generate new individuals to help the algorithm jump out of the current area. To achieve this goal, the premise and key are to generate better solutions based on the existing information in the current population; based on this, this embodiment uses the orthogonal strategy to generate better solutions; the main idea of this strategy is to use the method of orthogonal experimental design to sample a part of representative individuals and conduct multi-dimensional and multi-value experiments. Multi-dimensional represents the problem dimension, and multi-value means that there are multiple possible values for each dimension; in this embodiment, based on the above example, 2 individuals are selected to execute the orthogonal process, that is, the values are (1, 2) and the dimension is the dimension D of the individual itself to construct orthogonal individuals, where the two selected individuals are the optimal individual and a random one among the top 100p% individuals in the population; among them, the orthogonal table is a predefined table of mutually orthogonal, represented by M represents the combined array:

[0045] After having the orthogonal table, it can be used to determine the best individual. First, use the orthogonal table to select the corresponding combination, and perform factor analysis on it to determine the influence of each value on each dimension to determine the best value for each dimension. Take calculating the influence degree of the value 1 on the first dimension as an example: ; Step 9: Record the crossover and mutation operators that successfully generate better individuals, the purpose of which is to further generate better individuals using these operators later, and then linearly reduce the current population size, the purpose of which is to balance the exploration and exploitation capabilities of the algorithm and improve the convergence performance and search efficiency of the algorithm; Step 10: If the number of iterations T is less than the maximum number of iterations 100, return to Steps 4 to 9 to update the set of parent planning sequences, and the number of iterations TIncrement by 1; otherwise, output the latest set of parent planning sequences as the finally obtained optimized emergency task planning sequences. The purpose is to continuously repeat steps 4 to 9, update the parent planning sequences, and output the latest planning sequences as the finally obtained optimized task planning sequences of the present invention. The finally obtained optimized task planning sequences are shown in Table 1, details are as Figure 3 shown.

[0046] Table 1: Finally obtained optimized task planning sequences

[0047] This embodiment provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the above-mentioned satellite planning and scheduling method based on the improved differential evolution algorithm are implemented. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only memory (ROM), a random access memory (RAM), a flash memory, a hard disk drive (HDD) or a solid-state drive (SSD), etc.; the storage medium can also include a combination of the above-mentioned types of memories.

[0048] This embodiment provides a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, the steps of the above-mentioned satellite planning and scheduling method based on the improved differential evolution algorithm are implemented.

[0049] As Figure 4As shown, the computer device 120 may include: at least one processor 121, such as a Central Processing Unit (CPU), at least one communication interface 123, a memory 124, and at least one communication bus 122. Among them, the communication bus 122 is used to realize the connection and communication between these components. Among them, the communication interface 123 may include a display screen (Display) and a keyboard (Keyboard). Optionally, the communication interface 123 may also include a standard wired interface and a wireless interface. The memory 124 may be a high-speed random access memory (Random Access Memory, RAM), or a non-volatile memory, such as at least one disk memory. Optionally, the memory 124 may also be at least one storage device located far from the aforementioned processor 121. Among them, an application program is stored in the memory 124, and the processor 121 calls the program code stored in the memory 124 to execute any of the above method steps. Among them, the communication bus 122 may be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. The communication bus 122 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 4It is represented by only one line, but it does not mean that there is only one bus or one type of bus. Among them, the memory 124 may include volatile memory, such as random-access memory (RAM); the memory may also include non-volatile memory, such as flash memory, hard disk drive (HDD) or solid-state drive (SSD); the memory 124 may also include a combination of the above types of memory. Among them, the processor 121 may be a central processing unit (CPU), a network processor (NP), or a combination of a CPU and an NP. Among them, the processor 121 may further include a hardware chip. The above hardware chip may be an application-specific integrated circuit (ASIC), a programmable logic device (PLD), or a combination thereof. The above PLD may be a complex programmable logic device (CPLD), a field-programmable gate array (FPGA), a generic array logic (GAL), or any combination thereof. Optionally, the memory 124 is also used to store program instructions. The processor 121 may call the program instructions to implement the satellite planning and scheduling method based on the improved differential evolution algorithm as in this embodiment.

[0050] This embodiment provides a computer program product, including a computer program, which implements the steps of the above satellite planning and scheduling method based on the improved differential evolution algorithm when executed by a processor.

[0051] The embodiments of the present invention have been described above in conjunction with the accompanying drawings. However, the present invention is not limited to the above specific embodiments. The above specific embodiments are merely illustrative and not restrictive. Under the inspiration of the present invention, those of ordinary skill in the art can also make many forms without departing from the purpose of the present invention and the scope protected by the claims. These all belong to the protection scope of the present invention.

Claims

1. A satellite planning and scheduling method based on an improved differential evolution algorithm, characterized in that: The following steps are involved: S1: Determine the parameters of satellite planning and scheduling problems according to the satellite mission scenario; S2: According to the satellite planning and scheduling problem parameters, a feasible solution for emergency mission planning is obtained using a greedy method; S3: Obtain a parent generation task planning sequence set according to the existing emergency task planning results, the existing conventional task planning results and the feasible solution of the emergency task planning; S4: Calculate the fitness value according to the parent task planning sequence set to obtain an initial population fitness value; select the parent task planning sequence set according to the initial population fitness value to obtain a new parent task planning sequence set; S5: performing a mixed crossover mutation operation according to the new parent generation task planning sequence set to obtain a mutant individual; S6: performing a crossover operation on the variant individual and the parent individual to obtain a new offspring individual; S7: performing a selection operation on the new offspring individuals to obtain an updated parent generation task planning sequence set; S8: Calculate the population diversity according to the updated parent generation task planning sequence set; when the population diversity is greater than a preset diversity threshold, use the current population as the next generation population; when the population diversity is not greater than the preset diversity threshold, use the orthogonal strategy to generate better individuals to obtain the next generation population; S9: According to the next generation population, record the crossover mutation operator that successfully generates a better individual, use the crossover mutation operator that successfully generates a better individual to generate a better individual, and linearly reduce the current population size to update the current population; S10: When the number of iterations is less than the maximum number of iterations, return to step S4, update the parent generation planning sequence, and increase the number of iterations by 1; When the number of iterations is not less than the maximum number of iterations, the latest parent planning sequence is used as the final optimized emergency task planning sequence.

2. The satellite planning and scheduling method based on the improved differential evolution algorithm according to claim 1 is characterized in that: The satellite planning and scheduling problem parameters include simulation period, satellite resource set, conventional task set, emergency task set, and maximum number of iterations.

3. The satellite planning and scheduling method based on the improved differential evolution algorithm according to claim 1 is characterized in that: The fitness calculation is performed as follows: ， ， ， ， ， ， ， ， ， ， in, is the optimization evaluation target based on task benefits, For the i The task is in j Satellite k Boolean decision variables executed in time windows, Indicates the number of regular tasks, represents the number of emergency tasks, represents the number of satellites, Indicates satellite To the task The number of visible time windows, Indicates the task The weight of Indicates the degree of impact on the original task sequence. represents the optimization evaluation objective based on minimum disturbance, represents the degree of constraint violation, represents the tolerance parameter of the equality constraint, Representative i Inequality constraints, Representative j equality constraints, q represents the number of inequality constraints, m represents the number of equality constraints, represents the total constraint violation degree of the individual, represents the fitness function, represents the objective function of the constraint satisfaction model constructed for the emergency mission planning problem, is the penalty coefficient, Indicates i Task No. j Satellite k Available time windows, Representation and time window There is a set of time windows with potential resource contention conflicts. Indicates satellite To the task A collection of time windows, Represents a time window The start time of Represents a time window The end time of Express After further division, the corresponding l A small piece of time, Representation and time slice There is a set of time windows with potential resource contention conflicts. Represents a time window The start time of Represents a time window The end time of Indicates satellite Two adjacent tasks and The execution time of the conversion is Indicates the task The minimum observation duration required is Delegates and Time Slices The number of tasks with potential resource contention conflicts, Considering time slice The total weight of tasks with potential resource contention conflicts, Indicates satellite To the task A collection of time windows, represents the flexibility of scheduling the i-th task, represents the evaluation function value of the ith task considering the number of resource contention conflicting tasks, The evaluation function value of the i-th task representing the resource contention conflict task weight.

4. The satellite planning and scheduling method based on the improved differential evolution algorithm according to claim 1 is characterized in that: The hybrid crossover mutation operation is performed as shown in the formula: ， ， ， ， ， in, For the first t A randomly selected individual from the population, In the process of evolution t An individual is randomly selected from the first 100×P% best individuals in the generation population. In the process of evolution t The best individual archived in the generation population, Individual The fitness function of is the direction control function. Also defined as a direction control function, parameter wildcard Respectively represent ; is a random number between 0 and 1, is the threshold value, New individuals are generated by mutation operations on other individuals in the current population. is the scaling factor, For the first t An individual is randomly selected from the population.

5. The satellite planning and scheduling method based on the improved differential evolution algorithm according to claim 1 is characterized in that: The preset diversity threshold is 0.

5.

6. The satellite planning and scheduling method based on the improved differential evolution algorithm according to claim 1 is characterized in that: The population diversity is calculated as follows: ， in, is the population diversity index value, is the best individual in the current population, is the individual with medium fitness ranking in the current population, For the individual dimension, and They represent the first j The upper and lower bounds of the dimension.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the satellite planning and scheduling method based on the improved differential evolution algorithm as described in any one of claims 1 to 6 are implemented.

8. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the satellite planning and scheduling method based on the improved differential evolution algorithm as described in any one of claims 1-6 are implemented.

9. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the satellite planning and scheduling method based on the improved differential evolution algorithm described in any one of claims 1-6 are implemented.

Images