A multi-satellite collaborative task allocation method and system for non-time-sensitive mobile targets

Abstract

The present invention provides a multi-satellite collaborative task allocation method for non-time-sensitive mobile targets. The non-time-sensitive mobile target task set is annotated to the allocation main satellite, and then shared to each sub-satellite. Each sub-satellite calculates the visible window of each task respectively, which reduces the calculation pressure of the allocation main satellite. After obtaining the satellite set that can be allocated to each task, the rules are evolved based on GEP, and the optimal allocation rule is used to allocate the task to the most suitable satellite. It has been verified by experiments that the method of the present invention only needs to select the optimal allocation rule to perform task allocation, so the solution speed is faster and the allocation efficiency is higher.

Description

Technical Field

[0001] The invention belongs to the field of satellite mission planning, and in particular relates to a multi-satellite collaborative task allocation method and system for non-time-sensitive mobile targets. Background Art

[0002] Multi-satellite collaborative task assignment problem (MSCTAP) for non-time-sensitive mobile target tracking aims to achieve efficient collaboration of inter-satellite task planning through fast on-satellite task allocation and realize rapid response to the dynamic uncertainty of mobile targets.

[0003] The MSCTAP problem for non-time-sensitive mobile target tracking is a typical constrained optimization problem, which aims to maximize the benefits of constellation scheduling tasks by quickly and reasonably allocating tasks. Tasks are generated by the onboard autonomous management module, and usually contain attribute information such as priority, payload attributes, imaging duration, and windows with satellites. For satellites, there are different status information, including current attitude, working status, onboard power and storage, etc. For a mission, there may be more than one transiting satellite within a planning cycle. How to achieve fast and high-quality task allocation to ensure the completion of the mission with a high success rate.

[0004] With the enhancement of onboard capabilities (onboard image recognition, computing and communication capabilities), satellite autonomy has become a reality. In this case, the response to mobile targets can be achieved more quickly. However, whether facing mobile targets or conventional observation targets, there will be many tasks on the satellite. These tasks are dynamically arrived and also come from the observation targets dynamically generated on the satellite. In the face of many dynamic uncertainties, it is difficult to take into account both the high timeliness and high quality of the solution by using traditional operations optimization methods to solve onboard autonomous task planning. The heuristic-based methods used in the existing technology are difficult to obtain high-benefit solutions, and the use of precise or meta-heuristic algorithms needs to be improved in terms of solution efficiency. Rapid and high-quality solution of onboard task planning is the basis for responding to many uncertain factors on the satellite, and it is also an urgent need for autonomous collaboration on the satellite. Summary of the invention

[0005] The technical problem to be solved by the present invention is how to quickly and high-quality assign tasks to satellites for non-time-sensitive mobile targets to ensure the completion of tasks with a high success rate. A multi-satellite collaborative task assignment method and system for non-time-sensitive mobile targets are proposed.

[0006] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0007] A multi-satellite collaborative task allocation method for non-time-sensitive mobile targets comprises the following steps:

[0008] Step 1: Obtain the non-time-sensitive mobile target task set and the satellite set and state parameters involved in the planning;

[0009] Step 2: The non-time-sensitive mobile target task set is annotated to the assigned main satellite, and the non-time-sensitive mobile target task set is shared with each sub-satellite in the satellite set. The assigned main satellite refers to the satellite with the longest transit time, and the sub-satellite refers to other non-main satellites that are about to transit;

[0010] Step 3: Each satellite calculates the visible window of each task according to its own state and orbit information to obtain the visible window of each satellite for each task, and feeds back the calculation result of the visible window to the assigned main satellite;

[0011] Step 4: Allocate the main satellite to the multi-satellite tasks according to the visible window calculation results of each satellite for each task fed back by each sub-satellite.

[0012] Furthermore, in step 4, the method for allocating the main satellite to allocate multi-satellite tasks according to the visible window calculation results fed back by each sub-satellite is to construct a multi-satellite collaborative task allocation model for non-time-sensitive mobile target tracking, solve the multi-satellite collaborative task allocation model, and obtain a multi-satellite collaborative task allocation plan.

[0013] Furthermore, the multi-satellite collaborative task allocation model is:

[0014]

[0015]

[0016]

[0017] Ω j ={tsk i |x ij =1} (4)

[0018]

[0019]

[0020] Among them, n sat Indicates the number of satellites to be assigned, n tsk represents the number of tasks to be assigned, Φ(·) represents the scheduling function of satellite j, and returns the scheduling benefit, S j represents the state attribute of satellite j before allocation, Ω j represents the set of tasks assigned to satellite j, C j represents the set of constraints for satellite j, Ψ jrepresents the optimization target of satellite j;

[0021] x ij is a 0-1 decision variable, 1 means the i-th task tsk i Assigned to satellite j; v ij Represents a 0-1 parameter, 1 means there is a visible window between task i and satellite j, 0 means there is no visible window between task i and satellite j;

[0022] Indicates the mission imaging payload requirement, 0 for optical and 1 for SAR;

[0023] Indicates the satellite payload type, 0 is optical and 1 is SAR;

[0024] Task i imaging mode requirements, different working modes correspond to different widths;

[0025] Represents the set of satellite j working modes.

[0026] Furthermore, a method for solving the multi-satellite collaborative task allocation model is:

[0027] Step 4.1: Filter the set of satellites that can be assigned to each task through constraints from the calculation results of each satellite's visible window for each task;

[0028] Step 4.3: Use the optimal allocation rule to select the best satellite from the set of satellites that can be allocated to each task in turn to allocate to the task.

[0029] Furthermore, the optimal allocation rule is obtained based on the rule evolution method of GEP, specifically:

[0030] Step 4.3.1: Randomly generate an initial population, which contains n pop Chromosomes, each chromosome contains several coding genes, each coding gene consists of a head and a tail. The first code in the head must be a function set, and the other coding bits in the head can be function sets or terminal sets. The tail must be a terminal set.

[0031] Step 4.3.2: Evaluate the fitness of all chromosome individuals in the population;

[0032] Step 4.3.3: Perform individual selection, crossover, and mutation on the population to form a new population;

[0033] Step 4.3.4: After satisfying the number of iterations, select the individual with the largest fitness as the output of the optimal allocation rule.

[0034] Furthermore, the function set is +, -, *, sin, cos, max, min, and the endpoint set is the feature vector FV that characterizes the multi-satellite collaborative task allocation problem. ij ,

[0035]

[0036] 1) TN j It represents the proportion of the assigned tasks on satellite j to the total assigned tasks of all satellites;

[0037]

[0038] 2)RE j The remaining power percentage of satellite j;

[0039]

[0040] Egy max Indicates the maximum power of the satellite. Indicates the remaining power of the satellite;

[0041] 3) AP j represents the mean priority of the tasks assigned to satellite j,

[0042]

[0043] After normalization, it is expressed as:

[0044]

[0045] pri max Indicates the highest task priority;

[0046] 4) SP j represents the standard deviation of the priority of the tasks assigned to satellite j;

[0047]

[0048] After normalization, it is expressed as:

[0049]

[0050] 5) AR j represents the average value of the priority-imaging duration ratio of the assigned tasks on satellite j;

[0051]

[0052] After normalization:

[0053]

[0054] 6)SR jrepresents the priority-imaging duration ratio standard deviation of the assigned tasks on satellite j;

[0055]

[0056] 7) AO j represents the mean time of the overhead of the assigned tasks on satellite j;

[0057]

[0058] After normalization:

[0059]

[0060] 9) SO j represents the standard deviation of the time when the assigned task passes over the top of satellite j;

[0061]

[0062] The relative planning period Tplan is used for normalization to obtain the normalized

[0063]

[0064] 9) WP i Indicates the priority of task i to be assigned;

[0065] After normalization, it is expressed as:

[0066] 10) WR i It represents the priority of task i to be assigned - imaging duration;

[0067]

[0068] After normalization:

[0069]

[0070] 12) WO ij It indicates the time when satellite j passes through the top of the assigned task i;

[0071] Through the planning period Tplan, normalization is performed to obtain the normalized

[0072]

[0073] 12)WL ij It indicates the proportion of visible window length of satellite j executing the task i to be assigned;

[0074] WL ij =we ij-wb ij ;

[0075] After normalization:

[0076] 13)WC ij represents the conflict degree between the task to be assigned i and the task to be planned for satellite j;

[0077] WC ij =Cd ij ;

[0078]

[0079] After normalization:

[0080]

[0081] 14)WV ij represents the length of the non-overlapping time period in the visible window corresponding to the task to be assigned i and the visible time windows of other assigned tasks on satellite j;

[0082]

[0083] The function ||·|| represents the sum of the lengths of the window segments in the set;

[0084] After normalization:

[0085]

[0086] 15)AS ij represents the statistical mean of the observed slopes of the task to be assigned i and the assigned tasks on satellite j;

[0087] The observed slope sp between task i and task l il It can be defined as:

[0088]

[0089]

[0090] After normalization:

[0091]

[0092] 17) SS ij represents the standard deviation of the slope of observations between the task to be assigned i and the assigned task on satellite j;

[0093]

[0094] After normalization:

[0095]

[0096] Furthermore, the method for evaluating the fitness of the chromosome individuals in step 4.3.2 is:

[0097] Step 4.3.2.1: Decode the chromosome individuals to obtain the allocation rule tree, and convert the allocation rule tree into a regular arithmetic expression;

[0098] Step 4.3.2.2: For the nsc scenarios generated by training, use the rules expressed by the regular arithmetic expression to assign the tasks to be assigned to each scenario;

[0099] Step 4.3.2.3: Perform single-satellite task scheduling for the tasks assigned to a satellite;

[0100] Step 4.3.2.4: Calculate the total revenue of all satellites after executing task scheduling according to the regular task allocation expressed by the regular arithmetic expression in a scenario;

[0101] Step 4.3.2.5: Until the total benefits after task scheduling in nsc scenarios are obtained and the average value is calculated, the average value is taken as the fitness evaluation value of an individual.

[0102] Furthermore, the individual selection in step 4.3.3 uses the tournament sorting method. When constructing the new parent population, the proportion of the population size is pro el Keep the elites and randomly select n from the population tour Individuals, copy the one with the largest fitness and put it into the new parent generation;

[0103] The operators of crossover mutation are: single-point crossover operator, according to the single-point crossover probability p c1 Determine whether to select crossover; double-point crossover operator, according to the double-point crossover probability p c2 Determine whether crossover is needed; gene crossover operator, according to the gene crossover probability p ge Determine whether to select crossover;

[0104] The mutation operators are: single-point mutation operator according to the single-point mutation probability p ms To determine whether mutation is needed, the segment reversal operator randomly selects a coding segment in the chromosome for reversal, which is based on the segment reversal probability p mi Determine whether mutation is needed; the IS transposition operator randomly selects a segment from the coding segment and inserts it into the non-first position of another gene on the chromosome. is The RIS transposition operator randomly selects a segment starting from the function set from the head of a gene and inserts it into the starting position of the head of another gene.

[0105] The present invention also provides a multi-satellite collaborative task allocation system for non-time-sensitive mobile targets, comprising the following modules:

[0106] Information acquisition module: used to obtain the non-time-sensitive mobile target mission set and the satellite set involved in the planning;

[0107] Information annotation module: used to annotate the non-time-sensitive mobile target task set obtained by the information acquisition module to the assigned main satellite, and share the non-time-sensitive mobile target task set with each sub-satellite. The assigned main satellite refers to the satellite with the longest transit time, and the sub-satellite refers to other non-main satellites that are about to transit;

[0108] Mission visible window calculation module: used for each satellite in the satellite set to calculate the visible window of each mission in the mission set according to its own status and orbit information to obtain the visible window of each satellite for each mission, and feed back the visible window calculation result to the assigned main satellite;

[0109] Task allocation module: used to allocate multi-satellite tasks to the main satellite based on the visible window calculation results of each satellite for each task fed back by each sub-satellite.

[0110] By adopting the above technical solution, the present invention has the following beneficial effects:

[0111] The present invention provides a multi-satellite collaborative task allocation method and system for non-time-sensitive mobile targets. The non-time-sensitive mobile target task set is annotated to the allocation main satellite, and then shared to each sub-satellite. Each sub-satellite calculates the visible window of each task respectively, thereby reducing the calculation pressure of the allocation main satellite. After obtaining the satellite set that can be allocated to each task, the rules are evolved based on GEP, and the optimal allocation rule is used to allocate the task to the most suitable satellite. It has been verified by experiments that the method of the present invention only needs to select the optimal allocation rule to perform task allocation, so the solution speed is faster and the allocation efficiency is higher. BRIEF DESCRIPTION OF THE DRAWINGS

[0112] Figure 1 It is a flow chart of the system of the present invention;

[0113] Figure 2 This is a schematic diagram of the multi-satellite collaboration process under the centralized-distributed collaboration architecture;

[0114] Figure 3 It is a multi-satellite task allocation problem under serialized decision-making;

[0115] Figure 4 A framework for solving the MSCTAP problem for non-time-sensitive moving target tracking;

[0116] Figure 5 This is a schematic diagram of the chromosome encoding and decoding process;

[0117] Figure 6 Schematic diagram of the sequence deconstruction process for task assignment;

[0118] Figure 7 It is a flow chart based on the evolution of GEP rules;

[0119] Figure 8 Schematic diagram for selecting individuals in the population;

[0120] Fig. 9 Schematic diagram of the evolution of GEP at three different scales. DETAILED DESCRIPTION

[0121] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0122] The multi-satellite collaborative task assignment problem (MSCTAP) for non-time-sensitive mobile target tracking aims to achieve efficient collaboration of inter-satellite task planning through fast on-board task allocation and realize rapid response to dynamic uncertainties of mobile targets. The MSCTAP problem for non-time-sensitive mobile target tracking belongs to a typical constrained optimization problem, whose purpose is to maximize the benefits of constellation scheduling tasks by quickly and reasonably allocating tasks. The enhancement of on-board capabilities (on-board image recognition, computing and communication capabilities) makes satellite autonomy a reality. In this case, it is possible to respond to mobile targets more quickly. However, whether facing mobile targets or conventional observation targets, there will be many tasks on the satellite, which are dynamically arrived and also come from the observation targets dynamically generated on the satellite. Faced with many dynamic uncertainties, it is difficult to simultaneously take into account the high timeliness and high quality of the solution by using traditional operations research optimization methods to solve on-board autonomous task planning. In practice, it is difficult to obtain a high-benefit solution by using heuristic-based methods, and the efficiency of solving the solution by using exact or meta-heuristic algorithms needs to be improved. Rapid and high-quality solution of on-board mission planning is the basis for responding to many uncertain factors on the satellite, and is also an urgent need for autonomous collaboration on the satellite.

[0123] In order to solve the problem of the present invention, some assumptions are made. Specifically:

[0124] 1) The earth observation satellite system under consideration is equipped with visible light payload and SAR payload. No matter which payload is carried, the platform has agile three-axis attitude maneuverability and can form an observation window for the target.

[0125] 2) Only one payload can be working on the same satellite at the same time. For the SAR payload, it can only be in one imaging mode at any time.

[0126] 3) Ignore the satellite storage constraint. After the satellite has the image recognition capability, invalid images can be erased autonomously and valid images can be compressed. Therefore, the storage constraint will no longer be a tight constraint.

[0127] 4) Satellite imaging and attitude maneuvers will consume electricity, and the electricity consumption is proportional to the imaging time and attitude maneuvering time respectively.

[0128] 5) The satellite uses uniform ground-speed imaging to image the target. The observation time of the mission is proportional to the length of the generated strip (grid side length) of the mission, which has been taken into account when the mission is generated.

[0129] 6) Ignore the impact of the satellite's imaging observation angle on the imaging resolution.

[0130] Figures 1 to 9 FIG. 4 shows a specific embodiment of a multi-satellite collaborative task allocation method for non-time-sensitive mobile targets of the present invention, such as Figure 1 and Figure 4 The following steps are shown:

[0131] Step 1: Obtain the non-time-sensitive mobile target task set and the satellite set and status parameters involved in the planning; in this embodiment, the task attributes in the non-time-sensitive mobile target task set include attribute information such as priority, payload attributes, imaging duration, and window with the satellite. For satellites involved in the planning, satellite attributes include status information such as current attitude, working status, on-board power and storage. For a task, task allocation refers to how to achieve fast and high-quality task allocation and ensure the completion of the task with a high success rate when there may be more than one transiting satellite within a planning period.

[0132] Step 2: The non-time-sensitive mobile target task set is annotated to the assigned main satellite, and the non-time-sensitive mobile target task set is shared with each sub-satellite. The assigned main satellite refers to the satellite with the longest transit time, and the sub-satellite refers to other non-main satellites that are about to transit.

[0133] In this embodiment, the powerful capability of constellation distributed computing and the ability of on-board centralized task allocation to achieve global optimization are utilized. Figure 2 As shown in the figure, first, based on the satellite-ground tracking and control network, the target information of multi-source situation fusion can be injected into the satellite, and the satellite with the longest transit time (corresponding to the medium-orbit satellite) (referred to as the main satellite in this embodiment) serves as the global target maintenance satellite. At the same time, the target information is shared, but other satellites are not responsible for maintaining and managing the target. The main satellite distributes the target information of the tracking observation task to each sub-satellite that is about to pass by.

[0134] Step 3: Each satellite calculates the visible window of each task according to its own status and orbit information to obtain the visible window of each satellite for each task, and feeds back the calculation result of the visible window to the assigned main satellite.

[0135] In this embodiment, each satellite calculates the visible window of each task according to its own state and orbit information by using STK, or the patent document CN105893659A A method for quickly calculating satellite access prediction provides a method for calculating the visible window.

[0136] Step 4: The master satellite is assigned to the multi-satellite tasks according to the visible window calculation results of each satellite for each task fed back by each sub-satellite. In this embodiment, since the master satellite obtains global information, the allocation scheme is more excellent. After each sub-satellite obtains the task assigned by the master satellite, it performs autonomous scheduling to maximize the total observation benefit and feeds back the results of the execution observation to the master satellite. The centralized-distributed collaborative architecture adopted by the present invention takes into account the global information advantages of centralized task allocation and the efficiency of distributed scheduling calculation.

[0137] In this embodiment, the method for allocating the main satellite to allocate multi-satellite tasks according to the visible window calculation results fed back by each sub-satellite is to construct a multi-satellite collaborative task allocation model for non-time-sensitive mobile target tracking, solve the multi-satellite collaborative task allocation model, and obtain a multi-satellite collaborative task allocation plan.

[0138] The multi-satellite collaborative task allocation model is:

[0139]

[0140]

[0141]

[0142] Ω j ={tsk i |x ij =1} (4)

[0143]

[0144]

[0145] Among them, n sat Indicates the number of satellites to be assigned, n tsk represents the number of tasks to be assigned, Φ(·) represents the scheduling function of satellite j, and returns the scheduling benefit, S j represents the state attribute of satellite j before allocation, Ω j represents the set of tasks assigned to satellite j, C j represents the set of constraints for satellite j, Ψj represents the optimization target of satellite j;

[0146] x ij is a 0-1 decision variable, 1 means the i-th task tsk i Assigned to satellite j; v ij Represents a 0-1 parameter, 1 means there is a visible window between task i and satellite j, 0 means there is no visible window between task i and satellite j;

[0147] Indicates the mission imaging payload requirement, 0 for optical and 1 for SAR;

[0148] Indicates the satellite payload type, 0 is optical and 1 is SAR;

[0149] Task i imaging mode requirements, different working modes correspond to different widths;

[0150] Represents the set of satellite j working modes.

[0151] In this embodiment, the method for solving the multi-satellite collaborative task allocation model is:

[0152] Step 4.1: Filter the set of satellites that can be assigned to each task through constraints from the calculation results of each satellite's visible window for each task;

[0153] Step 4.2: Use the optimal allocation rule to select the best satellite from the set of satellites that can be allocated to each task and allocate it to the task.

[0154] When a new task is generated for a time-insensitive mobile target, the previous task has become a thing of the past and invalid even if it has not been completed, and a complete replanning needs to be performed. The efficiency and quality of the algorithm are particularly important. The multi-satellite collaborative task allocation for tracking time-insensitive mobile targets aims to quickly provide input data for single-satellite complete rescheduling through rapid on-board allocation. Speed ​​and high quality are the key points of multi-satellite collaborative task allocation, and are also the key to achieving rapid response to time-insensitive mobile targets. Since the MSCTAP problem for tracking time-insensitive mobile targets can be classified as a classic node matching problem, its solution space complexity is exponential. For a system with n tsk Tasks and n sat For the MSCTAP problem of non-time-sensitive mobile target tracking, leaving aside the constraints such as the visible time window, there are n matching options for each task. sat Therefore, the space complexity of this problem is Moreover, the spatial complexity of this problem has shown an explosive growth with the growth of tasks and satellite scales. It is difficult to obtain the optimal solution in a short time using an exact solution algorithm, and classic optimization solution software like CPLEX cannot handle black box functions such as Φ(·). For the current metaheuristic solution method, the timeliness of the solution is still not high. Heuristics have high timeliness for fast solutions, but the solution quality is generally poor. Therefore, the idea of ​​solving the multi-satellite collaborative task allocation model in the present invention is to model the problem as a sequential decision problem, and gradually construct the optimal allocation plan for all tasks and satellites by sequentially deciding the best allocation satellite for each task. Figure 3 As shown, the assignment of tasks is considered one by one. In each decision step, the set of assignable satellites is first screened by constraints (such as Figure 3 Considering the current allocation status, use the allocation rules to allocate the tasks to be allocated and establish a matching relationship (such as Figure 3 The MSCTAP problem for non-time-sensitive mobile target tracking is solved through multi-step serialized decision making. The most critical issue is how to formulate allocation rules and describe the current allocation status.

[0155] In this embodiment, due to the multiple attributes of tasks and satellites, it is an important step to select which attributes and weight them to obtain a good allocation rule for task allocation. Therefore, the method for formulating allocation rules is based on the evolutionary method GEP of gene expression programming to train the multi-satellite task allocation rules. The allocation rules obtained through training are used to generate the optimal solution to obtain the score of the satellite under a specific task allocation. Finally, the satellite with the largest benefit is selected for allocation based on the score. The weighted rule is obtained through GEP evolution. Figure 6 As shown,

[0156] In this embodiment, the evolution method of multi-satellite collaborative task allocation rules based on GEP designs a variety of test scenarios of task attributes and satellite states, extracts the tasks and satellite parameters in each test scenario as problem instances, and then models the problem instances. According to the 16 attributes mentioned below, feature extraction is performed, and then rule evolution is performed to obtain the optimal allocation rule. The specific training process of the allocation rule is:

[0157] Step 4.3.1: Randomly generate an initial population, which contains n pop chromosomes, each chromosome contains several coding genes, each coding gene consists of a head and a tail, the first code in the head must be a function set, the other coding bits in the head can be function sets or endpoint sets, and the tail must be an endpoint set. In this embodiment, the function set is +, -, *, sin, cos, max, min, and the endpoint set is the feature vector FV that characterizes the multi-satellite collaborative task allocation problem.ij ,

[0158]

[0159] 1) TN j It represents the proportion of the assigned tasks on satellite j to the total assigned tasks of all satellites;

[0160]

[0161] 2)RE j The remaining power percentage of satellite j;

[0162]

[0163] Egy max Indicates the maximum power of the satellite. Indicates the remaining power of the satellite;

[0164] 3) AP j represents the mean priority of the tasks assigned to satellite j,

[0165]

[0166] After normalization, it is expressed as:

[0167]

[0168] pri max Indicates the highest task priority;

[0169] 4) SP j represents the standard deviation of the priority of the tasks assigned to satellite j;

[0170]

[0171] After normalization, it is expressed as:

[0172]

[0173] 5) AR j represents the average value of the priority-imaging duration ratio of the assigned tasks on satellite j;

[0174]

[0175] After normalization:

[0176]

[0177] 6)SR j represents the priority-imaging duration ratio standard deviation of the assigned tasks on satellite j;

[0178]

[0179] 7) AO j represents the mean time of the overhead of the assigned tasks on satellite j;

[0180]

[0181] After normalization:

[0182]

[0183] 8) SO j represents the standard deviation of the time when the assigned task passes over the top of satellite j;

[0184]

[0185] The relative planning period Tplan is used for normalization to obtain the normalized

[0186]

[0187] Features 1) to 8) are satellite status related features.

[0188] 9) WP i Indicates the priority of task i to be assigned;

[0189] After normalization, it is expressed as:

[0190] 10) WR i It represents the priority of task i to be assigned - imaging duration;

[0191]

[0192] After normalization:

[0193]

[0194] Features 9) and 10) are the features of the tasks to be assigned.

[0195] 11) WO ij It indicates the time when satellite j passes through the top of the assigned task i;

[0196] Through the planning period Tplan, normalization is performed to obtain the normalized

[0197]

[0198] 12)WL ij It indicates the proportion of visible window length of satellite j executing the task i to be assigned;

[0199] WLij =we ij -wb ij ;

[0200] After normalization:

[0201] 13)WC ij represents the conflict degree between the task to be assigned i and the task to be planned for satellite j;

[0202] WC ij =Cd ij ;

[0203]

[0204] After normalization:

[0205]

[0206] 14)WV ij represents the length of the non-overlapping time period in the visible window corresponding to the task to be assigned i and the visible time windows of other assigned tasks on satellite j;

[0207]

[0208] The function ||·|| represents the sum of the lengths of the window segments in the set;

[0209] After normalization:

[0210]

[0211] 15)AS ij represents the statistical mean of the observed slopes of the task to be assigned i and the assigned tasks on satellite j;

[0212] The observed slope spil of task i and task l can be defined as:

[0213]

[0214]

[0215] After normalization:

[0216]

[0217] 16) SS ij represents the standard deviation of the slope of observations between the task to be assigned i and the assigned task on satellite j;

[0218]

[0219] After normalization:

[0220]

[0221] Features 11) to 16) are satellite-related features. The 16 attribute features in this embodiment are normalized to improve the applicability and scalability of the evolution rules to the problem. Gene encoding is performed through these 16 attributes and function sets, such as Figure 5 Schematic diagram of chromosome encoding and decoding process shown.

[0222] Step 4.3.2: Evaluate the fitness of all chromosome individuals in the population.

[0223] In this embodiment, the method for evaluating the fitness of the chromosome individuals is:

[0224] Step 4.3.2.1: Decode the chromosome individuals to obtain the allocation rule tree, and convert the allocation rule tree into a regular arithmetic expression.

[0225] Step 4.3.2.2: For n generated by training sc The rules expressed by the regular arithmetic expression are used to allocate the tasks to be allocated in each scenario;

[0226] Step 4.3.2.3: Perform single-satellite task scheduling for the tasks assigned to a satellite; commonly used single-satellite task scheduling methods include heuristic (priority sorting, earliest executable time sorting and other rules), meta-heuristic (genetic, ant colony, etc.) scheduling methods, etc.

[0227] Step 4.3.2.4: Calculate the total revenue of all satellites after executing task scheduling according to the regular task allocation expressed by the regular arithmetic expression in a scenario;

[0228] Step 4.3.2.5: Until the total benefits after task scheduling in nsc scenarios are obtained and the average value is calculated, the average value is taken as the fitness evaluation value of an individual.

[0229] In this embodiment, fitness evaluation is the process of evaluating the evolution rules using test scenarios under the rules corresponding to the given chromosome individuals. Fitness evaluation largely determines the evolutionary direction of the population and must be consistent with the optimization goal of the problem. The function EvRule(·) is used to describe the arithmetic expression of the rule corresponding to each chromosome. The function outputs the score for each task i matching with satellite j. The decision-making method using the chromosome corresponding rule is:

[0230]

[0231] For the training generated n sc The allocation is done according to the rules expressed by the arithmetic expression. The allocation is done by using the sequence matching construction process as follows: Figure 7 As shown. After allocation, single satellite scheduling is used to obtain the scheduling benefit Φkj(·) of each satellite j in scenario k. For the fitness of the rule, the present invention uses the average benefit of the test scenario to define the fitness fu of rule u as shown in the following formula.

[0232]

[0233] Step 4.3.3: Perform individual selection, crossover, and mutation on the population to form a new population;

[0234] In this embodiment, the individual selection method is a tournament sorting method. When constructing a new parent population, the proportion of the population size is selected. el Keep the elites and randomly select n from the population tour Individuals with the largest fitness are copied and placed into the new parent generation. The selection of individuals in the population as the new parent generation of the next generation requires retaining elite individuals while retaining a certain number of inferior individuals to avoid falling into the local optimum. Figure 8 shown.

[0235] The crossover operator is designed to maintain population diversity, and it searches for rules by exchanging chromosome coding segments. The crossover operator in this embodiment uses:

[0236] The single-point crossover operator is based on the single-point crossover probability p c1 Determining whether crossover is needed is the classic 1-opt operator;

[0237] The two-point crossover operator is based on the two-point crossover probability p c2 Determining whether crossover is needed is the classic 2-opt operator;

[0238] Gene crossover operator, according to the gene crossover probability p ge Determine whether to select crossover.

[0239] The mutation operator is also used to search for rules and maintain population diversity. Compared with the crossover operator acting on two chromosomes, the mutation operator acts on a single chromosome. The mutation operator in this embodiment uses:

[0240] Single-point mutation operator: According to the single-point mutation probability p ms Determine whether mutation is needed;

[0241] The segment reversal operator randomly selects a coding segment in the chromosome for reversal, and the reversal probability p mi Determine whether mutation is needed;

[0242] The IS transposition operator randomly selects a segment from the coding segment and inserts it into the non-first position of another gene on the chromosome.is conduct;

[0243] The RIS transposition operator randomly selects a segment starting from the function set from the head of a gene and inserts it into the beginning position of the head of another gene.

[0244] Step 4.3.4: After satisfying the number of iterations, select the individual with the largest fitness as the output of the optimal allocation rule.

[0245] Experimental verification:

[0246] The present invention aims to design a fast, efficient and high-quality solution algorithm for the MSCTAP problem of non-time-sensitive mobile target tracking, so as to achieve a fast response to the dynamic changes of mobile targets and realize the rapid allocation of tasks on the satellite. The test platform is a desktop computer (CPU: Intel (R), Core (TM) i5-4460, main frequency 3.2GHz; RAM: 12GB).

[0247] The evolution process analysis is mainly to check the convergence of the GEP evolution process and the evolution results. During the evolution process, this experiment conducted regular evolution for three task scales: sc100-s4-t160, sc100-s7-t280, and sc100-s10-t400. These three scales represent the number of scenarios n. sc The scenario scales of 100 satellites, 4 satellites, and 160 tasks to be assigned; 7 satellites, 280 tasks to be assigned; and 10 satellites, 400 tasks to be assigned, represent the rule evolution of small, medium, and large-scale scenarios, respectively. During the evolution process, the algorithm parameter settings are shown in Table 1.

[0248] Through the fitness evaluation scenarios of three scale settings, the experiment counted the maximum, minimum and average fitness of each generation of individuals in the evolution process, and obtained the final statistical results of the evolution process as follows: Fig. 9 As shown, from Fig. 9It can be seen that the fitness values ​​of the three scales of scenes rise very fast in the first 50 generations, indicating that the algorithm has a significant effect on improving rule evolution in the early stage. For small-scale scenes, the evolution gradually slows down after the number of generations reaches 50, converges around 200 generations, and then shows a stable trend of stopping growth. Although the medium and large scales tend to be stable after 100 generations, they increase slightly after 350 and 300 generations respectively, and then stop growing and tend to be stable. During the entire evolution process, the average fitness of the population maintains the same trend as the maximum value, but the minimum value is basically in a fluctuating stage, which is also conducive to maintaining population diversity and avoiding falling into local optimality. Through experimental analysis, it can be seen that both convergence and trend consistency of evolution rules are guaranteed in the GEP algorithm. The evolution effect of GEP on solving the rules of the MSCTAP problem for non-time-sensitive mobile target tracking is obvious.

[0249] Table 1 Parameters of GEP evolution process

[0250] Parameter name Parameter Value Parameter name Parameter Value <![CDATA[Population size n pop > 50 <![CDATA[Evolutionary algebra n iter > 600 <![CDATA[Number of fitness evaluation scenarios n sc > 100 <![CDATA[Single-point crossover probability p cl > 0.7 <![CDATA[Double - point crossover probability p c2 > 0.7 <![CDATA[Gene crossover probability p ge > 0.7 <![CDATA[Single-point crossover probability p ms > 0.05 <![CDATA[Fragment inversion probability p mi > 0.1 <![CDATA[IS transposition probability p is > 0.1 <![CDATA[RIS transposition probability p ris > 0.1 <![CDATA[Elite individual retention ratio pro el > 0.1 <![CDATA[Sorted extraction of the number of individuals n tour > 3

[0251] The satellite passes over the mission for a period of time, and the mission can only be assigned to the satellite if there is a window during this period. Judging from the number of tasks assigned to the satellite in each beam of assigned task windows, except for the relatively small number of 33 assigned tasks for satellite 3 and 34 assigned tasks for satellite 9 in large-scale scenarios, the number of tasks assigned to satellites in other scale scenarios is basically maintained at around 40. This is not only uniform in distribution, but also reduces the conflict between tasks in single-satellite task scheduling for each satellite. If the distribution lacks uniformity, there will inevitably be a high conflict on a certain satellite, the success rate of task scheduling will be reduced, and the overall benefits of multi-satellite task planning will be reduced. In addition, in terms of the uniformity of the distribution of task windows, the distribution of windows for assigning tasks on each satellite is relatively uniform, and there is no phenomenon of excessive accumulation in a certain section of the satellite window. This shows that when executing the allocation, the evolutionary rule will try to reduce the overlap between the windows of tasks and the assigned satellites and reduce the conflict between tasks. This trend in task allocation obviously increases the possibility of successful task scheduling. In addition, this experiment counted the allocation time in three scale instance scenarios, which were 2.324s, 6.744s and 13.199s respectively. From the actual usage, the timeliness of the algorithm allocation meets the autonomous needs on the satellite and can respond to the uncertainty caused by the motion changes of non-time-sensitive mobile targets. The above analysis shows that the Constructed Heuristic Method for multi-satellite collaborative task assignment based on GEP evolution (CHMGEP) is feasible in solving the multi-satellite task allocation problem.

[0252] In addition, 50 examples were generated for the scenarios of scales s4-t160, s7-t280, and s10-t400 for testing, and the scheduling benefits (SP) and scheduling time (ST) in each scenario were finally counted. It should be noted that the scheduling time includes the multi-satellite allocation time and the single-satellite scheduling time, and the single-satellite scheduling time is serial (in the centralized-distributed collaborative architecture designed in the study, the single-satellite scheduling can be parallelized, so the actual application can greatly reduce the calculation time). Through experiments, the test results of 50 examples of three scales were obtained. In the small, medium and large scale scenario examples, the CHMGEP algorithm has achieved a comprehensive surpassing of the other three heuristic algorithms in terms of SP indicators. The three heuristic algorithms have their own advantages and disadvantages in the quality of solving problems, and the LVTD algorithm is the best. It can be seen that the CHMGEP algorithm has advantages over a single heuristic algorithm.

[0253] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A multi-satellite collaborative task allocation method for non-time-sensitive mobile targets, characterized in that: The following steps are involved: Step 1: Obtain the non-time-sensitive mobile target task set and the satellite set and state parameters involved in the planning; Step 2: The non-time-sensitive mobile target task set is annotated to the assigned main satellite, and the non-time-sensitive mobile target task set is shared with each sub-satellite in the satellite set. The assigned main satellite refers to the satellite with the longest transit time, and the sub-satellite refers to other non-main satellites that are about to transit; Step 3: Each satellite calculates the visible window of each task according to its own state and orbit information to obtain the visible window of each satellite for each task, and feeds back the calculation result of the visible window to the assigned main satellite; Step 4: Allocate the main satellite to the multi-satellite tasks according to the visible window calculation results of each satellite for each task fed back by each sub-satellite.

2. The distribution method according to claim 1, characterized in that: The method for allocating the main satellite in step 4 to allocate multi-satellite tasks according to the visible window calculation results fed back by each sub-satellite is to construct a multi-satellite collaborative task allocation model for non-time-sensitive mobile target tracking, solve the multi-satellite collaborative task allocation model, and obtain a multi-satellite collaborative task allocation plan.

3. The distribution method according to claim 2, characterized in that: The multi-satellite collaborative task allocation model is: Ω j ={tsk i |x ij =1} (4) Among them, n sat Indicates the number of satellites to be assigned, n tsk represents the number of tasks to be assigned, Φ(·) represents the scheduling function of satellite j, and returns the scheduling benefit, S j represents the state attribute of satellite j before allocation, Ω j represents the set of tasks assigned to satellite j, C j represents the set of constraints for satellite j, Ψ j represents the optimization target of satellite j; x ij is a 0-1 decision variable, 1 means the i-th task tsk i Assigned to satellite j; v ij Represents a 0-1 parameter, 1 means there is a visible window between task i and satellite j, 0 means there is no visible window between task i and satellite j; Indicates the mission imaging payload requirement, 0 for optical and 1 for SAR; Indicates the satellite payload type, 0 is optical and 1 is SAR; Task i imaging mode requirements, different working modes correspond to different widths; Represents the set of satellite j working modes.

4. The distribution method according to claim 3, characterized in that: The method for solving the multi-satellite collaborative task allocation model is: Step 4.1: Filter the set of satellites that can be assigned to each task through constraints from the calculation results of each satellite's visible window for each task; Step 4.3: Use the allocation rule to select the best satellite from the set of satellites that can be allocated to each task in turn to allocate to the task.

5. The distribution method according to claim 4, characterized in that: The allocation rule is obtained based on the multi-star allocation rule evolution method of GEP, specifically: Step 4.3.1: Randomly generate an initial population, which contains n pop Chromosomes, each chromosome contains several coding genes, each coding gene consists of a head and a tail. The first code in the head must be a function set, and the other coding bits in the head can be function sets or terminal sets. The tail must be a terminal set. Step 4.3.2: Evaluate the fitness of all chromosome individuals in the population; Step 4.3.3: Perform individual selection, crossover, and mutation on the population to form a new population; Step 4.3.4: After satisfying the number of iterations, select the individual with the largest fitness as the output of the optimal allocation rule.

6. The distribution method according to claim 5, characterized in that: The function set is +, -, *, sin, cos, max, min, and the endpoint set is the feature vector FV that characterizes the multi-satellite collaborative task allocation problem. ij , 1) TN j It represents the proportion of the assigned tasks on satellite j to the total assigned tasks of all satellites; 2)RE j The remaining power percentage of satellite j; Egy max Indicates the maximum power of the satellite. Indicates the remaining power of the satellite; 3) AP j represents the mean priority of the tasks assigned to satellite j, After normalization, it is expressed as: pri max Indicates the highest task priority; 4) SP j represents the standard deviation of the priority of the tasks assigned to satellite j; After normalization, it is expressed as: 5) AR j represents the average value of the priority-imaging duration ratio of the assigned tasks on satellite j; After normalization: 6)SR j represents the priority-imaging duration ratio standard deviation of the assigned tasks on satellite j; 7) AO j represents the mean time of the overhead of the assigned tasks on satellite j; After normalization: 8) SO j represents the standard deviation of the time when the assigned task passes over the top of satellite j; Adopting relative planning period T plan Normalize it and get the normalized 9) WP i Indicates the priority of task i to be assigned; After normalization, it is expressed as: 10) WR i It represents the priority of task i to be assigned - imaging duration; After normalization: 11) WO ij It indicates the time when satellite j passes through the top of the assigned task i; Through the planning cycle T plan Normalize it and get the normalized 12)WL ij It indicates the proportion of visible window length of satellite j executing the task i to be assigned; WL ij =we ij -wb ij ; After normalization: 13)WC ij represents the conflict degree between the task to be assigned i and the task to be planned for satellite j; Toilets ij =Cont ij ; After normalization: 14)WV ij represents the length of the non-overlapping time period in the visible window corresponding to the task to be assigned i and the visible time windows of other assigned tasks on satellite j; The function ||·|| represents the sum of the lengths of the window segments in the set; After normalization: 15)AS ij represents the statistical mean of the observed slopes of the task to be assigned i and the assigned tasks on satellite j; The observed slope sp between task i and task l il It can be defined as: After normalization: 16) SS ij represents the standard deviation of the slope of observations between the task to be assigned i and the assigned task on satellite j; After normalization:

7. The distribution method according to claim 6, characterized in that: The method for evaluating the fitness of the chromosome individuals in step 4.3.2 is: Step 4.3.2.1: Decode the chromosome individuals to obtain the allocation rule tree, and convert the allocation rule tree into a regular arithmetic expression; Step 4.3.2.2: For the nsc scenarios generated by training, use the rules expressed by the regular arithmetic expression to assign the tasks to be assigned to each scenario; Step 4.3.2.3: Perform single-satellite task scheduling for the tasks assigned to a satellite; Step 4.3.2.4: Calculate the total revenue of all satellites after executing task scheduling according to the regular task allocation expressed by the regular arithmetic expression in a scenario; Step 4.3.2.5: Until you get n sc The total benefits after task scheduling in each scenario are averaged and the average value is used as an individual fitness evaluation value.

8. The distribution method according to claim 6, characterized in that: The individual selection in step 4.3.3 uses the tournament sorting method. When constructing the new parent population, the proportion of the population size is pro el Keep the elites and randomly select n from the population tour Individuals, copy the one with the largest fitness and put it into the new parent generation; The operators of crossover mutation are: single-point crossover operator, according to the single-point crossover probability p c1 Determine whether to select crossover; double-point crossover operator, according to the double-point crossover probability p c2 Determine whether crossover is needed; gene crossover operator, according to the gene crossover probability p ge Determine whether to select crossover; The mutation operators are: single-point mutation operator according to the single-point mutation probability p ms To determine whether mutation is needed, the segment reversal operator randomly selects a coding segment in the chromosome for reversal, which is based on the segment reversal probability p mi Determine whether mutation is needed; the IS transposition operator randomly selects a segment from the coding segment and inserts it into the non-first position of another gene on the chromosome. is The RIS transposition operator randomly selects a segment starting from the function set from the head of a gene and inserts it into the starting position of the head of another gene.

9. A multi-satellite collaborative task allocation system for non-time-sensitive mobile targets, characterized in that: Includes the following modules: Information acquisition module: used to obtain the non-time-sensitive mobile target mission set and the satellite set involved in the planning; Information annotation module: used to annotate the non-time-sensitive mobile target task set obtained by the information acquisition module to the assigned main satellite, and share the non-time-sensitive mobile target task set with each sub-satellite. The assigned main satellite refers to the satellite with the longest transit time, and the sub-satellite refers to other non-main satellites that are about to transit; Mission visible window calculation module: used for each satellite in the satellite set to calculate the visible window of each mission in the mission set according to its own status and orbit information to obtain the visible window of each satellite for each mission, and feed back the visible window calculation result to the assigned main satellite; Task allocation module: used to allocate multi-satellite tasks to the main satellite based on the visible window calculation results of each satellite for each task fed back by each sub-satellite.

Images