A method and device for hybrid optimization of packing branches for multi-spacecraft-to-multi-satellite traversal rendezvous

By decomposing the multi-spacecraft traversal and rendezvous task planning problem into outer target allocation and inner transfer trajectory planning, the tree search framework and global optimization operators are used to solve the computational complexity of multi-spacecraft mission planning in the existing technology. The problem of sexual and local optimality achieves better task schemes and more efficient calculations.

CN118798581BActive Publication Date: 2024-11-26NAT UNIV OF DEFENSE TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411265313.X
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-09-10
Publication Date
2024-11-26
Estimated Expiration
2044-09-10

AI Technical Summary

Technical Problem

The existing technology is difficult to coordinate the optimization of the target allocation, rendezvous sequence and time variables in multi-spacecraft traversal rendezvous missions under acceptable computational volume, resulting in huge solution space and large local optimality, lacking effective global optimal solution. method.

Method used

Decompose the multi-spacecraft-to-multi-star traversal rendezvous task planning problems into outer layer planning problems and inner layer transfer trajectory planning problems. By modeling the spacecraft as a mission box and discrete time points as the basic unit, the problem is transformed into boxing combination optimization Problem, use the tree search framework and global optimization operator to adjust the target distribution, and optimize the time variables through the gradient method to obtain the optimal intersection time.

Benefits of technology

It realizes the better task scheme with a smaller calculation amount, fully explores the time domain and the combined domain, and improves the optimality and efficiency of the task scheme.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118798581B_ABST
    Figure CN118798581B_ABST
Patent Text Reader

Abstract

The present application relates to a box-packing branch hybrid optimization method and device for multi-spacecraft multi-satellite traversal rendezvous. The method includes: decomposing the multi-spacecraft multi-satellite traversal rendezvous mission planning problem into outer target allocation, rendezvous sequence and time planning problems and inner transfer trajectory planning problems; modeling the spacecraft as a mission box, and discretizing the flight mission timeline into evenly distributed time points as the basic unit for placing targets; assigning targets to the mission boxes, and constructing a traversal rendezvous mission solution tree, the solution tree is integrated into the time domain representation, and the branches represent the optimal rendezvous sequence and time; adjusting and improving the target distribution between each task box and reconstructing the solution tree through global optimization operators such as movement, permutation, and redistribution; using the optimized rendezvous time point as the initial value, and accurately optimizing the time variable through the gradient method; based on the precise time when each spacecraft rendezvouses each target, solving the inner optimal end-to-end transfer trajectory respectively. A better mission plan can be obtained by using this method.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the technical field of aerospace mission planning, and in particular to a method and device for hybrid optimization of box-packing branches for multi-spacecraft-to-multi-star traversal rendezvous. Background Art

[0002] For tasks such as space multi-debris removal, multi-asteroid exploration, and constellation on-orbit service, multiple spacecraft can cooperate to traverse and rendezvous with a large number of space targets within a given time to achieve batch access to large-scale targets, such as Figure 1 As shown. Designing such a mission requires global optimization of target allocation, rendezvous sequence and time, and the transfer trajectory of the spacecraft between targets to minimize the flight cost or flight time. This problem is a type of vehicle path planning problem with time-varying state characteristics under orbital dynamics. The position and velocity of space targets change continuously with time, and the transfer cost between targets is a continuous function of time. Target allocation and rendezvous sequence planning are combinatorial optimization problems involving discrete variables; transfer time and trajectory planning are nonlinear optimization problems involving continuous variables. The mixture of the two results in a huge solution space and many local optima.

[0003] Traditional solution strategies are divided into two categories. The first category is the multi-stage decomposition strategy, which decomposes the entire solution process into multiple stages such as target allocation, sorting, time scheduling and trajectory planning, and uses the simplified model of the next stage for approximate solution in each stage; the error caused by model approximation restricts the global optimality of the solution. The second category is the target state staticization strategy, which transforms the time-varying problem into a static problem by locally optimizing the time variable; the mixture of global combinatorial optimization and local time optimization causes excessive computational complexity and low solution efficiency; insufficient exploration of the time domain limits the optimality of the obtained solution. At present, there is still a lack of a method to comprehensively optimize all variables through a single solution process, fully explore the time domain during the optimization process, and ensure the optimality of allocation, sequence and time variables at the same time with an acceptable amount of computation. Summary of the invention

[0004] Based on this, it is necessary to provide a box-packing branch hybrid optimization method and device for multi-spacecraft to multi-star traversal rendezvous in response to the above-mentioned technical problems.

[0005] A box-packing branch hybrid optimization method for multi-spacecraft to multi-satellite traversal rendezvous, the method comprising:

[0006] The multi-spacecraft multi-star traversal rendezvous mission planning problem is decomposed into an outer planning problem and an inner transfer trajectory planning problem, and a mathematical model of the two-level planning problem is constructed; the outer planning problem is the problem of target allocation, rendezvous sequence and time planning.

[0007] The spacecraft is modeled as a mission box, and the flight mission timeline of a single spacecraft is discretized into uniformly distributed time points. Discrete time points are used as the basic units for placing targets, and the complex discrete-continuous mixed optimization problem is transformed into a simple box packing combinatorial optimization problem.

[0008] The targets to be rendezvoused are randomly assigned to task boxes. For each task box, a solution tree of the single spacecraft-to-multi-target rendezvous mission plan in each task box is constructed. The time domain representation is integrated into the traditional tree search framework so that the branches in the solution tree can simultaneously represent the optimal rendezvous sequence and time; the solution tree contains multiple high-quality boxing solution branches.

[0009] The target distribution between each task box is improved by adjusting the global optimization operator; the solution tree of the adjusted task box is reconstructed by deleting operators and inserting operators to obtain the intersection time point of each target in each task box; the global optimization operators include: moving operator, permutation operator, and redistribution operator.

[0010] The rendezvous time point of each target in each mission box is taken as the initial value, and the time variable is accurately optimized through the gradient method to obtain the optimal time for each spacecraft to rendezvous with each target in the continuous parameter domain.

[0011] Based on the optimal time for each spacecraft to rendezvous with each target in the continuous parameter domain, the original problem is transformed into multiple independent inner orbit transfer trajectory planning problems, and the optimal end-to-end transfer trajectory is solved separately; the original problem is a global optimization problem of multi-spacecraft-to-multi-satellite traversal rendezvous trajectory.

[0012] A box-packing branch hybrid optimization device for multi-spacecraft to multi-satellite traversal rendezvous, the device comprising:

[0013] The modeling module of the two-level planning problem is used to decompose the multi-spacecraft multi-star traversal rendezvous mission planning problem into an outer planning problem and an inner transfer trajectory planning problem, and to construct a mathematical model of the two-level planning problem; the outer planning problem is the problem of target allocation, rendezvous sequence and time planning.

[0014] The problem conversion module is used to model the spacecraft as a mission box and discretize the flight mission timeline of a single spacecraft into evenly distributed time points. The discrete time points are used as the basic units for placing targets, and the complex discrete-continuous mixed optimization problem is converted into a simple bin packing combinatorial optimization problem.

[0015] The solution tree construction module of the multi-target rendezvous mission plan is used to construct the solution tree of the multi-target rendezvous mission plan for a single spacecraft in each mission box, integrating the time domain representation into the traditional tree search framework so that the branches in the solution tree can simultaneously represent the optimal rendezvous sequence and time; the solution tree contains multiple high-quality boxing solution branches.

[0016] The target packing distribution optimization and adjustment module improves the target distribution between each task box through global optimization operator adjustment; reconstructs the solution tree of the adjusted task box by deleting operators and inserting operators to obtain the intersection time point of each target in each task box; the global optimization operators include: moving operator, permutation operator, and redistribution operator.

[0017] The optimal rendezvous time determination module is used to take the rendezvous time point of each target in each mission box as the initial value, accurately optimize the time variable through the gradient method, and obtain the optimal time for each spacecraft to rendezvous with each target in the continuous parameter domain.

[0018] The optimal end-to-end transfer trajectory solving module is used to transform the original problem into multiple independent inner orbit transfer trajectory planning problems based on the optimal time for each spacecraft to rendezvous with each target in the continuous parameter domain, and solve the optimal end-to-end transfer trajectory respectively; the original problem is the global optimization problem of multi-spacecraft to multi-star traversal rendezvous trajectory.

[0019] The above-mentioned bin-packing branch hybrid optimization method and device for multi-spacecraft multi-satellite traversal rendezvous transforms the complex discrete-continuous hybrid optimization problem into a simple bin-packing combinatorial optimization problem. Through a tree search method that incorporates time domain representation, the combinatorial domain and the time domain are fully explored at the same time, and the target allocation, rendezvous sequence and time variables can be comprehensively optimized. First, the multi-spacecraft multi-satellite traversal rendezvous mission planning problem is decomposed into the outer target allocation, rendezvous sequence and time planning problem and the inner transfer trajectory planning problem. Secondly, the spacecraft is modeled as a task box, and the flight mission timeline is discretized into uniformly distributed time points as the basic unit for placing targets. Then, the targets are assigned to the task boxes, and the traversal rendezvous mission solution tree is constructed. The solution tree incorporates the time domain representation, and the branches simultaneously represent the optimal rendezvous sequence and time; the target distribution between each task box is adjusted and improved through global optimization operators such as movement, permutation, and redistribution, and the solution tree is reconstructed. Finally, the optimized rendezvous time point is used as the initial value, and the time variable is accurately optimized through the gradient method; based on the precise time of rendezvous of each spacecraft and each target obtained in the outer layer, the optimal end-to-end transfer trajectory of the inner layer is solved respectively. This method can obtain a better mission plan than the existing method. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 This is a schematic diagram of a multi-spacecraft multi-satellite traversal rendezvous mission;

[0021] Figure 2 A flow chart of a bin-packing branch hybrid optimization method for multi-spacecraft to multi-satellite traversal rendezvous in another embodiment;

[0022] Figure 3 It is a schematic diagram of inner and outer layer decomposition of a multi-spacecraft to multi-star traversal rendezvous mission planning problem in another embodiment;

[0023] Figure 4 A schematic diagram of converting a discrete-continuous mixed optimization problem into a combinatorial packing problem in another embodiment;

[0024] Figure 5 A flow chart of a bin-packing branch hybrid optimization method in another embodiment;

[0025] Figure 6 A schematic diagram of a single-task box solution tree incorporating time domain representation in another embodiment;

[0026] Figure 7 Schematic diagram of tree reconstruction after deleting and inserting targets in another embodiment, wherein (a) is a schematic diagram of deleting target 3, and (b) is a schematic diagram of inserting target 5;

[0027] Figure 8 A schematic diagram of redundant target reallocation in another embodiment;

[0028] Fig. 9 It is a structural block diagram of a box-packing branch hybrid optimization device for multi-spacecraft multi-star traversal rendezvous in one embodiment. DETAILED DESCRIPTION

[0029] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0030] In one embodiment, Figure 2 As shown, a box-packing branch hybrid optimization method for multi-spacecraft to multi-satellite traversal rendezvous is provided, and the method comprises the following steps:

[0031] Step 100: Decompose the multi-spacecraft to multi-star traversal rendezvous mission planning problem into an outer planning problem and an inner transfer trajectory planning problem, and construct a mathematical model of the two-layer planning problem; the outer planning problem is the target allocation, rendezvous sequence and time planning problem.

[0032] Specifically, Figure 3 As shown. The spacecraft is The goal is are the position and velocity of the spacecraft, are the position and velocity of the target respectively.

[0033] The inner transfer trajectory planning problem is a nonlinear trajectory optimization problem, which is the optimization and evaluation of the end-to-end transfer flight trajectory; the outer planning problem is a time-varying vehicle path planning problem, which requires determining all the departure and arrival times of each spacecraft transferring between their respective targets.

[0034] Step 102: Model the spacecraft as a mission box, and discretize the flight mission timeline of a single spacecraft into evenly distributed time points. Use the discrete time points as the basic unit for placing targets, and transform the complex discrete-continuous hybrid optimization problem into a simple bin packing combinatorial optimization problem.

[0035] Specifically, the spacecraft is modeled as a task box, and the flight mission timeline of a single spacecraft is discretized into uniformly distributed time points as the basic unit for placing targets, thereby transforming the outer planning problem with discrete-continuous mixed characteristics into a box packing problem with pure combinatorial characteristics, such as Figure 4 shown.

[0036] The original problem is to locate the target at a certain moment on a task timeline. The transformed problem is to group the targets into task boxes, build a solution tree containing the optimal task solution branch in each box, and improve the global optimality by continuously adjusting the target distribution between task boxes and reconstructing the solution tree.

[0037] Steps 104 to 110 are the detailed process of the binning branch hybrid optimization method, as shown in FIG. Figure 5 shown.

[0038] Step 104: Randomly assign the targets to be rendezvoused to task boxes. For each task box, construct a solution tree of the single spacecraft-to-multi-target rendezvous mission plan in each task box. Integrate the time domain representation into the traditional tree search framework so that the branches in the solution tree can simultaneously represent the optimal rendezvous sequence and time. The solution tree contains multiple high-quality boxing solution branches.

[0039] For the targets to be rendezvoused in a mission box, a packing solution tree is constructed. Each branch of the solution tree represents a solution for placing each target on the mission timeline. The traversal rendezvous mission planning problem of a single spacecraft to multiple satellites can be modeled as a time-varying traveling salesman problem under orbital dynamics. The tree model is a classic model for representing the solution space of the traveling salesman problem. In order to adapt to the time-varying characteristics of the orbital target state, a time domain branch is introduced on the basis of the sequence branch, such as Figure 6 As shown. Each node in the solution tree It not only contains the information of the rendezvous sequence S, but also contains a variety of high-quality rendezvous time plans corresponding to the sequence S. Information and timetable Corresponding cost indicators. For example, Figure 6 In the , expand the node, time plan The characters “→2, →300.0 days, 73.3 kg” in the figure indicate that the spacecraft spent 210 days and 73.3 kg of fuel to transfer from target 3 to target 2. .

[0040] Step 106: Adjust and improve the target distribution between each task box through the global optimization operator; reconstruct the solution tree of the adjusted task box by deleting the operator and inserting the operator to obtain the intersection time point of each target in each task box; the global optimization operator includes: moving operator, permutation operator, and redistribution operator.

[0041] Step 108: Taking the rendezvous time point of each target in each mission box as the initial value, the time variable is accurately optimized by the gradient method to obtain the optimal time for each spacecraft to rendezvous with each target in the continuous parameter domain.

[0042] Specifically, in this embodiment, a Sequential Quadratic Programing (SQP) algorithm is used to accurately optimize the time.

[0043] Step 110: Based on the optimal time for each spacecraft to rendezvous with each target in the continuous parameter domain, the original problem is transformed into multiple independent inner orbit transfer trajectory planning problems, and the optimal end-to-end transfer trajectory is solved separately; the original problem is a global optimization problem of multi-spacecraft to multi-star traversal rendezvous trajectory.

[0044] In the above-mentioned bin-packing branch hybrid optimization method for multi-spacecraft multi-satellite traversal rendezvous, this method transforms the complex discrete-continuous hybrid optimization problem into a simple bin-packing combinatorial optimization problem. Through a tree search method that incorporates time domain representation, the combinatorial domain and the time domain are fully explored at the same time, and the target allocation, rendezvous sequence and time variables can be comprehensively optimized. First, the multi-spacecraft multi-satellite traversal rendezvous mission planning problem is decomposed into the outer layer target allocation, rendezvous sequence and time planning problem and the inner layer transfer trajectory planning problem. Secondly, the spacecraft is modeled as a task box, and the flight mission timeline is discretized into uniformly distributed time points as the basic unit for placing targets. Then, the targets are assigned to the task boxes, and the traversal rendezvous mission solution tree is constructed. The solution tree incorporates the time domain representation, and the branches simultaneously represent the optimal rendezvous sequence and time; the target distribution between each task box is adjusted and improved through global optimization operators such as movement, permutation, and redistribution, and the solution tree is reconstructed. Finally, the optimized rendezvous time point is used as the initial value, and the time variable is accurately optimized by the gradient method; based on the precise time of each spacecraft rendezvous with each target obtained in the outer layer, the inner layer optimal end-to-end transfer trajectory is solved respectively. This method can be used to obtain a better task solution than the existing methods.

[0045] In one embodiment, step 100 includes: the inner layer transfer trajectory planning problem is a nonlinear trajectory optimization problem, and the design variables are the maneuvers of the spacecraft over time:

[0046] (1)

[0047] in, T is the maximum thrust amplitude constant, is the spacecraft thrust vector, is the normalized thrust magnitude variable, is the thrust direction unit vector variable, , is the time variable.

[0048] In the inner transfer trajectory planning problem, the starting mass of the spacecraft is , departure star Track and departure time , Arrival at the rendezvous star The trajectory and arrival time of Fixed, the objective function is to minimize fuel consumption:

[0049] (2)

[0050] in, is the fuel consumption, is the spacecraft engine specific impulse, is the gravitational acceleration constant of the central body;

[0051] If both the transfer fuel and the transfer time are considered, the objective function of the inner transfer trajectory planning problem is:

[0052] (3)

[0053] in, is the objective function of the inner transfer trajectory planning problem, is the end-to-end evaluation function, is the transfer duration, and are the weight coefficients corresponding to the transfer fuel and transfer time, respectively, and their values ​​depend on the maneuverability of the spacecraft, the mission constraints and the definition of the objective function in the specific problem. Here, the weighted sum is used as the function Simple and effective example, but not limiting Definition of .

[0054] The constraints of the inner transfer trajectory planning problem are:

[0055] (4)

[0056] (5)

[0057] (6)

[0058] in, is the upper limit of fuel for a single end-to-end transfer, and They are the lower and upper limits of a single end-to-end transfer duration, and are the terminal position and velocity of the spacecraft in this transfer phase, and are the terminal position and velocity of the target star, and are the maximum permissible terminal position and velocity errors respectively.

[0059] The outer planning problem is a time-varying vehicle path planning problem, which requires determining all the departure and arrival times of each spacecraft transferring between their respective targets. From the perspective of the target, the design variables of the outer planning problem are expressed as:

[0060] (7)

[0061] in, represents the design variable, Indicates the target number, Represents the intersection target The spacecraft number, Represents spacecraft Intersection Target moment, and Represents spacecraft The start and end time of the entire sequential rendezvous mission; In order to decouple the discrete sequence combination information and the continuous flight time information in the design space, from the perspective of the spacecraft mission, the design variables in formula (7) can be transformed into:

[0062] (8)

[0063] in, represents the design variable after deformation, Represents spacecraft The sequence number of the rendezvous target, Represents spacecraft The total number of targets encountered, Represents spacecraft The first The number of the target, Represents spacecraft No. The duration of the segment transfer; the design variables in formula (7) and the design variables in formula (8) The conversion relationship between them is:

[0064] (9)

[0065] (10)

[0066] Objective function of the outer programming problem is the sum of the objective function values ​​of all spacecraft missions; the objective function of the outer planning problem is:

[0067] (11)

[0068] (12)

[0069] (13)

[0070] (14)

[0071] in, is the objective function of the outer programming problem, It's a spacecraft The objective function of the task, and Spacecraft Fuel consumption and mission time, where function The weighted sum function used is also just a simple and effective example and does not limit Definition of; Represents spacecraft The sequence number of the rendezvous target, and Spacecraft No. The fuel consumption and duration of the segment transfer, It's a spacecraft Total number of targets encountered;

[0072] The constraints of the outer planning problem are:

[0073] (15)

[0074] (16)

[0075] in, It's a spacecraft The upper limit constraint of fuel consumption is It's a spacecraft The upper limit of task time is constrained.

[0076] The global optimization problem of the multi-spacecraft cooperation to rendezvous with a large number of asteroids and mine resources in the 12th Global Trajectory Optimization Competition (GTOC-12) is used as an example. The details and data of the problem can be found at https: / / gtoc12.tsinghua.edu.cn / competition. The mission objective function of this example is

[0077] (17)

[0078] in, It is from The mass of ore collected from an asteroid is defined as 10 kg of ore mined per year of mining:

[0079] (18)

[0080] in, is the moment when the spacecraft rendezvouses the asteroid and recovers the minerals, which is the solution variable in the problem; is the moment when the miner starts working, and is a given constant in the problem. This embodiment uses the spacecraft and asteroids involved in the results submitted by the Jet Propulsion Laboratory (JPL), the champion of the competition, as test examples, including 35 spacecraft and 313 target asteroids.

[0081] In the outer layer optimization problem, the method in reference 1 (Zhu, YH, and Luo, YZ, “FastEvaluation of Low-Thrust Transfers via Multilayer Perceptions.” Journal ofGuidance, Control, and Dynamics, Vol. 42, No. 1, 2019, pp. 2627–2637.) is used as the function in formula (2) g, estimate the end-to-end trajectory transfer fuel consumption value. In the inner optimization problem, the method in reference 2 (Jiang, FH, Baoyin, HX, and Li, JF, “Practical Techniques for Low-Thrust Trajectory Optimization with Homotopic Approach.” Journal of Guidance,Control, and Dynamics, Vol. 35, No. 1, 2012, pp. 245–258.) is used to accurately optimize the end-to-end transfer trajectory.

[0082] In one embodiment, step 104 includes: Discretize the mission timeline of a single spacecraft into evenly distributed time points; for each mission box, initialize the initial state of the single spacecraft as the root node of the solution tree, and initialize the set of targets to be intersected; expand to the next layer of nodes in the solution tree, add orbit transfer for a certain target; traverse all remaining targets to be intersected; traverse each target to be intersected All discrete transfer times of , calculate the fuel consumption , spacecraft fuel consumption , spacecraft mission time and the objective function of the spacecraft mission ; For the same terminal moment, only the objective function of the spacecraft mission is retained The best time plan If the number of time plans is greater than the preset value, the optimal one is retained. time plan; among them, Represents an integer greater than 0; delete the target to be intersected from the target set to be intersected ; If there are too many single-layer nodes, keep the best nodes; among them, represents an integer greater than 0; if the total flight time or fuel consumption exceeds the upper limit, or the target set to be rendezvoused is empty, the loop traversal is stopped to obtain the optimal rendezvous sequence and time plan solution tree.

[0083] Specifically, for each rendezvous time plan , calculate the total fuel consumption of formula (13) and the total flight time of formula (14) according to the fuel consumption results and timeline results. Calculate the time plan through formula (12) The objective function value of .node The objective function value of Defined as the optimal value of all time schemes in this node, node The objective function value of for:

[0084] (19)

[0085] To fully explore the time domain, Discretize the mission timeline of a single spacecraft into evenly distributed time points. When expanding a new rendezvous target from a node, traverse the calculation to is the total plan for the transfer duration. represents the floor operator, In a certain node, for the time plan with the same terminal time, only the indicator value is retained The optimal time solution. This discrete strategy transforms the continuous nonlinear programming problem into a discrete combinatorial problem, which can roughly traverse the time domain and find the approximate location of the optimal time solution in the solution space. The obtained discrete time is used as the initial value of the gradient-based optimization algorithm, which converges to the exact result in the continuous domain.

[0086] Algorithm 1 is a method for constructing a tree of solutions for the rendezvous sequence and time plan of a single spacecraft for multiple targets. Figure 6 The pseudo code of the algorithm for solving the tree is as follows:

[0087] Algorithm 1: Construct a solution tree for the rendezvous sequence and time plan of a single spacecraft for multiple targets;

[0088] Input: A single spacecraft s and its target set ;

[0089] Output: optimal intersection sequence and time solution tree;

[0090] 1: Initialize the root node with the spacecraft initial state s and initialize the target set to be intersected ;

[0091] 2: Expand to the next layer of nodes in the solution tree and add track transfer for a certain target;

[0092] 3: Traverse all remaining targets to be intersected ;

[0093] 4: Traverse all discrete transfer durations ;

[0094] 5: Calculation and ;

[0095] 6: For the same terminal time, only keep The best time plan ;

[0096] 7: If there are too many time plans, only keep the best one Time plan ;

[0097] 8: From the target set to be intersected Delete a target ;

[0098] 9: If there are too many nodes in a single layer, only the best ones are retained. Nodes ;

[0099] 10: If the total flight time or fuel consumption exceeds the upper limit, or the target set to be rendezvoused is empty, stop the process.

[0100] In order to reduce the algorithm complexity of constructing the solution tree, at most Index value Better nodes; in each node, keep at most Index value Better time plan. This pruning strategy can avoid the exponential growth of algorithm complexity as the total number of targets and the number of solution trees increase. With this strategy, even if the total number of targets and the number of solution trees are large, solution trees with high-quality branches can be efficiently generated. Need to be configured for specific problems and ,Appropriate parameter configuration can make the algorithm both optimal and efficient.

[0101] In the embodiment, the solution tree construction algorithm parameters are configured for the GTOC-12 problem. Based on experience and a large number of test cases, the configuration parameters are shown in Table 1. This configuration can make the algorithm both optimal and efficient.

[0102] Table 1 Parameter configuration of tree construction algorithm

[0103]

[0104] In one embodiment, step 106 includes: in one round of iteration, adjusting the target distribution between each task box by moving operators and permuting operators, and reconstructing the solution tree of the adjusted task box by deleting operators and inserting operators; collecting the targets that cannot be loaded in each task box into a redundant set, and inserting the redundant targets into other task boxes by redistributing operators, and saving the optimal result; continuing to execute the next round of iteration, if the objective function of the outer planning problem is J If there is no change in multiple rounds of iterations, the iteration is terminated and the intersection time point of each target in each task box is obtained.

[0105] Specifically, Algorithm 2 gives the iterative optimization process of the bin-packing branch hybrid optimization method. The input of Algorithm 2 is kThe solution tree initialized in step 104 in the task box is output as k Spacecraft rendezvous n The optimal sequence and time of each target. The pseudo code of Algorithm 2 is as follows:

[0106] Algorithm 2: Iterative optimization process of bin-packing and branch-mixing optimization method;

[0107] Input: solution trees initialized in step 104 in k task boxes;

[0108] Output: The optimal sequence and time for k spacecraft to rendezvous with n targets;

[0109] 1: Iterative optimization of adjusting the target distribution between boxes and reconstructing the solution tree of each box;

[0110] 2: Adjust the target distribution between task boxes by moving and permuting operators;

[0111] 3: Reconstruct the solution tree of the adjusted task box by deleting and inserting operators;

[0112] 4: Collect the targets that cannot be loaded in each task box into the redundant set;

[0113] 5: Insert redundant targets into other task boxes by redistributing operators;

[0114] 6: Save the optimal result (k intersection sequences and n intersection moments);

[0115] 7: If indicator J does not change in multiple iterations, end the iteration.

[0116] In each round of iteration, the target distribution between each task box is adjusted by moving and permuting operators, and the solution tree of the adjusted task box is reconstructed by deleting and inserting operators. There will be redundant targets in the adjusted task box that cannot be loaded because they cannot meet the upper limit constraint of fuel consumption in formula (15) and the upper limit constraint of flight time in formula (16). Collect the redundant targets that cannot be arranged into feasible branches into the redundant set. Insert the targets in the redundant set into other more suitable task boxes through the redistribution operator. Repeat the above iterative optimization process of coordinating the target loading of each task box. If the total index is obtained J If a better result is found, save it.

[0117] In one embodiment, the target distribution between the task boxes is adjusted by moving operators and permuting operators, and the solution tree of the adjusted task boxes is reconstructed by deleting operators and inserting operators. The specific steps of moving operators in this step include:

[0118] Calculate the target loading density of each task box to ε - Greedy strategy selection task box A , that is, ε ∈[0,1], select the task box with the lowest target loading density, with a probability of 1- ε The probability of randomly sampling task boxes from each task box is A ; Task Box A is the source task box of the target moved by the mobile operator; the task box z Medium target loading density for:

[0119] (20)

[0120] in, It's a task box Medium target loading density, is the optimal time solution in a single task box The corresponding total fuel consumption is is the optimal time solution in a single task box The corresponding total flight time; and are the weight coefficients corresponding to the transfer fuel and transfer duration, It's a spacecraft The upper limit constraint of fuel consumption is It's a spacecraft The upper limit of task time is constrained.

[0121] Computational Task Box A The best time plan The objective function of the inner transfer trajectory planning problem for each segment of the transfer is J b2b ;pass ε -The greedy strategy selects the objective function of the inner transfer trajectory planning problem J b2b The biggest goal x Otherwise, from the task box A Random sampling target x ;Target x is the target moved by the mobile operator; calculate the task box A The target insertion evaluation values ​​for all other task boxes are:

[0122] (twenty one)

[0123] in, Q is the target insertion evaluation value of the task box, and A task box is inserted x The best time plan before and after The objective function value of .

[0124] If there is at least one task box with a positive target insertion evaluation value, ε - Greedy strategy selects the target with the highest evaluation value as the task box B , otherwise, randomly sample from the task box with a positive target insertion evaluation value; if there is no task box with a positive target insertion evaluation value, cancel this move operation.

[0125] Use the delete operator to remove the task box A Delete a target x And reconstruct the solution tree; use the insertion operator to insert the target x Insert to Taskbox B And reconstruct the solution tree.

[0126] Specifically, the mobile operator moves a task box A Target x Move to another task box B In essence, it is from A Lieutenant General x Delete and x insert B As shown in Algorithm 3.

[0127] Algorithm 3: Move a target x in a task box A to another task box B, and reconstruct the solution trees of task boxes A and B respectively;

[0128] Input: Each task box and the solution tree constructed in each task box;

[0129] Output: The solution tree reconstructed after task boxes A and B are moved by target x;

[0130] 1: Select task box A through ε-greedy strategy;

[0131] 2: By probability ε A , calculate the loading density of each task box defined by formula (20) , select the loading density The lowest one is used as task box A;

[0132] 3: Otherwise, randomly sample task box A from each task box;

[0133] 4: Select the target x in task box A through the ε-greedy strategy;

[0134] 5: Based on probability , calculate the optimal time plan in task box A defined by formula (3) The objective function of each transfer arc segment ,choose The worst target is taken as x;

[0135] 6: Otherwise, randomly sample target x from task box A;

[0136] 7: For all task boxes except task box A, calculate {Q} defined by equation (21);

[0137] 8: If there is a task box with a positive Q value;

[0138] 9: Selecting task box B through the ε-greedy strategy;

[0139] 10: By probability , select the one with the highest value Q as task box B;

[0140] 11: Otherwise, randomly sample task box B from each task box except task box A;

[0141] 12: Otherwise, cancel this move operation;

[0142] 13: Delete the target x from the task box A and reconstruct the solution tree through Algorithm 5;

[0143] 14: Insert the target x into task box B and reconstruct the solution tree through Algorithm 6.

[0144] In one embodiment, the target distribution between the task boxes is adjusted by moving operators and permuting operators, and the solution tree of the adjusted task boxes is reconstructed by deleting operators and inserting operators. The specific steps of permuting operators in this step include:

[0145] Select Task Box C and goals y , and select the task box in the mobile operator A and goals x The method is the same as that of C Each target and task box of each task box other than C Medium Target y The replacement benefit value is:

[0146] (twenty two)

[0147] in, is the revenue value, They are the optimal time plans for task boxes C and D before replacing the target. The objective function value of They are the optimal time plans for task boxes C and D after replacing the target. The objective function value of .

[0148] If there is at least one set of task boxes D and goals z combination( D , z )ofQ CD Positive, through ε - Greedy strategy chooses the replacement with the largest benefit value ( D , z ), otherwise, from is positive ( D , z ) is randomly sampled; if it does not exist is positive ( D , z ), cancel the replacement operation; use the delete operator to remove the task box C Delete a target y , from the task box D Delete a target z , and reconstruct the solution tree; by inserting the operator y Insert Task Box D The target z Insert Task Box C , and reconstruct the solution tree.

[0149] Specifically, the permutation operator converts a task box C Target y And another task box D Target z Exchange, in essence, from the task box C Lieutenant General y Delete from task box D Lieutenant General z Delete and y Insert Task Box D In z Insert Task Box C The permutation operator is shown in Algorithm 4, and its pseudo code is:

[0150] Algorithm 4: Replace a target x in a task box C with a target y in another task box D, and reconstruct the solution trees of task boxes C and D respectively;

[0151] Input: each task box and the solution tree constructed in each task box;

[0152] Output: The solution tree reconstructed after task boxes C and D are replaced by targets x and y;

[0153] 1: Select the task box C and the target x through lines 1-6 of Algorithm 3;

[0154] 2: For each target of the task box other than C, calculate the replacement benefit value defined by formula (22) ;

[0155] 3: If there is a corresponding Positive values ​​(D, z);

[0156] 4: Determine the task box D and target y through the ε-greedy strategy;

[0157] 5: Based on probability , select the replacement benefit value highest (D, z);

[0158] 6: Otherwise, the return value from the replacement Random sampling from (D, z);

[0159] 7: Otherwise, cancel this replacement operation;

[0160] 8: Delete target x from task box C and target y from task box D through Algorithm 5, and reconstruct the solution tree;

[0161] 9: Insert target x into task box D and target y into task box C through Algorithm 6, and reconstruct the solution tree.

[0162] Permutation operator selection task box C and goals y Method and mobile operator selection task box A and goals x The same method is used. Formula (22) is used to calculate the task box C Each target and task box of each task box other than C Medium Target y The replacement benefit value.

[0163] If there is at least one set of task boxes D and goals z combination( D , z )of Positive, through ε - Greedy strategy chooses the replacement with the largest benefit value ( D , z ), otherwise, from is positive ( D , z ) is randomly sampled; if it does not exist is positive ( D , z ), cancel the replacement operation; use the delete operator to remove the task box C Delete a target y , from the task box D Delete a target z , and reconstruct the solution tree; by inserting the operator y Insert Task Box D The target z Insert Task Box C , and reconstruct the solution tree.

[0164] In one embodiment, a delete operator is used to remove the task box. A Delete a target x And reconstruct the solution tree, including: traversing the task box from the first level to the last level A Corresponding solution tree, find the target x Corresponding node, delete the node and the lower branch of the node; if the target does not exist in the solution tree x If the corresponding node is found, the traversal ends.

[0165] In one embodiment, the target x Insert to Taskbox B Reconstruct the solution tree, including: traversing the task box from the first level to the last level B The corresponding solution tree traverses all nodes in the current layer and converts the target x Connect to the node and execute the steps of constructing the solution tree of the single spacecraft multi-target rendezvous mission plan in Algorithm 1, starting from the target x Start expanding branches; if a better branch is generated, update the best branch.

[0166] Specifically, the operators for deleting the target and reconstructing the solution tree in line 13 of Algorithm 3 and line 8 of Algorithm 4 and the operators for inserting the target and reconstructing the solution tree in line 14 of Algorithm 3 and line 9 of Algorithm 4 are as follows: Figure 7 As shown. The deletion operator and the insertion operator reconstruct the solution tree with the minimum complexity. Algorithm 5 gives the pseudo code of the deletion operator. For a deleted target, delete the relevant nodes in the original solution tree and all the branches under the nodes, as shown in Figure 7 (a) shows a schematic diagram of deleting target 3. Algorithm 6 gives the pseudo code of the insertion operator. Connect the newly inserted target to each node of the original solution tree, as follows Figure 7 (b) shows a schematic diagram of inserting target 5, and executing Algorithm 1 with the new node as the starting point to expand feasible and high-quality branches. In the process of reconstructing the solution tree after inserting the target, only branches are expanded on the new node, and only Indicators Better nodes.

[0167] The pseudo code of Algorithm 5 is:

[0168] Algorithm 5: Delete a target from a task box and reconstruct the solution tree;

[0169] Input: a task box and a selected target x in it;

[0170] Output: The reconstructed solution tree after deleting the target x;

[0171] 1: Traverse the solution tree from the first level to the last level;

[0172] 2: Find the node corresponding to the target x;

[0173] 3: Delete the node and its lower branches;

[0174] 4: If there is no node corresponding to the target x in the solution tree, the algorithm process ends.

[0175] The pseudo code of Algorithm 6 is:

[0176] Algorithm 6: Insert a certain goal into a certain task box and reconstruct the solution tree;

[0177] Input: a target x and a selected task box;

[0178] Output: The reconstructed solution tree after inserting the target x;

[0179] 1: Traverse the solution tree from the first level to the last level;

[0180] 2: Traverse all nodes in this layer;

[0181] 3: Connect the target x to the node;

[0182] 4: Execute Algorithm 1 and expand the branch starting from target x;

[0183] 5: If a better branch is generated, update the optimal branch.

[0184] After the target movement and replacement operations between boxes, it is necessary to check the fuel consumption upper limit constraint of formula (15) and the flight time upper limit constraint of formula (16), collect the redundant targets that cannot be arranged into the branch into the redundant set, and redistribute the redundant targets through the redistribution operator and insert them into other more suitable task boxes, such as Figure 8 shown.

[0185] Specifically, Algorithm 7 gives a redundant target reallocation algorithm. The pseudo code of Algorithm 7 is:

[0186] Algorithm 7: Redistribute the redundant targets that cannot be loaded in each box to other task boxes;

[0187] Input: redundant target x;

[0188] Output: Reassign x to a task box and reconstruct the solution tree;

[0189] 1: Traverse all task boxes and calculate the benefit value of inserting x;

[0190] 2: Calculate the optimal time branch after inserting x The objective function value of

[0191] 3: Calculate the insertion benefit value defined by equation (21);

[0192] 4: If there is at least one task box with a benefit value > 0, insert x into the task box with the largest benefit value through Algorithm 6.

[0193] In one embodiment, the targets that cannot be loaded in each task box are collected into a redundant set, and the redundant targets are inserted into other task boxes through a redistribution operator, including:

[0194] Collect the targets that cannot be loaded in each task box into the redundant set; traverse all task boxes and calculate the redundant targets to be inserted into the redundant set x Post-optimal time branch The objective function value is calculated based on the calculated result to insert redundant objectives x If there is at least one task box with a positive benefit value, the insertion operator is used to insert the redundant target into the task box with the largest benefit value; otherwise, the redundant target is left in the redundant set and the next round of iteration is entered. When the iterative optimization converges, the historical optimal solution in the optimization is taken.

[0195] It should be understood that although Figure 2 The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. Moreover, Figure 2 At least part of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.

[0196] In a verification example, for the same spacecraft initial state and rendezvous target, a solution with higher index than the JPL champion solution was obtained. The sequence and time of 35 spacecraft traversing and rendezvousing 313 asteroids in the obtained solution are shown in Tables 2 and 3 respectively. The total objective function value is 25279.4 kg, which is higher than JPL's 25190.1 kg. This result fully proves the optimality of the box-packing branch hybrid optimization method for multi-spacecraft multi-star traversal rendezvous proposed in this application in the design of multi-spacecraft cooperative traversal rendezvous mission.

[0197] Table 2 Number and sequence of rendezvous targets for each spacecraft in the solution obtained by this application method

[0198]

[0199] Table 3 The target time of rendezvous of each spacecraft and the value of the objective function in the solution obtained by this application method

[0200]

[0201] In one embodiment, Fig. 9 As shown, a packing branch hybrid optimization device for multi-spacecraft multi-satellite traversal rendezvous is provided, including: a modeling module for two-layer planning problems, a problem conversion module, a solution tree construction module for multi-objective rendezvous mission plans, a target packing distribution optimization adjustment module, an optimal rendezvous time determination module and an optimal end-to-end transfer trajectory solution module, wherein:

[0202] The modeling module of the two-level planning problem is used to decompose the multi-spacecraft multi-star traversal rendezvous mission planning problem into an outer planning problem and an inner transfer trajectory planning problem, and to construct a mathematical model of the two-level planning problem; the outer planning problem is the problem of target allocation, rendezvous sequence and time planning.

[0203] The problem conversion module is used to model the spacecraft as a mission box and discretize the flight mission timeline of a single spacecraft into evenly distributed time points. The discrete time points are used as the basic units for placing targets, and the complex discrete-continuous mixed optimization problem is converted into a simple bin packing combinatorial optimization problem.

[0204] The solution tree construction module of the multi-target rendezvous mission plan is used to construct the solution tree of the multi-target rendezvous mission plan for a single spacecraft in each mission box, integrating the time domain representation into the traditional tree search framework so that the branches in the solution tree can simultaneously represent the optimal rendezvous sequence and time; the solution tree contains multiple high-quality boxing solution branches.

[0205] The target packing distribution optimization and adjustment module improves the target distribution between each task box through global optimization operator adjustment; reconstructs the solution tree of the adjusted task box by deleting operators and inserting operators to obtain the intersection time point of each target in each task box; the global optimization operators include: moving operator, permutation operator, and redistribution operator.

[0206] The optimal rendezvous time determination module is used to take the rendezvous time point of each target in each mission box as the initial value, accurately optimize the time variable through the gradient method, and obtain the optimal time for each spacecraft to rendezvous with each target in the continuous parameter domain.

[0207] The optimal end-to-end transfer trajectory solving module is used to transform the original problem into multiple independent inner orbit transfer trajectory planning problems based on the optimal time for each spacecraft to rendezvous with each target in the continuous parameter domain, and solve the optimal end-to-end transfer trajectory respectively; the original problem is the global optimization problem of multi-spacecraft to multi-star traversal rendezvous trajectory.

[0208] In one embodiment, the modeling module of the two-level planning problem is also used to model the inner problem as a nonlinear trajectory optimization problem, and the design variables are the maneuvers of the spacecraft over time, as shown in formula (1).

[0209] In the inner transfer trajectory planning problem, the starting mass of the spacecraft is , departure star Track and departure time , Arrival at the rendezvous star The trajectory and arrival time of The objective function is to minimize the fuel consumption as shown in formula (2). If both the transfer fuel and the transfer time are considered, the objective function of the inner transfer trajectory planning problem is as shown in formula (3). The constraints of the inner transfer trajectory planning problem are shown in formulas (4) to (6).

[0210] The outer planning problem is a time-varying vehicle path planning problem, which requires determining all departure and arrival times of each spacecraft transferring between its respective targets. From the perspective of the target, the design variables of the outer planning problem are shown in formula (7). In order to decouple the discrete sequence combination information and continuous flight time information in the design space, from the perspective of the spacecraft mission, the design variables in formula (7) can be transformed into formula (8). The design variables in formula (7) are X and the design variables in formula (8) X The transformation relationship between 1 is shown in formula (9) and formula (10). The objective function of the outer planning problem is J is the sum of the objective function values ​​of all spacecraft missions; the objective function of the outer planning problem is shown in formulas (11) to (14); the constraints of the outer planning problem are shown in formulas (15) and (16).

[0211] In one embodiment, the solution tree construction module of the multi-target rendezvous mission plan is also used

[0212] By time interval Discretize the mission timeline of a single spacecraft into evenly distributed time points; for each mission box, initialize the initial state of the single spacecraft as the root node of the solution tree, and initialize the set of targets to be intersected; expand to the next layer of nodes in the solution tree, add orbit transfer for a certain target; traverse all remaining targets to be intersected; traverse each target to be intersected All discrete transfer times of , calculate the fuel consumption , spacecraft fuel consumption , spacecraft mission time and the objective function of the spacecraft mission ; For the same terminal moment, only the objective function of the spacecraft mission is retained The best time plan If the number of time plans is greater than the preset value, the optimal one is retained. time plan; among them, Represents an integer greater than 0; delete the target to be intersected from the target set to be intersected ; If there are too many single-layer nodes, keep the best nodes; among them, represents an integer greater than 0; if the total flight time or fuel consumption exceeds the upper limit, or the target set to be rendezvoused is empty, the loop traversal is stopped to obtain the optimal rendezvous sequence and time plan solution tree.

[0213] In one embodiment, the target packing distribution optimization adjustment module is also used to adjust the target distribution between each task box by moving operators and permutation operators in one round of iteration, and reconstruct the solution tree of the adjusted task box by deleting operators and inserting operators; collect the targets that cannot be loaded in each task box into a redundant set, and insert the redundant targets into other task boxes by redistributing operators to save the optimal result; continue to execute the next round of iteration, if the objective function of the outer planning problem J If there is no change in multiple rounds of iterations, the iteration is terminated and the intersection time point of each target in each task box is obtained.

[0214] In one embodiment, the specific steps of moving operators in the target packing distribution optimization adjustment module include: calculating the target loading density of each task box, ε - Greedy strategy selection task box A , that is, ε ∈[0,1], select the task box with the lowest target loading density, with a probability of 1- ε The probability of randomly sampling task boxes from each task box is A ; Task Box A is the source task box of the target moved by the mobile operator; the task box z Medium target loading density As shown in formula (20). Calculate the task box A The best time plan The objective function of the inner transfer trajectory planning problem for each segment of the transfer is ;pass ε -The greedy strategy selects the objective function of the inner transfer trajectory planning problem The biggest goal x Otherwise, from the task box A Random sampling target x ;Target x is the target moved by the mobile operator; Formula (21) is used to calculate the task box A All other task boxes except the target insert evaluation value Q If there is at least one task box with a positive target insertion evaluation value, ε - Greedy strategy selects the target with the highest evaluation value as the task box BOtherwise, randomly sample from the task box with a positive target insertion evaluation value; if there is no task box with a positive target insertion evaluation value, cancel this move operation. A Delete a target x And reconstruct the solution tree; use the insertion operator to insert the target x Insert to Taskbox B And reconstruct the solution tree.

[0215] In one embodiment, the specific steps of the permutation operator in the target bin distribution optimization adjustment module include: selecting a task bin C and goals y , and select the task box in the mobile operator A and goals x The method is the same as that of the task box; Formula (22) is used to calculate C Each target and task box of each task box other than C Medium Target y The replacement income value ; If there is at least one set of task boxes D and goals z combination( D , z )of Positive, through ε - Greedy strategy chooses the replacement with the largest benefit value ( D , z ), otherwise, from is positive ( D , z ) is randomly sampled; if it does not exist is positive ( D , z ), cancel the replacement operation; use the delete operator to remove the task box C Delete a target y , from the task box D Delete a target z , and reconstruct the solution tree; by inserting the operator y Insert Task Box D The target z Insert Task Box C , and reconstruct the solution tree.

[0216] In one embodiment, the deletion operator of the target packing distribution optimization adjustment module is also used to traverse the task boxes from the first layer to the last layer. A Corresponding solution tree, find the target x Corresponding node, delete the node and the lower branch of the node; if the target does not exist in the solution tree x If the corresponding node is found, the traversal ends.

[0217] In one embodiment, the insertion operator of the target packing distribution optimization adjustment module is also used to traverse the task boxes from the first layer to the last layer. B The corresponding solution tree traverses all nodes in the current layer and converts the target x Connect to the node and execute the steps of constructing the solution tree of the multi-target rendezvous mission solution for a single spacecraft. x Start expanding branches; if a better branch is generated, update the best branch.

[0218] In one embodiment, the target packing distribution optimization adjustment module is further used to collect the targets that cannot be loaded in each task box into a redundant set; traverse all task boxes, calculate and insert redundant targets into the redundant set x Post-optimal time branch The objective function value is calculated based on the calculated result to insert redundant objectives x If there is at least one task box with a positive benefit value, the insertion operator is used to insert the redundant target into the task box with the largest benefit value; otherwise, the redundant target is left in the redundant set and the next round of iteration is entered. When the iterative optimization converges, the historical optimal solution in the optimization is taken.

[0219] For the specific definition of the bin-packing branch hybrid optimization device for multi-spacecraft to multi-star traversal rendezvous, please refer to the definition of the bin-packing branch hybrid optimization method for multi-spacecraft to multi-star traversal rendezvous above, which will not be repeated here. Each module in the above-mentioned bin-packing branch hybrid optimization device for multi-spacecraft to multi-star traversal rendezvous can be fully or partially implemented by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0220] The technical features of the above embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0221] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A box-packing branch hybrid optimization method for multi-spacecraft to multi-satellite traversal rendezvous, characterized in that: The method comprises: The multi-spacecraft multi-satellite traversal rendezvous mission planning problem is decomposed into an outer planning problem and an inner transfer trajectory planning problem, and a mathematical model of the two-layer planning problem is constructed; the outer planning problem is a target allocation, rendezvous sequence and time planning problem; The spacecraft is modeled as a mission box, and the flight mission timeline of a single spacecraft is discretized into uniformly distributed time points. The discrete time points are used as the basic unit for placing targets, and the complex discrete-continuous hybrid optimization problem is transformed into a simple bin packing combinatorial optimization problem. The targets to be rendezvoused are randomly assigned to task boxes, and for each task box, a solution tree of the rendezvous mission plan for a single spacecraft to multiple targets in each task box is constructed, and the time domain representation is integrated into the traditional tree search framework so that the branches in the solution tree can simultaneously represent the optimal rendezvous sequence and time; the solution tree contains multiple high-quality boxing solution branches; The target distribution between each task box is improved by adjusting the global optimization operator; the solution tree of the adjusted task box is reconstructed by deleting the operator and inserting the operator to obtain the intersection time point of each target in each task box; the global optimization operator includes: moving operator, permutation operator, and redistribution operator; The time point of rendezvous of each target in each mission box is taken as the initial value, and the time variable is accurately optimized by the gradient method to obtain the optimal time for each spacecraft to rendezvous with each target in the continuous parameter domain; Based on the optimal time for each spacecraft to rendezvous with each target in the continuous parameter domain, the original problem is transformed into multiple independent inner orbit transfer trajectory planning problems, and the optimal end-to-end transfer trajectory is solved respectively; the original problem is a global optimization problem of multi-spacecraft-to-multi-satellite traversal rendezvous trajectory; Among them, the multi-spacecraft multi-satellite traversal rendezvous mission planning problem is decomposed into an outer planning problem and an inner transfer trajectory planning problem, and a mathematical model of the two-layer planning problem is constructed, including: The inner transfer trajectory planning problem is a nonlinear trajectory optimization problem, and the design variables are the maneuvers of the spacecraft over time: in, T is the maximum thrust amplitude constant, is the spacecraft thrust vector, [0, 1] is the normalized thrust magnitude variable, is the thrust direction unit vector variable, , t is the time variable; In the inner transfer trajectory planning problem, the starting mass of the spacecraft is m d , departure star i d Track and departure time t d , Arrival at the rendezvous star i a The trajectory and arrival time of t a Fixed, the objective function is to minimize fuel consumption: in, is the fuel consumption, is the spacecraft engine specific impulse, is the gravitational acceleration constant of the central body; If both the transfer fuel and the transfer time are considered, the objective function of the inner transfer trajectory planning problem is: in, is the objective function of the inner transfer trajectory planning problem, is the end-to-end evaluation function, Δ t is the transfer duration, and They are the weight coefficients corresponding to the transfer fuel and transfer duration respectively; The constraints of the inner transfer trajectory planning problem are: in, is the upper limit of fuel for a single end-to-end transfer, and They are the lower and upper limits of a single end-to-end transfer duration, and are the terminal position and velocity of the spacecraft in this transfer phase, and are the terminal position and velocity of the target star, and are the maximum allowed terminal position and velocity errors respectively; The outer planning problem is a time-varying vehicle path planning problem, which requires determining all the departure and arrival times of each spacecraft transferring between their respective targets. From the perspective of the target, the design variables of the outer planning problem are expressed as: in, X represents the design variable, i = 1, 2, …, n Indicates the target number, Z i Represents the intersection target i The spacecraft number, t i = Represents spacecraft Z i Intersection Target i moment, and Respectively represent spacecraft Z i The start and end time of the entire sequential rendezvous mission; in order to decouple the discrete sequence combination information and the continuous flight time information in the design space, from the perspective of the spacecraft mission, the design variables are transformed into: in, represents the design variable after deformation, j = 1, 2, …, n z Represents spacecraft z The sequence number of the rendezvous target, n z Represents spacecraft z The total number of targets encountered, Represents spacecraft z The first j The number of the target, Represents spacecraft z No. j The duration of segment transfer; the transformation relationship between the design variables and the design variables after deformation is: Objective function of the outer programming problem J is the sum of the objective function values ​​of all spacecraft missions; the objective function of the outer planning problem is: in, J is the objective function of the outer programming problem, is the objective function of the spacecraft z mission, J Δm,z and J Δt,z are the fuel consumption and mission time of spacecraft z respectively; j = 1, 2, …, n z Represents spacecraft z The sequence number of the rendezvous target, Δ m z,j and Δ t z,j Spacecraft z No. j The fuel consumption and duration of the segment transfer, n z It's a spacecraft z Total number of targets encountered; The constraints of the outer planning problem are: in, J Δm,z,max is the upper limit constraint on the fuel consumption of spacecraft z, J Δt,z,max is the upper limit constraint on the mission time of spacecraft z.

2. The method for hybrid optimization of multi-spacecraft-to-multi-satellite traversal rendezvous based on box packing branches according to claim 1, characterized in that: The targets to be rendezvoused are randomly assigned to task boxes. For each task box, a solution tree of the single-spacecraft-multi-target rendezvous mission plan is constructed. The time domain representation is integrated into the traditional tree search framework so that the branches in the solution tree can simultaneously represent the optimal rendezvous sequence and time, including: At time interval Δ t u Discretize the mission timeline of a single spacecraft into evenly distributed time points; For each mission box, the single spacecraft initial state is initialized as the root node of the solution tree, and the target set to be rendezvoused is initialized; Expand to the next level of nodes in the solution tree and add track transfer for a certain target; Traverse all remaining targets to be intersected; Traverse each target to be intersected x i For all discrete transfer times, calculate the fuel consumption, the fuel consumption of the spacecraft, the mission time of the spacecraft, and the objective function of the spacecraft mission; for the same terminal moment, only keep the optimal time plan for the objective function of the spacecraft mission If the number of time plans is greater than the preset value, the optimal one is retained. time plan; among them, Represents an integer greater than 0; Delete the target to be intersected from the target set to be intersected x i ; If there are too many single-layer nodes, keep the best nodes; If the total flight time or fuel consumption exceeds the upper limit, or the target set to be rendezvoused is empty, the loop traversal is stopped to obtain the optimal rendezvous sequence and time plan solution tree.

3. The method for hybrid optimization of multi-spacecraft-to-multi-satellite traversal rendezvous based on box packing branches according to claim 1, characterized in that: Improve the target distribution between task boxes through global optimization operator adjustment; By deleting operators and inserting operators, the solution tree of the adjusted task box is reconstructed to obtain the intersection time points of each target in each task box, including: In one round of iteration, the target distribution between task boxes is adjusted by moving operators and permuting operators, and the solution tree of the adjusted task box is reconstructed by deleting operators and inserting operators; Collect the targets that cannot be loaded in each task box into the redundant set, and insert the redundant targets into other task boxes through the redistribution operator to save the optimal result; Continue to execute the next round of iteration. If the objective function of the outer planning problem does not change in multiple rounds of iteration, the iteration is terminated to obtain the intersection time point of each target in each task box.

4. The method for hybrid optimization of multi-spacecraft-to-multi-satellite traversal rendezvous by packing branches according to claim 3, characterized in that: The target distribution between task boxes is adjusted by moving operators and permuting operators, and the solution tree of the adjusted task box is reconstructed by deleting operators and inserting operators. The specific steps of moving operators in this step include: Calculate the target loading density of each task box. ε - Greedy strategy selection task box A , that is, ε ∈[0,1], select the task box with the lowest target loading density, with a probability of 1- ε The probability of randomly sampling task boxes from each task box is A ; The task box A is the source task box of the target moved by the mobile operator; the task box z The target loading density is: in, It's a task box z Medium target loading density, is the optimal time solution in a single task box The corresponding total fuel consumption is is the optimal time solution in a single task box The corresponding total flight time; γ Δm ∈[0,1] and γ Δt ∈[0,1] are the weight coefficients corresponding to the transfer fuel and transfer duration respectively, J Δm,z,max is the upper limit constraint on the fuel consumption of spacecraft z, J Δt,z,max is the upper limit constraint on the mission time of spacecraft z; Computational Task Box A The best time plan The objective function of the inner transfer trajectory planning problem for each transfer segment; pass ε - The greedy strategy selects the largest objective function of the inner transfer trajectory planning problem as the target x Otherwise, from the task box A Random sampling target x ; The stated goal x is the target moved by the movement operator; Calculate the task box A The target insertion evaluation values ​​for all other task boxes are: in, Q is the target insertion evaluation value of the task box, and A task box is inserted x The best time plan before and after The objective function value of If there is at least one task box with a positive target insertion evaluation value, ε - Greedy strategy selects the target with the highest evaluation value as the task box B , otherwise, randomly sample from the task box with a positive target insertion evaluation value; if there is no task box with a positive target insertion evaluation value, cancel this move operation; Use the delete operator to remove the task box A Delete a target x And reconstruct the solution tree; Use the insertion operator to insert the target x Insert to Taskbox B And reconstruct the solution tree.

5. The method for hybrid optimization of multi-spacecraft-to-multi-satellite traversal rendezvous based on box packing branches according to claim 4, characterized in that: The target distribution between task boxes is adjusted by moving operators and permuting operators, and the solution tree of the adjusted task box is reconstructed by deleting operators and inserting operators. The specific steps of the permutation operator in this step include: Select Task Box C and goals y , and select the task box in claim 4 A and goals x The method is the same as that of C Each target and task box of each task box other than C Medium Target y The replacement benefit value is: in, is the revenue value, , They are the optimal time plans for task boxes C and D before replacing the target. The objective function value of , They are the optimal time plans for task boxes C and D after replacing the target. The objective function value of If there is at least one set of task boxes D and goals z combination( D , z )of Q CD is positive, through ε - Greedy strategy chooses the replacement with the largest benefit value ( D , z ), otherwise, from is positive ( D , z ) is randomly sampled; if it does not exist is positive ( D , z ), cancel this replacement operation; Use the delete operator to remove the task box C Delete a target y , from the task box D Delete a target z , and reconstruct the solution tree; By inserting the operator y Insert Task Box D The target z Insert Task Box C , and reconstruct the solution tree.

6. The method for hybrid optimization of multi-spacecraft-to-multi-satellite traversal rendezvous by packing branches according to claim 4, characterized in that: Use the delete operator to remove the task box A Delete a target x And reconstruct the solution tree, including: Traverse the task box from the first layer to the last layer A Corresponding solution tree, find the target x Corresponding node, delete the node and the lower branch of the node; if the target does not exist in the solution tree x If the corresponding node is found, the traversal ends.

7. The method for hybrid optimization of multi-spacecraft-to-multi-satellite traversal rendezvous by packing branches according to claim 4, characterized in that: Use the insertion operator to insert the target x Insert to Taskbox B The solution tree is reconstructed, including: Traverse the task box from the first layer to the last layer B The corresponding solution tree traverses all nodes in the current layer and converts the target x Connect to the node and execute the tree solution steps of the single spacecraft multi-target rendezvous mission plan, from the target x Start expanding branches; if a better branch is generated, update the best branch.

8. The method for hybrid optimization of multi-spacecraft-to-multi-satellite traversal rendezvous by packing branches according to claim 3, characterized in that: Collect the targets that cannot be loaded in each task box into the redundant set, and insert the redundant targets into other task boxes through the redistribution operator, including: Collect the targets that cannot be loaded in each task box into a redundant set; Traverse all task boxes and calculate the redundant targets to be inserted into the redundant set x Post-optimal time branch The objective function value is calculated based on the calculated result to insert redundant objectives x The revenue value; If there is at least one task box with a positive benefit value, the insertion operator is used to insert the redundant target into the task box with the largest benefit value; Otherwise, the redundant target is left in the redundant set and the next round of iteration is entered. When the iterative optimization converges, the historical optimal solution in the optimization is taken.

9. A box-packing branch hybrid optimization device for multi-spacecraft multi-satellite traversal rendezvous, characterized in that: The device comprises: A modeling module for two-layer planning problems, used to decompose the multi-spacecraft multi-star traversal rendezvous mission planning problem into an outer planning problem and an inner transfer trajectory planning problem, and to construct a mathematical model of the two-layer planning problem; the outer planning problem is a problem of target allocation, rendezvous sequence and time planning; The problem conversion module is used to model the spacecraft as a task box and discretize the flight mission timeline of a single spacecraft into evenly distributed time points. The discrete time points are used as the basic unit for placing targets, and the complex discrete-continuous hybrid optimization problem is converted into a simple bin packing combinatorial optimization problem. A solution tree construction module for multi-target rendezvous mission solutions is used to construct a solution tree for a single spacecraft to multi-target rendezvous mission solution in each mission box, and integrate the time domain representation into the traditional tree search framework so that the branches in the solution tree can simultaneously represent the optimal rendezvous sequence and time; the solution tree contains multiple high-quality binning solution branches; The target packing distribution optimization and adjustment module adjusts and improves the target distribution between each task box through the global optimization operator; reconstructs the solution tree of the adjusted task box through the deletion operator and the insertion operator to obtain the intersection time point of each target in each task box; the global optimization operator includes: movement operator, permutation operator, and redistribution operator; The optimal rendezvous time determination module is used to take the rendezvous time point of each target in each mission box as the initial value, accurately optimize the time variable through the gradient method, and obtain the optimal time for each spacecraft to rendezvous with each target in the continuous parameter domain; The optimal end-to-end transfer trajectory solving module is used to transform the original problem into multiple independent inner orbit transfer trajectory planning problems based on the optimal time for each spacecraft to rendezvous with each target in the continuous parameter domain, and solve the optimal end-to-end transfer trajectory respectively; the original problem is the global optimization problem of multi-spacecraft-to-multi-satellite traversal rendezvous trajectory; Among them, the modeling module of the two-level planning problem is also used for the inner-level transfer trajectory planning problem, which is a nonlinear trajectory optimization problem, and the design variables are the maneuvers of the spacecraft over time: in, T is the maximum thrust amplitude constant, is the spacecraft thrust vector, [0, 1] is the normalized thrust magnitude variable, is the thrust direction unit vector variable, , t is the time variable; In the inner transfer trajectory planning problem, the starting mass of the spacecraft is m d , departure star i d Track and departure time t d , Arrival at the rendezvous star i a The trajectory and arrival time of t a Fixed, the objective function is to minimize fuel consumption: in, is the fuel consumption, is the spacecraft engine specific impulse, is the gravitational acceleration constant of the central body; If both the transfer fuel and the transfer time are considered, the objective function of the inner transfer trajectory planning problem is: in, is the objective function of the inner transfer trajectory planning problem, is the end-to-end evaluation function, Δ t is the transfer duration, and They are the weight coefficients corresponding to the transfer fuel and transfer duration respectively; The constraints of the inner transfer trajectory planning problem are: in, is the upper limit of fuel for a single end-to-end transfer, and They are the lower and upper limits of a single end-to-end transfer duration, and are the terminal position and velocity of the spacecraft in this transfer phase, and are the terminal position and velocity of the target star, and are the maximum allowed terminal position and velocity errors respectively; The outer planning problem is a time-varying vehicle path planning problem, which requires determining all the departure and arrival times of each spacecraft transferring between their respective targets. From the perspective of the target, the design variables of the outer planning problem are expressed as: in, X represents the design variable, i = 1, 2, …, n Indicates the target number, Z i Represents the intersection target i The spacecraft number, t i = Represents spacecraft Z i Intersection Target i moment, and Respectively represent spacecraft Z i The start and end time of the entire sequential rendezvous mission; in order to decouple the discrete sequence combination information and the continuous flight time information in the design space, from the perspective of the spacecraft mission, the design variables are transformed into: in, represents the design variable after deformation, j = 1, 2, …, n z Represents spacecraft z The sequence number of the rendezvous target, n z Represents spacecraft z The total number of targets encountered, Represents spacecraft z The first j The number of the target, Represents spacecraft z No. j The duration of segment transfer; the transformation relationship between the design variables and the design variables after deformation is: Objective function of the outer programming problem J is the sum of the objective function values ​​of all spacecraft missions; the objective function of the outer planning problem is: in, J is the objective function of the outer programming problem, is the objective function of the spacecraft z mission, J Δm,z and J Δt,z are the fuel consumption and mission time of spacecraft z respectively; j = 1, 2, …, n z Represents spacecraft z The sequence number of the rendezvous target, Δ m z,j and Δ t z,j Spacecraft z No. j The fuel consumption and duration of the segment transfer, n z It's a spacecraft z Total number of targets encountered; The constraints of the outer planning problem are: in, J Δm,z,max is the upper limit constraint on the fuel consumption of spacecraft z, J Δt,z,max is the upper limit constraint on the mission time of spacecraft z.