A spacecraft on-orbit service large-scale task scheduling method and related equipment

Abstract

The application provides a spacecraft on-orbit service large-scale task scheduling method and related equipment, comprising: establishing an orbit dynamics equation of a single service spacecraft and a single target satellite to determine the out-of-plane transfer phase and the coplanar transfer phase of the service spacecraft to the target satellite; taking the time and fuel consumption of the service spacecraft as performance indicators to establish a cost function; sampling multiple phase angles of the target satellite and performing nonlinear mapping on the orbit characteristics extracted based on the cost function to obtain a comprehensive cost, a time cost and an optimal transfer orbit; based on the Monte Carlo tree search method, the service spacecraft is assigned to each target satellite, and the optimal transfer orbit is selected through the established large-scale task scheduling model to schedule each service spacecraft to the corresponding target satellite; the exploration of the task state space in the process of the spacecraft on-orbit service large-scale task scheduling is avoided to fall into a local optimum, and the speed and accuracy of the large-scale on-orbit service task scheduling are improved.

Description

Technical Field

[0001] This invention relates to the field of satellite on-orbit servicing mission scheduling technology, and in particular to a method and related equipment for large-scale on-orbit servicing mission scheduling of spacecraft. Background Technology

[0002] On-orbit servicing technology refers to the technology of spacecraft approaching space targets via orbital transfer to perform payload replacement, upgrades, and other operations. This technology can effectively enhance the capabilities of space systems and extend the operational lifespan of satellites. With the rapid development of the space industry, the number of space satellites has multiplied, and the scenarios for spacecraft on-orbit servicing have become increasingly complex. The focus has shifted from a single spacecraft serving a single target satellite to multiple spacecraft serving multiple target satellites. Large-scale on-orbit servicing mission scheduling has become a pressing issue. Currently, most on-orbit servicing mission scheduling methods suffer from problems such as excessive computation time and large search spaces. Changes in target or environmental conditions also limit the effectiveness of mission allocation, exacerbate the risk of space orbital conflicts, and fail to meet the requirements of rapid decision-making and real-time scheduling.

[0003] For the spacecraft on-orbit servicing mission scheduling problem, existing technologies mostly employ integer programming, dynamic programming, heuristic algorithms, and machine learning-based methods for solution. While these methods can achieve mission scheduling and orbit transfer, integer programming methods suffer from typical NP-hard problems when dealing with large-scale mission scheduling. NP-hard problems (NP, Nondeterminism Polynomial) are problems for which no polynomial-time algorithm solution has been found, requiring a very long time to obtain a feasible or optimal solution, which is not conducive to rapid decision-making by on-orbit spacecraft. Heuristic algorithms such as genetic algorithms suffer from problems such as excessively large search spaces and high dependence on fitness functions when dealing with large-scale problems, making them prone to getting trapped in local optima. Machine learning-based methods require a large amount of training data, suffer from low algorithm convergence, and due to the payload constraints of spacecraft, it is difficult to provide sufficient computational resources. Summary of the Invention

[0004] This invention provides a method and related equipment for large-scale on-orbit servicing mission scheduling of spacecraft. Its purpose is to avoid the exploration of the mission state space getting stuck in local optima during the large-scale on-orbit servicing mission scheduling process, thereby improving the speed and accuracy of large-scale on-orbit servicing mission scheduling.

[0005] To achieve the above objectives, this invention provides a method for large-scale on-orbit servicing mission scheduling of spacecraft, including multiple servicing spacecraft and multiple target satellites. The mission scheduling method includes:

[0006] Step 1: Establish the orbital dynamics equations of a single service spacecraft and a single target satellite, and determine the non-planar transfer phase and the coplanar transfer phase of the service spacecraft to the target satellite based on the orbital dynamics equations of the service spacecraft and the target satellite.

[0007] Step 2: Using the time and fuel consumed by the servicing spacecraft during the non-plane transfer phase as performance indicators, establish the first cost function for the non-plane transfer phase. Using the time and fuel consumed by the servicing spacecraft during the coplane transfer phase as performance indicators, establish the second cost function for the coplane transfer phase. The first and second cost functions are used to evaluate the quality of the transfer orbit.

[0008] Step 3: Select the optimal surface transfer orbit for transferring the service spacecraft to the target satellite using the first cost function;

[0009] Step 4: Sample the phase angle difference of the target satellite at equal intervals to obtain multiple phase angles, and extract the second cost function and transfer time corresponding to each phase angle as orbital features;

[0010] Step 5: Nonlinearly map multiple phase angles to orbital characteristics to obtain the comprehensive cost and time cost of the optimal orbit in the coplanar transfer phase. This cost is used to generate the optimal coplanar transfer orbit. Based on the optimal coplanar transfer orbit and the optimal coplanar transfer orbit, the optimal transfer orbit for each service spacecraft to transfer to the corresponding target satellite is obtained.

[0011] Step 6: Assign a service spacecraft to each target satellite based on the Monte Carlo tree search method, and select the optimal transfer orbit from the optimal transfer orbit of each service spacecraft to the corresponding target satellite through the established large-scale mission scheduling model. Schedule each service spacecraft to the corresponding target satellite through the selected optimal transfer orbit.

[0012] Furthermore, step 1 includes:

[0013] The orbital dynamics equations for a single servicing spacecraft are as follows:

[0014]

[0015] Where, r Si and v Si Let u and μ represent the position vector and velocity vector of the i-th service spacecraft in the geocentric coordinate system, respectively, where μ represents the Earth's gravitational constant. Si ={U 1,Sij (t 1,ij2 ),U 2,Sij (t 2,ij1 ),U 2,Sij (t 2,ij2 )},u SiU represents the pulse control vector. 1,Sij (t 1,ij2 U represents the pulse control vector required for the non-plane transfer stage. 2,Sij (t 2,ij1 ),U 2,Sij (t 2,ij2 ) represents the pulse control vector required for the coplanar transfer phase, and m represents the number of servicing spacecraft;

[0016] The orbital dynamics equations for a single target satellite are:

[0017]

[0018] Where, r Tj and v Tj Let represent the position vector and velocity vector of the j-th target satellite in the geocentric coordinate system, respectively, and n represent the number of target satellites;

[0019] Based on the orbital dynamics equations of the servicing spacecraft and the target satellite, the orbital intersection line between the servicing spacecraft and the target satellite is determined;

[0020] The process of the service spacecraft reaching the orbital intersection line is defined as the non-plane transfer phase of the service spacecraft's transfer to the target satellite;

[0021] The process of the service spacecraft approaching the target satellite from the orbital intersection line is defined as the coplanar transfer phase of the service spacecraft's transfer to the target satellite.

[0022] Furthermore, step 2 includes:

[0023] Using the time and fuel consumed by the servicing spacecraft during the off-plane transfer phase as performance indicators, a first cost function J for the off-plane transfer phase is established. 1,ij for:

[0024] J 1,ij =CJ 1,vij +(1-C)J 1,tij

[0025] Among them, J 1,vij J represents the normalized variable representing fuel consumption during the non-planar transfer phase. 1,tij The normalized variable representing the time consumption of the non-plane transfer stage, where C represents the scaling factor;

[0026] Using the time and fuel consumed by the servicing spacecraft during the coplanar transfer phase as performance indicators, a second cost function J for the coplanar transfer phase is established. 2,ij for:

[0027] J 2,ij =CJ 2,tij +(1-C)J 2,vij

[0028] Among them, J 2,vij Table J represents the normalized variable representing fuel consumption during the coplanar transfer phase. 2,tij Normalized variables representing the time consumption of the coplanar transition phase;

[0029] Both the first and second cost functions are used to evaluate the quality of the transfer orbit.

[0030] Furthermore, by performing a nonlinear mapping between multiple phase angles and orbital characteristics, the comprehensive cost and time cost of the optimal orbit during the coplanar transfer stage are obtained, including:

[0031] Multiple phase angles and orbital features are input into a multilayer perceptron. The multilayer perceptron performs nonlinear mapping processing on the multiple phase angles and orbital features to obtain the comprehensive cost and time cost of the optimal orbit in the coplanar transfer stage, which is used to generate the optimal coplanar transfer orbit.

[0032] A multilayer perceptron consists of an input layer, a first fully connected layer, a second fully connected layer, and an output layer connected in sequence.

[0033] Furthermore, by using a multilayer perceptron to perform nonlinear mapping between multiple phase angles and orbital features, the comprehensive cost and time cost of the optimal orbit during the coplanar transfer stage are obtained, including:

[0034] Multiple phase angles and orbital features are transmitted from the input layer to the first fully connected layer via forward propagation. The expression h1 for forward propagation is:

[0035]

[0036] The first output matrix E1 is obtained by performing a full connection through the first fully connected layer:

[0037]

[0038] By performing a full connection through a second fully connected layer, the second output matrix E2 is obtained as follows:

[0039]

[0040] The output vector through the output layer is:

[0041]

[0042] Where h1 represents the weighted sum of the data, w b The weight representing the phase angle, Δθ a Let represent the a-th phase angle, ε1, ε2, and ε3 represent offsets, N represent the number of input data, M represent the number of neurons b in the first and second fully connected layers, g(·) represent the activation function, and D represent the number of output data. and This indicates the combined cost and the time cost.

[0043] Furthermore, based on the Monte Carlo tree search method, service spacecraft are assigned to each target satellite, including:

[0044] Treating each service spacecraft as an independent object, the allocation of target satellites is viewed as a group game process;

[0045] The pairing of a service spacecraft with a target satellite is defined as a state. The set of all states at any given time constitutes the nodes of a Monte Carlo tree. The nodes include the root node, which represents that no service spacecraft is paired with the target satellite, the child nodes, which represent that some service spacecraft are paired with the target satellite, and the leaf nodes, which represent that all service spacecraft are paired with the target satellite.

[0046] Start the search from the root node, using the formula The upper confidence interval values ​​for the child nodes are calculated, where, S represents the average reward value of child node k, and s represents the total number of explorations. k C represents the number of times child node k has been explored. p Indicates the exploration coefficient;

[0047] Generate new child nodes for child nodes with a non-zero exploration count;

[0048] Starting from a new child node, the game proceeds through random simulation of state transitions until the game is completed, and the reward value is evaluated.

[0049] The information of each node on the Monte Carlo tree is updated through backpropagation, including reward value and number of explorations.

[0050] Furthermore, the expression for the large-scale task scheduling model is:

[0051]

[0052] Where, α ij Represents the allocation coefficient, when α ij =1 When service spacecraft i is assigned to target satellite j, when α ij =0 When the service spacecraft i is not assigned to the target satellite j, J(α) ij J*(α) represents the decision performance function. ij ) represents the optimal decision performance function;

[0053] Distribution coefficient α ij The following conditions must be met:

[0054]

[0055] in, This means that each target satellite must be assigned a service spacecraft. This means that each service spacecraft can dock with a maximum of one target satellite.

[0056] Furthermore, each service spacecraft is scheduled to its corresponding target satellite via a selected optimal transfer orbit, including:

[0057] The state space is established with the target satellite as the state:

[0058]

[0059] The action space is established by selecting the target satellite and the service spacecraft as the action set:

[0060]

[0061] The comprehensive cost of the optimal transfer trajectory is used as the immediate reward value of the action set, and the reward function is designed as follows:

[0062]

[0063] The action sequence is obtained as the scheduling result by solving the following equation:

[0064]

[0065] in, Let the j-th target satellite be represented as the state. Let R represent the i-th service spacecraft as the action set, and let R(·) represent the reward function;

[0066] For each State selection action Receive timely reward points If the task scheduling result R does not satisfy the allocation coefficient α ij If the conditions are met, a penalty of -1 will be imposed.

[0067] The present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements a method for large-scale mission scheduling of spacecraft on-orbit servicing.

[0068] The present invention also provides a terminal device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements a method for large-scale mission scheduling of spacecraft on-orbit servicing.

[0069] The above-described solution of the present invention has the following beneficial effects:

[0070] This invention determines the non-planar transfer phase and the coplanar transfer phase of a service spacecraft to a target satellite by establishing the orbital dynamics equations of a single service spacecraft and a single target satellite. It uses the time and fuel consumed by the service spacecraft in the non-planar and coplanar transfer phases as performance indicators to establish a first cost function and a second cost function. The optimal coplanar transfer orbit is selected using the first cost function. The phase angle difference of the target satellite is sampled at equal intervals, and the resulting multiple phase angles are nonlinearly mapped with the extracted orbital features to obtain the comprehensive cost and time cost for generating the optimal orbit in the coplanar transfer phase, thus obtaining the optimal transfer orbit for each service spacecraft to its corresponding target satellite. A service spacecraft is assigned to each target satellite based on the Monte Carlo tree search method, and the optimal transfer orbit is selected from the optimal transfer orbits for each service spacecraft to its corresponding target satellite using the established large-scale mission scheduling model. Each service spacecraft is then scheduled to its corresponding target satellite using the selected optimal transfer orbit. Compared with existing technologies, this invention avoids the exploration of the mission state space getting trapped in local optima during the large-scale mission scheduling of on-orbit servicing, thus improving the speed and accuracy of large-scale on-orbit servicing mission scheduling.

[0071] Other beneficial effects of the present invention will be described in detail in the following detailed description section. Attached Figure Description

[0072] Figure 1 This is a flowchart illustrating an embodiment of the present invention;

[0073] Figure 2 This is a schematic diagram of the structure of the multilayer sensor in an embodiment of the present invention;

[0074] Figure 3 This is a schematic diagram of a single iteration process of the Monte Carlo search tree in an embodiment of the present invention;

[0075] Figure 4 This is a schematic diagram of the node depth of the Monte Carlo tree in an embodiment of the present invention. Detailed Implementation

[0076] To make the technical problems, solutions, and advantages of this invention clearer, a detailed description will be provided below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0077] In the description of this invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0078] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a locking connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0079] Furthermore, the technical features involved in the different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0080] This invention addresses existing problems by providing a method and related equipment for large-scale on-orbit servicing of spacecraft.

[0081] like Figure 1 As shown, embodiments of the present invention provide a method for large-scale on-orbit servicing mission scheduling of spacecraft, including multiple servicing spacecraft and multiple target satellites. The mission scheduling method includes:

[0082] Step 1: Establish the orbital dynamics equations of a single service spacecraft and a single target satellite, and determine the non-planar transfer phase and the coplanar transfer phase of the service spacecraft to the target satellite based on the orbital dynamics equations of the service spacecraft and the target satellite.

[0083] Step 2: Using the time and fuel consumed by the servicing spacecraft during the non-plane transfer phase as performance indicators, establish the first cost function for the non-plane transfer phase. Using the time and fuel consumed by the servicing spacecraft during the coplane transfer phase as performance indicators, establish the second cost function for the coplane transfer phase. The first and second cost functions are used to evaluate the quality of the transfer orbit.

[0084] Step 3: Select the optimal surface transfer orbit for transferring the service spacecraft to the target satellite using the first cost function;

[0085] Step 4: Sample the phase angle difference of the target satellite at equal intervals to obtain multiple phase angles, and extract the second cost function and transfer time corresponding to each phase angle as orbital features;

[0086] Step 5: Nonlinearly map multiple phase angles to orbital characteristics to obtain the comprehensive cost and time cost of the optimal orbit in the coplanar transfer phase. This cost is used to generate the optimal coplanar transfer orbit. Based on the optimal coplanar transfer orbit and the optimal coplanar transfer orbit, the optimal transfer orbit for each service spacecraft to transfer to the corresponding target satellite is obtained.

[0087] Step 6: Assign a service spacecraft to each target satellite based on the Monte Carlo tree search method, and select the optimal transfer orbit from the optimal transfer orbit of each service spacecraft to the corresponding target satellite through the established large-scale mission scheduling model. Schedule each service spacecraft to the corresponding target satellite through the selected optimal transfer orbit.

[0088] Specifically, step 1 includes:

[0089] For m service spacecraft, the orbital dynamics equations for a single service spacecraft are established as follows:

[0090]

[0091] Where, r Si and v Si Let u and μ represent the position vector and velocity vector of the i-th service spacecraft in the geocentric coordinate system, respectively, where μ represents the Earth's gravitational constant. Si This represents the pulse control vector. Since both the non-plane and coplanar transfer phases of the servicing spacecraft require pulse control vectors, u... Si ={U 1,Sij (t 1,ij2 ),U 2,Sij (t 2,ij1 ),U 2,Sij (t 2,ij2 )}, U 1,Sij (t 1,ij2 U represents the pulse control vector required for the non-plane transfer stage, i.e., the pulse control vector required to complete one non-plane transfer. 2,Sij (t 2,ij1 ),U 2,Sij (t 2,ij2 ) represents the pulse control vector required for the coplanar transfer phase, i.e., the pulse control vector required to complete one non-coplanar transfer, and m represents the number of servicing spacecraft;

[0092] Assuming there are n target satellites, the orbital dynamics equation for a single target satellite is:

[0093]

[0094] Where, r Tj and v Tj Let represent the position vector and velocity vector of the j-th target satellite in the geocentric coordinate system, respectively, and n represent the number of target satellites;

[0095] The position and velocity of a single target satellite can be determined from its orbital dynamics equations. The target satellite floats unpowered in its orbit and is not subject to artificial pulse control; therefore, when r... Si =r Tj v Si =v Tj At that time, it was considered that the service spacecraft and the target satellite had rendezvoused;

[0096] Based on the orbital dynamics equations of the servicing spacecraft and the target satellite, the orbital intersection line between the servicing spacecraft and the target satellite is determined, which can be divided into two stages:

[0097] The process of the service spacecraft reaching the orbital intersection line is defined as the first stage, which is the non-plane transfer stage of the service spacecraft's transfer to the target satellite.

[0098] The process of the service spacecraft approaching the target satellite from the orbital intersection line is defined as the second stage, which is the coplanar transfer stage of the service spacecraft's transfer to the target satellite.

[0099] Specifically, step 2 includes:

[0100] Using the time and fuel consumed by the servicing spacecraft during the off-plane transfer phase as performance indicators, a first cost function J for the off-plane transfer phase is established. 1,ij for:

[0101] J 1,ij =CJ 1,vij +(1-C)J 1,tij

[0102] Among them, J 1,vij J represents the normalized variable representing fuel consumption during the non-planar transfer phase. 1,tij The normalized variable representing the time consumption of the non-plane transfer stage, where C represents the scaling factor;

[0103] Using the time and fuel consumed by the servicing spacecraft during the coplanar transfer phase as performance indicators, a second cost function J for the coplanar transfer phase is established. 2,ij for:

[0104] J 2,ij =CJ 2,tij +(1-C)J 2,vij

[0105] Among them, J 2,vijTable J represents the normalized variable representing fuel consumption during the coplanar transfer phase. 2,tij Normalized variables representing the time consumption of the coplanar transition phase;

[0106] Both the first and second cost functions are used to evaluate the quality of the transfer orbit. The less time and fuel consumed, the higher the quality of the orbit is considered.

[0107] Among them, the first cost function can comprehensively evaluate the time consumption and fuel consumption required for the transfer of the non-plane orbit, so as to select the best plane transfer orbit that consumes less fuel and less time.

[0108] The second cost function can comprehensively evaluate the time and fuel consumption required for coplanar orbit transfer, thereby selecting the optimal coplanar transfer orbit that consumes less fuel and less time.

[0109] In the non-plane transfer phase of this invention embodiment, it is assumed that the serving spacecraft is located in orbit 1 and the target satellite is located in orbit 2. The two orbits are not on the same plane, and their planes intersect. When the serving spacecraft reaches the intersection of the orbital planes, it applies a minimal non-plane maneuver to enter the target satellite's orbital plane while maintaining its orbital radius. Therefore, the time consumption Δt of the non-plane transfer phase is... 1,ij It can be represented as:

[0110] Δt 1,ij =(t 1,ij2 -t 1,ij1 )

[0111] Among them, t 1,ij1 t represents the start time of the non-plane transition phase. 1,ij2 Indicates the end time of the non-plane transition phase;

[0112] Solve for the normalized variable J of fuel consumption during the non-planar transfer stage. 1,vij The normalized variable J of time consumption 1,tij The expression is:

[0113] J 1,vij =(||U 1,Sij (t 1,ij2 )||-u min ) / (u max -(u min )

[0114] J 1,tij =(Δt) 1,ij -t0) / (t f -t0)

[0115] Among them, u min u represents the minimum value of the control vector pulse.max The value of the control vector pulse is represented by t0, and the minimum orbital transfer time is represented by t. f Indicates the maximum orbital transfer time;

[0116] After the off-plane transfer phase, the service spacecraft and the target satellite are on the same plane.

[0117] In this embodiment of the invention, the coplanar transfer stage adopts the Lambert transfer method. Assuming r1 and r2 are the position vectors of the two target satellites, v1 and v2 represent the velocity vectors of the two target satellites, and Δθ represents the difference in the coplanar phase angles of the two target satellites, the transfer process is as follows:

[0118] Through formula Calculate r1 and r2;

[0119] Calculate the difference Δθ between the coplanar phase angles of the two target satellites and determine whether the transfer orbit is retrograde or prograde.

[0120] Through formula Calculate the constant A;

[0121] Use Newton's method to determine whether it is an elliptical orbit type;

[0122] From the formula Calculate y, where z = aχ 2 ;

[0123] Calculate the Lagrange function f, g and

[0124] Calculate v1 and v2;

[0125] The orbital elements are calculated using r1 and r2.

[0126] Therefore, the time consumption Δt of the coplanar transition phase 2,ij It can be represented as:

[0127] Δt 2,ij =(t 2,ij2 -t 2,ij1 )

[0128] Among them, t 2,ij1 This represents the start time of the coplanar transition phase, which satisfies t 2,ij1 =t 1,ij2 , t 2,ij2 This indicates the end of the coplanar transfer phase, signifying that the service spacecraft and the target satellite have rendezvoused.

[0129] Solving for the normalized variable J of fuel consumption during the coplanar transfer phase 2,vij The normalized variable J of time consumption2,tij The expression is:

[0130] J 2,vij =(||U 2,Sij (t 2,ij1 )||+||U 2,Sij (t 2,ij2 )||-u min ) / (u max -u min

[0131] J 2,tij =(Δt) 2,ij -t0) / (t f -t0)

[0132] Among them, u min u represents the minimum value of the control vector pulse. max The value of the control vector pulse is represented by t0, and the minimum orbital transfer time is represented by t. f This indicates the maximum orbital transfer time.

[0133] Combining the non-planar transfer phase and the coplanar transfer phase, the cost function of a single-service spacecraft for a single-target satellite can be obtained as follows:

[0134] J tij =J 1,tij +J 2,tij

[0135] J ij =J 1,ij +J 2,ij

[0136] Among them, J tij J represents the time it takes for a service spacecraft to reach its target satellite. ij This indicates the overall cost.

[0137] When the cost function of the coplanar transfer phase reaches its minimum value, the cost function of the entire trajectory transfer process also reaches its minimum value, meaning the entire transfer process represents the optimal trajectory. Therefore, the time consumption cost of the optimal transfer trajectory is... and overall consumption cost It can be represented as:

[0138]

[0139]

[0140] in, This represents the time cost of transferring along the optimal coplanar transfer trajectory. This represents the overall cost of transferring along the optimal coplanar transfer trajectory.

[0141] Specifically, multiple phase angles are nonlinearly mapped to orbital characteristics to obtain the comprehensive cost and time cost of the optimal orbit during the coplanar transfer stage, including:

[0142] Multiple phase angles and orbital features are input into a multilayer perceptron. The multilayer perceptron performs nonlinear mapping processing on the multiple phase angles and orbital features to obtain the comprehensive cost and time cost of the optimal orbit in the coplanar transfer stage, which is used to generate the optimal coplanar transfer orbit.

[0143] like Figure 2 As shown, the multilayer perceptron includes an input layer, a first fully connected layer, a second fully connected layer, and an output layer connected in sequence.

[0144] Solving for coplanar orbit transfers to obtain a large number of optimal coplanar transfer orbits under different initial conditions requires extensive iterative calculations, which severely impacts the decision-making speed of large-scale mission scheduling. To avoid iterative calculations, a large amount of sample data of coplanar transfer orbits can be stored offline within the servicing spacecraft. However, this method consumes a significant amount of storage resources, and the second cost function J of the coplanar transfer phase... 2,ij and transition time Δt 2,ij Only the phase angle difference Δθ with the target satellite ij related.

[0145] The embodiments of the present invention address Δθ ij Equal-interval sampling is performed in the range [0, 360°], and the second cost function J is extracted from a large number of sample data of coplanar transfer trajectories. 2,ij and transition time Δt 2,ij As orbital features, a multi-layer perceptron (MLP) is used to represent Δθ = Δθ. ij Using / 360 as input, a nonlinear mapping is performed with the orbital features, and the corresponding output is obtained through offline optimization. and This allows for the rapid planning of the optimal coplanar transfer trajectory.

[0146] Specifically, a multilayer perceptron is used to perform nonlinear mapping between multiple phase angles and orbital features to obtain the comprehensive cost and time cost of the optimal orbit during the coplanar transfer stage, including:

[0147] Multiple phase angles and orbital features are transmitted from the input layer to the first fully connected layer via forward propagation. The expression h1 for forward propagation is:

[0148]

[0149] A full connection is achieved by using the first fully connected layer, i.e., activating h1 with an activation function, resulting in the first output matrix E1:

[0150]

[0151] The sigmoid function is used to activate h1, and the expression is:

[0152]

[0153] Its derivative is:

[0154] σ'(h1)=σ(h1)[1-σ(h1)]

[0155] Substituting e1 = g(h1) into the equation, we get:

[0156] g'(h1)=e1(1-e1)

[0157] By performing a full connection through a second fully connected layer, the second output matrix E2 is obtained as follows:

[0158]

[0159] The output vector through the output layer is:

[0160]

[0161] Where h1 represents the weighted sum of the data, w b The weight representing the phase angle, Δθ a Let represent the a-th phase angle, ε1, ε2, and ε3 represent offsets, N represent the number of input data, M represent the number of neurons b in the first and second fully connected layers, g(·) represent the activation function, and D represent the number of output data. and This indicates the combined cost and the time cost.

[0162] In this embodiment of the invention, the training process of the multilayer perceptron is as follows:

[0163] The loss function is established using the sum of squared errors. for:

[0164]

[0165] Among them, E t Indicates the baseline numerical result;

[0166] For loss function Find the partial derivative with respect to the weight w:

[0167]

[0168] The partial derivative of the right-hand side of the equation with respect to h can be expressed as:

[0169]

[0170] The increment term of the output layer is obtained as follows:

[0171]

[0172] Update the weights of the output layer:

[0173]

[0174] In the formula, η represents the learning rate, which is used to adjust the speed of loss convergence;

[0175] The incremental terms δ1 of the first fully connected layer and δ2 of the second fully connected layer are obtained as follows:

[0176]

[0177] Update the weights of the first and second fully connected layers:

[0178]

[0179] Where I3 represents a unit vector and I2 represents an identity matrix.

[0180] Specifically, based on the Monte Carlo tree search method, service spacecraft are assigned to each target satellite, including:

[0181] Treating each service spacecraft as an independent object, the allocation of target satellites is viewed as a group game process;

[0182] The pairing of a service spacecraft with a target satellite is defined as a state. The set of all states at any given time constitutes the nodes of a Monte Carlo tree. The nodes include the root node, which represents that no service spacecraft is paired with the target satellite, the child nodes, which represent that some service spacecraft are paired with the target satellite, and the leaf nodes, which represent that all service spacecraft are paired with the target satellite.

[0183] Start the search from the root node, using the formula The upper confidence interval values ​​for the child nodes are calculated, where, S represents the average reward value of child node k, and s represents the total number of explorations. k C represents the number of times child node k has been explored. p Indicates the exploration coefficient;

[0184] Generate new child nodes for child nodes with a non-zero exploration count;

[0185] Starting from a new child node, the game proceeds through random simulation of state transitions until the game is completed, and the reward value is evaluated.

[0186] The information of each node in the Monte Carlo tree is updated through backpropagation, including the average reward value. And number of explorations S k Its expression is:

[0187] S k =S k +1

[0188]

[0189] Where, ∑Q k This represents the sum of the reward values ​​for the current node k.

[0190] Specifically, the expression for the large-scale task scheduling model is:

[0191]

[0192] Where, α ij Represents the allocation coefficient, when α ij =1 When service spacecraft i is assigned to target satellite j, when α ij =0 When the service spacecraft i is not assigned to the target satellite j, J(α) ij J*(α) represents the decision performance function. ij ) represents the optimal decision performance function;

[0193] Distribution coefficient α ij The following conditions must be met:

[0194]

[0195] in, This means that each target satellite must be assigned a service spacecraft. This means that each service spacecraft can dock with a maximum of one target satellite.

[0196] Specifically, each service spacecraft is scheduled to its corresponding target satellite via a selected optimal transfer orbit, including:

[0197] In large-scale task allocation, the number of service spacecraft exceeds the number of target satellites, i.e., m > n. The allocation process can be interpreted as "target satellites selecting service spacecraft." Therefore, the state space is established with the target satellites as the state:

[0198]

[0199] The action space is established by selecting the target satellite and the service spacecraft as the action set:

[0200]

[0201] The comprehensive cost of the optimal transfer trajectory is used as the immediate reward value of the action set, and the reward function is designed as follows:

[0202]

[0203] The action sequence is obtained as the scheduling result by solving the following equation:

[0204]

[0205] in, Let the j-th target satellite be represented as the state. Let R represent the i-th service spacecraft as the action set, and let R(·) represent the reward function;

[0206] For each State selection action Receive timely reward points If the task scheduling result R does not satisfy the allocation coefficient α ij If the conditions are met, a penalty of -1 will be imposed.

[0207] The simulation parameters for the large-scale on-orbit service task scheduling problem in this embodiment of the invention are set as follows:

[0208] Assuming the target satellite is stably operating in its current orbit and no orbital transfers or other orbital maneuvers are performed, the orbital six-element settings for the service spacecraft and the target satellite are as follows:

[0209] The target satellite's orbital altitude is L S =20200km, orbital inclination α S =55°, the orbital altitude of the service spacecraft is L T =2000km, orbital inclination α T = 89.421°. Considering near-circular transfer, the eccentricity between the servicing spacecraft and the target satellite is e. S =e T =0, the argument of perigee is ω S =ω S =0°, Right Ascension Ω of the ascending node of the service spacecraft and the target satellite S ,Ω T And true near point angle All are randomly generated.

[0210] The experiment was conducted with a service spacecraft count of m = 100 and a target satellite count of n = 60; variable parameters were set in the MCTS algorithm. The search tree width is limited to L=5 selectable nodes, and the total number of iterations is 30000. The simulation results are shown in Table 1 below, and the node information is shown in Table 2 below.

[0211] Table 1

[0212] Target satellite 1 2 3 4 5 6 7 8 9 10 11 12 Service spacecraft 38 55 62 16 98 35 34 13 78 85 26 45 Target satellite 13 14 15 16 17 18 19 20 21 22 23 24 Service spacecraft 97 70 65 89 36 30 68 39 66 17 67 58 Target satellite 25 26 27 28 29 30 31 32 33 34 35 36 Service spacecraft 8 2 4 43 87 56 23 79 41 84 90 28 Target satellite 37 38 39 40 41 42 43 44 45 46 47 48 Service spacecraft 51 10 52 71 93 69 86 6 80 48 50 5 Target satellite 49 50 51 52 53 54 55 56 57 58 59 60 Service spacecraft 75 82 73 47 60 42 100 25 27 64 15 76

[0213] Table 2

[0214]

[0215] As shown in the table above, there are 60 action sequences for "pairing service spacecraft i with target satellite j", where i ∈ [1, 100] and j ∈ [1, 60]. In this pairing scheme, the optimal combined cost of time and fuel is... The value is 6.223; the advantages of the embodiments of the present invention are as follows:

[0216] 1. Based on MLP, the optimal transfer orbit features are stored, and the optimal transfer orbit is quickly solved for service spacecraft and target satellites in different positions and states;

[0217] 2. For large-scale task scheduling models, the MCTS algorithm has a fast solution speed and can realize rapid decision-making for spacecraft clusters; the generated allocation scheme is of high quality and achieves the optimal solution in terms of time and fuel costs.

[0218] In summary, this invention determines the non-planar transfer phase and the coplanar transfer phase of the service spacecraft to the target satellite by establishing the orbital dynamics equations of a single service spacecraft and a single target satellite; it establishes a first cost function and a second cost function by using the time and fuel consumed by the service spacecraft in the non-planar and coplanar transfer phases as performance indicators; it selects the optimal coplanar transfer orbit through the first cost function; it samples the phase angle difference of the target satellite at equal intervals, and performs nonlinear mapping between the obtained multiple phase angles and the extracted orbital features to obtain the comprehensive cost and time cost for generating the optimal orbit in the coplanar transfer phase, and obtains the optimal transfer orbit for each service spacecraft to the corresponding target satellite; it assigns a service spacecraft to each target satellite based on the Monte Carlo tree search method, and selects the optimal transfer orbit from the optimal transfer orbits of each service spacecraft to the corresponding target satellite through the established large-scale mission scheduling model, and schedules each service spacecraft to the corresponding target satellite through the selected optimal transfer orbit; compared with the prior art, this invention avoids the exploration of the mission state space getting trapped in local optima during the large-scale mission scheduling of spacecraft on-orbit servicing, and improves the speed and accuracy of large-scale on-orbit servicing mission scheduling.

[0219] This invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements a method for large-scale on-orbit servicing scheduling of spacecraft.

[0220] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the embodiments of the present invention can implement all or part of the processes in the methods described above by instructing related hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include at least: any entity or device capable of carrying the computer program code to a building device / terminal device, a recording medium, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium. Examples include USB flash drives, portable hard drives, magnetic disks, or optical disks. In some jurisdictions, according to legislation and patent practice, computer-readable media cannot be electrical carrier signals or telecommunication signals.

[0221] This invention also provides a terminal device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements a method for large-scale on-orbit servicing scheduling of spacecraft.

[0222] It should be noted that the terminal device can be a mobile phone, tablet computer, laptop computer, Ultra-mobile Personal Computer (UMPC), netbook, Personal Digital Assistant (PDA), etc. For example, the terminal device can be a station (ST) in a WLAN, a cellular phone, cordless phone, Session Initiation Protocol (SIP) phone, Wireless Local Loop (WLL) station, PDA, handheld device with wireless communication capabilities, computing device or other processing device connected to a wireless modem, computer, laptop computer, handheld communication device, handheld computing device, satellite wireless device, etc. This embodiment of the invention does not impose any restrictions on the specific type of terminal device.

[0223] The processor referred to can be a Central Processing Unit (CPU), but it can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor.

[0224] In some embodiments, the memory may be an internal storage unit of the terminal device, such as a hard drive or RAM. In other embodiments, the memory may be an external storage device of the terminal device, such as a plug-in hard drive, smart media card (SMC), secure digital card (SD), flash card, etc. Furthermore, the memory may include both internal and external storage units of the terminal device. The memory is used to store the operating system, applications, boot loader, data, and other programs, such as the program code of the computer program. The memory can also be used to temporarily store data that has been output or will be output.

[0225] It should be noted that the information interaction and execution process between the above-mentioned devices / units are based on the same concept as the method embodiments of the present invention. For details on their specific functions and technical effects, please refer to the method embodiments section, which will not be repeated here.

[0226] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for large-scale mission scheduling of spacecraft on-orbit servicing, characterized in that, The mission scheduling method includes multiple service spacecraft and multiple target satellites, and comprises: Step 1: Establish the orbital dynamics equations for a single servicing spacecraft and a single target satellite, and based on these equations, determine the non-planar transfer phase and the coplanar transfer phase of the servicing spacecraft's transfer to the target satellite, including: The orbital dynamics equations for a single servicing spacecraft are established as follows: in, and They represent the first The position vector and velocity vector of a service spacecraft in the geocentric coordinate system Represents the Earth's gravitational constant. , Represents the pulse control vector. This represents the pulse control vector required for the non-plane transfer stage. This represents the pulse control vector required for the coplanar transfer phase. Indicates the number of serviced spacecraft; The orbital dynamics equations for a single target satellite are: in, and They represent the first The position vector and velocity vector of the target satellite in the geocentric coordinate system Indicates the number of target satellites; Based on the orbital dynamics equations of the servicing spacecraft and the target satellite, the orbital intersection line between the servicing spacecraft and the target satellite is determined; The process of the service spacecraft moving to the orbital intersection line is defined as the non-plane transfer phase of the service spacecraft's transfer to the target satellite; The process of the service spacecraft approaching the target satellite from the orbital intersection line is defined as the coplanar transfer phase of the service spacecraft's transfer to the target satellite. Step 2: Using the time and fuel consumed by the servicing spacecraft during the non-planar transfer phase as performance indicators, establish a first cost function for the non-planar transfer phase. Using the time and fuel consumed by the servicing spacecraft during the coplanar transfer phase as performance indicators, establish a second cost function for the coplanar transfer phase. The first and second cost functions are used to evaluate the quality of the transfer orbit, including: Using the time and fuel consumed by the service spacecraft during the off-plane transfer phase as performance indicators, a first cost function for the off-plane transfer phase is established. for: in, The normalized variable representing fuel consumption during the non-planar transfer phase. Normalized variables representing the time consumption of the non-plane transition phase. Indicates the proportionality coefficient; Using the time and fuel consumed by the service spacecraft during the coplanar transfer phase as performance indicators, a second cost function for the coplanar transfer phase is established. for: in, The table represents the normalized variable for fuel consumption during the coplanar transfer phase. Normalized variables representing the time consumption of the coplanar transition phase; Both the first cost function and the second cost function are used to evaluate the quality of the transfer orbit; Step 3: Select the optimal surface transfer orbit for the service spacecraft to transfer to the target satellite using the first cost function; Step 4: Sample the phase angle difference of the target satellite at equal intervals to obtain multiple phase angles, and extract the second cost function and transfer time corresponding to each phase angle as orbital features; Step 5: Nonlinearly map the multiple phase angles to the orbital features to obtain the comprehensive cost and time cost of the optimal orbit in the coplanar transfer phase, which is used to generate the optimal coplanar transfer orbit. Based on the optimal coplanar transfer orbit and the optimal coplanar transfer orbit, the optimal transfer orbit for each service spacecraft to transfer to the corresponding target satellite is obtained. Step 6: Assign a service spacecraft to each target satellite based on the Monte Carlo tree search method, and select the optimal transfer orbit from the optimal transfer orbit of each service spacecraft to the corresponding target satellite through the established large-scale mission scheduling model. Schedule each service spacecraft to the corresponding target satellite through the selected optimal transfer orbit.

2. The spacecraft on-orbit servicing large-scale mission scheduling method according to claim 1, characterized in that, By performing a nonlinear mapping between multiple phase angles and the orbital characteristics, the comprehensive cost and time cost of the optimal orbit in the coplanar transfer stage are obtained, including: Multiple phase angles and orbital features are input into a multilayer perceptron. The multilayer perceptron performs nonlinear mapping processing on the multiple phase angles and orbital features to obtain the comprehensive cost and time cost of the optimal orbit in the coplanar transfer stage, which is used to generate the optimal coplanar transfer orbit. The multilayer perceptron includes an input layer, a first fully connected layer, a second fully connected layer, and an output layer connected in sequence.

3. The spacecraft on-orbit servicing large-scale mission scheduling method according to claim 2, characterized in that, The multilayer perceptron performs nonlinear mapping processing on multiple phase angles and orbital features to obtain the comprehensive cost and time cost of the optimal orbit in the coplanar transfer stage, including: The multiple phase angles and orbital features are transmitted from the input layer to the first fully connected layer via forward propagation, and the forward propagation expression is... for: The first output matrix is ​​obtained by performing a full connection through the first fully connected layer. for: A second fully connected layer is used to perform a full connection, resulting in the second output matrix. for: The output vector through the output layer is: in, This represents a weighted sum of data. The weights representing the phase angles, Indicates the first One phase angle, , , Both represent offsets. Indicates the number of input data. Neurons representing the first and second fully connected layers quantity, This represents the activation function. Indicates the number of output data. and This indicates the combined cost and the time cost.

4. The spacecraft on-orbit servicing large-scale mission scheduling method according to claim 1, characterized in that, The method of assigning service spacecraft to each target satellite based on Monte Carlo tree search includes: Treating each of the service spacecraft as an independent object, the allocation of target satellites is viewed as a group game process; The pairing of a service spacecraft with a target satellite is defined as a state. The set of all states at any given time constitutes the nodes of a Monte Carlo tree. The nodes include a root node representing that none of the service spacecraft are paired with the target satellite, child nodes representing that some of the service spacecraft are paired with the target satellite, and leaf nodes representing that all the service spacecraft are paired with the target satellite. The search begins from the root node, using the formula... The upper confidence interval values ​​for the child nodes are calculated, where, Represents child nodes The average reward value Indicates the total number of explorations. Represents child nodes Number of explorations Indicates the exploration coefficient; Generate new child nodes for child nodes with a non-zero exploration count; Starting from a new child node, the game proceeds through random simulation of state transitions until the game is completed, and the reward value is evaluated. The information of each node on the Monte Carlo tree is updated through backpropagation, including reward value and number of explorations.

5. The spacecraft on-orbit servicing large-scale mission scheduling method according to claim 1, characterized in that, The expression for the large-scale task scheduling model is: in, Represents the allocation coefficient, when Time service spacecraft Assigned to target satellite ,when Time service spacecraft Not assigned to the target satellite , Represents the decision performance function. Describes the optimal decision performance function; The allocation coefficient The following conditions must be met: in, This means that each target satellite must be assigned a service spacecraft. This means that each service spacecraft can dock with a maximum of one target satellite.

6. The spacecraft on-orbit servicing large-scale mission scheduling method according to claim 1, characterized in that, Each of the service spacecraft is scheduled to its corresponding target satellite via a selected optimal transfer orbit, including: The state space is established by taking the target satellite as the state: The action space is established by selecting the service spacecraft for the target satellite as the action set: The comprehensive cost of the optimal transfer trajectory is used as the immediate reward value of the action set, and the reward function is designed as follows: The action sequence is obtained as the scheduling result by solving the following equation: in, Indicates the first Each target satellite is considered as a state. Indicates the first Each service spacecraft is considered as a set of actions. Represents the reward function; For each State selection action Receive timely reward points If the task scheduling result The allocation coefficient is not satisfied. If the conditions are met, a penalty of -1 will be imposed.

7. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the spacecraft on-orbit servicing large-scale mission scheduling method as described in any one of claims 1 to 6.

8. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the spacecraft on-orbit servicing large-scale mission scheduling method as described in any one of claims 1 to 6.

Images