A rolling prediction non-cooperative orbit pursuit-evasion game maneuver control method

Abstract

The application belongs to the technical field of spacecraft control. The application provides a rolling prediction non-cooperative orbit pursuit and escape game maneuver control method. The method comprises: a generative prediction module taking the past positions and control instructions of a pursuit spacecraft and an escape spacecraft as inputs to predict the positions of the pursuit spacecraft in the future for a period of time. A control strategy solution module based on adaptive dynamic programming takes the prediction results of the generative prediction model as inputs, constructs a Critic network to solve the approximate optimal strategy of the pursuit spacecraft, and then uses the approximate optimal strategy as the control instruction of the pursuit spacecraft. The embodiments of the disclosure predict the positions of the pursuit spacecraft and the escape spacecraft based on the generative prediction model, solve the approximate optimal control strategy of the pursuit spacecraft in the game process by using the adaptive dynamic programming method for the prediction results, compensate for the dynamic control instruction error caused by the time delay, and ensure that the pursuit spacecraft occupies a dominant position in the game.

Description

Technical Field

[0001] The disclosed embodiments relate to the field of spacecraft control technology, and more particularly to a rolling prediction non-cooperative orbital pursuit-escape game maneuver control method. Background Art

[0002] With the advancement of space technology, in-orbit spacecraft are playing an increasingly important role in both military and civilian applications. The number of spacecraft launched into space is increasing, and orbital resources are becoming increasingly scarce. Some spacecraft, even after reaching the end of their lifespan or failing, continue to occupy orbit, becoming space debris. Furthermore, some spacecraft will not provide information or cooperate with other spacecraft, and may even pose a threat to space security. Approaching these non-cooperative targets to remove or capture them requires orbital maneuvers.

[0003] The process of competing for dominance with a non-cooperative target is essentially an orbital game: the escaping spacecraft attempts to escape pursuit, while the pursuing spacecraft constantly monitors the position of the escaping spacecraft and its own position, calculating control commands to minimize the distance between them. The transmission of control commands between the ground and the spacecraft involves time delays, resulting in a mismatch between the pursuing spacecraft's position and ground commands. Furthermore, the escaping spacecraft may maneuver during the transmission of control commands from the pursuing spacecraft to escape pursuit. The information loss or asymmetry between the two parties caused by time delays can lead to lags or errors in decision-making and response during the game, resulting in inaccurate or ineffective game strategies, making it easier for the escaping spacecraft to evade the pursuing spacecraft's approach. Therefore, an appropriate delay compensation mechanism is needed to mitigate the adverse effects of delays during the orbital game. Compared with delay predictive control, commonly used delay compensation methods, such as delay-tolerant control and forward error correction coding, lack flexibility, real-time performance, and applicability. However, traditional predictive control methods are limited in their ability to provide long-term predictions. Summary of the Invention

[0004] In order to avoid the shortcomings of the existing technology, the present application provides a rolling prediction non-cooperative orbital pursuit and escape game maneuvering control method to solve the problem of time delay in the transmission of pursuit spacecraft control instructions between the earth and the sky during orbital maneuvers in the existing technology.

[0005] According to an embodiment of the present disclosure, a non-cooperative track pursuit-escape game maneuver control method with rolling prediction is provided, the method comprising:

[0006] Obtaining the position and control instructions of the chasing spacecraft and the position and control instructions of the escaping spacecraft within a preset time period, and processing the position and control instructions of all the chasing spacecraft and the position and control instructions of the escaping spacecraft to obtain a sequence matrix;

[0007] The sequence matrix is ​​processed using a generative prediction model to obtain the encoder input, decoder input, and target sequence;

[0008] Obtaining the current estimated position of the chasing spacecraft and the current estimated position of the escaping spacecraft according to the target sequence;

[0009] Based on adaptive dynamic programming, a critic network is constructed and used to solve the control strategy for the current estimated position of the chasing spacecraft and the current estimated position of the escaping spacecraft during the current delay period, so as to obtain the predicted control instructions for the chasing spacecraft and the escaping spacecraft at each moment during the current delay period.

[0010] According to the predicted control instructions of the chasing spacecraft and the predicted control instructions of the escaping spacecraft at each moment in the time delay duration period, the predicted position of the chasing spacecraft and the predicted position of the escaping spacecraft at each moment in the next time delay duration period are obtained.

[0011] Furthermore, the step of obtaining the position and control instructions of the chasing spacecraft and the position and control instructions of the escaping spacecraft within a preset time period, and processing the position and control instructions of all the chasing spacecraft and the position and control instructions of the escaping spacecraft to obtain a sequence matrix includes:

[0012] Get 0~l tra Tracking the spacecraft's position within seconds and control instructions Location of the escaping spacecraft and control instructions And the position of each spacecraft at each time point and control instructions Location of the escaping spacecraft and control instructions Form a state row vector x (t) ;

[0013] All state row vectors x (t) Stack by time series to get the sequence matrix X train .

[0014] Furthermore, the step of processing the sequence matrix using the generative prediction model to obtain the encoder input, the decoder input, and the target sequence includes:

[0015] The generative prediction model is derived from the sequence matrix X train Select t~t+l seq The state row vector between composes the input of the encoder

[0016] The generative prediction model takes the input from the encoder Select the last l abl state row vectors, and t+l seq ~t+l seq +l pre The state row vectors between are stacked to form the input of the decoder And the predicted target in this data is 0-encoded, and the prediction is from t+l seq The initial length is l pre Target sequence

[0017] Furthermore, the step of obtaining the current estimated position of the chasing spacecraft and the current estimated position of the escaping spacecraft according to the target sequence includes:

[0018] from Remove the front del The state row vectors form the current estimated position of the chasing spacecraft during the current delay period and the current estimated position of the escaping spacecraft

[0019] Furthermore, based on adaptive dynamic programming, a critic network is constructed, and the critic network is used to solve the control strategy for the current estimated positions of the chasing spacecraft and the escaping spacecraft during the current time delay duration period, so as to obtain the predicted control instructions for the chasing spacecraft and the escaping spacecraft at each moment during the current time delay duration period, including the following steps:

[0020] The current estimated position of the chasing spacecraft at the current moment obtained by the chasing spacecraft and the current estimated position of the escaping spacecraft Solve the control strategy based on the Critic network and obtain the control instructions for the chasing spacecraft at each moment during the current delay period and control instructions for the escape spacecraft

[0021] Furthermore, the step of obtaining the predicted position of the chasing spacecraft and the predicted position of the escaping spacecraft at each moment in the next delay duration period based on the predicted control instructions of the chasing spacecraft and the predicted control instructions of the escaping spacecraft at each moment in the delay duration period includes:

[0022] The chasing spacecraft and the escaping spacecraft respectively execute the control instructions of the chasing spacecraft and control instructions for the escape spacecraft

[0023] Based on the control instructions of the chasing spacecraft and control instructions for the escape spacecraft Use the dynamic model to calculate the predicted position of the chasing spacecraft at the next moment and the predicted position of the escaping spacecraft

[0024] Furthermore, the method further comprises:

[0025] The above length is l del The predicted control instructions and predicted estimated positions are added to the sequence matrix, and the first l of the sequence matrix is ​​removed. del row vectors to update the sequence matrix;

[0026] Perform rolling forecast iterations based on the updated sequence matrix.

[0027] The technical solutions provided by the embodiments of the present disclosure may have the following beneficial effects:

[0028] In the embodiment of the present disclosure, the non-cooperative orbital pursuit-escape game maneuver control method based on the above-mentioned rolling prediction is as follows: on the one hand, the generative prediction module uses the past positions and control instructions of the chasing spacecraft and the escaping spacecraft as input to predict the position of the chasing spacecraft in the future. The control strategy solving module based on adaptive dynamic programming uses the prediction results of the generative prediction model as input to construct a Critic network to solve the approximate optimal strategy of the chasing spacecraft, which is then used as the control instruction of the chasing spacecraft. On the other hand, the positions of the chasing spacecraft and the escaping spacecraft are predicted based on the generative prediction model, and the adaptive dynamic programming method is used to solve the approximate optimal control strategy of the chasing spacecraft during the game process for the prediction results, so as to compensate for the dynamic control instruction error caused by time delay and ensure that the chasing spacecraft occupies an advantageous position in the game. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] The accompanying drawings are incorporated into and constitute a part of the specification, illustrate embodiments consistent with the present disclosure, and together with the specification, are used to explain the principles of the present disclosure. Obviously, the drawings described below are only some embodiments of the present disclosure, and those skilled in the art can derive other drawings based on these drawings without inventive effort.

[0030] Figure 1 A diagram showing the steps of a rolling prediction non-cooperative track pursuit-escape game maneuvering control method in an exemplary embodiment of the present disclosure;

[0031] Figure 2 shows a predicted trajectory of a chasing spacecraft in an exemplary embodiment of the present disclosure;

[0032] Figure 3 shows the prediction error of the chasing spacecraft in an exemplary embodiment of the present disclosure;

[0033] Figure 4 shows a predicted trajectory of an escaping spacecraft in an exemplary embodiment of the present disclosure;

[0034] Figure 5 shows the prediction error of the escape spacecraft in an exemplary embodiment of the present disclosure;

[0035] Figure 6 FIG. 2 shows a trajectory prediction tracking error on the x-axis in an exemplary embodiment of the present disclosure;

[0036] Figure 7 shows the trajectory prediction tracking error on the y-axis in an exemplary embodiment of the present disclosure;

[0037] Figure 8 FIG. 4 shows the trajectory prediction tracking error on the z-axis in an exemplary embodiment of the present disclosure. DETAILED DESCRIPTION

[0038] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be embodied in many forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that this disclosure will be thorough and complete and will fully convey the concepts of the example embodiments to those skilled in the art. The described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

[0039] In addition, the accompanying drawings are merely schematic illustrations of embodiments of the present disclosure and are not necessarily drawn to scale. Like reference numerals in the figures represent like or similar parts, and thus repeated descriptions thereof will be omitted. Some of the blocks shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically separate entities.

[0040] This example embodiment provides a rolling prediction non-cooperative track pursuit and escape game maneuver control method. Figure 1 As shown in , the rolling predicted non-cooperative track pursuit and escape game maneuver control method may include: steps S101 to S105.

[0041] Step S101: Acquire the position and control instructions of the chasing spacecraft and the position and control instructions of the escaping spacecraft within a preset time period, and process all the position and control instructions of the chasing spacecraft and the position and control instructions of the escaping spacecraft to obtain a sequence matrix;

[0042] Step S102: Processing the sequence matrix using a generative prediction model to obtain encoder input, decoder input, and a target sequence;

[0043] Step S103: obtaining the current estimated position of the chasing spacecraft and the current estimated position of the escaping spacecraft according to the target sequence;

[0044] Step S104: Based on adaptive dynamic programming, a critic network is constructed, and the critic network is used to solve the control strategy for the current estimated position of the chasing spacecraft and the current estimated position of the escaping spacecraft during the current delay duration period, so as to obtain the predicted control instructions for the chasing spacecraft and the predicted control instructions for the escaping spacecraft at each moment during the current delay duration period;

[0045] Step S105: According to the predicted control instructions of the chasing spacecraft and the predicted control instructions of the escaping spacecraft at each moment in the time delay duration period, the predicted position of the chasing spacecraft and the predicted position of the escaping spacecraft at each moment in the next time delay duration period are obtained.

[0046] Using the aforementioned rolling prediction non-cooperative orbital pursuit-escape game maneuvering control method, a generative prediction module uses the past positions and control commands of the pursuing and escaping spacecraft as input to predict the position of the pursuing spacecraft within a certain period of time. The adaptive dynamic programming-based control strategy solution module uses the prediction results of the generative prediction model as input to construct a critic network to solve for the approximate optimal strategy of the pursuing spacecraft, which is then used as the control command for the pursuing spacecraft. Furthermore, the positions of the pursuing and escaping spacecraft are predicted based on the generative prediction model. The predicted results are then used to solve for the approximate optimal control strategy of the pursuing spacecraft during the game using an adaptive dynamic programming method. This method compensates for the dynamic control command errors caused by time delays and ensures that the pursuing spacecraft maintains a dominant position in the game.

[0047] Below, we will refer to Figures 1 to 8 Each step of the above-mentioned rolling prediction non-cooperative track pursuit-escape game maneuver control method in this example embodiment is described in more detail.

[0048] In steps S101 and S102, in the spacecraft state prediction module, a generative prediction model is used to predict the states of the pursuit spacecraft and the escape spacecraft. The encoder of the generative prediction model can effectively capture long-range dependencies in long sequence inputs. At each time t, the positions and control commands of the pursuit spacecraft and the escape spacecraft are combined into a row vector x (t) ,in represents the position of the chasing spacecraft on the x-axis, y-axis, and z-axis, Indicates the control instructions of the chasing spacecraft on the x-axis, y-axis and z-axis, represents the position of the escape spacecraft on the x-axis, y-axis, and z-axis, Indicates the control instructions of the escape spacecraft on the x-axis, y-axis and z-axis. Encoder input The length of the column is l, starting from the tth row vector. seqThe matrix can be expressed as

[0049] The prediction model uses a generative decoder that can predict the entire long time series at once, rather than doing it step by step. This approach greatly improves the inference speed of long sequence outputs. Take the last l abl row vectors, plus the Start taking l pre The row vectors form the input of the decoder in X tar is a sequence matrix composed of target columns to be predicted and marked with placeholder 0, X to for The final output is obtained through a fully connected layer, and the output size depends on the number of columns of the target column.

[0050] In step S103, in the control strategy solution module, the critic network is used to solve the approximate optimal strategy for chasing the spacecraft. During the pursuit process, the distance between the spacecraft is much smaller than the orbital altitude, and its dynamics can be described by the CW equation:

[0051]

[0052] in,

[0053]

[0054] in, Represents the average speed of the reference track. After removing the time label, the differential game model of formula (1) can be expressed as is equivalent to:

[0055]

[0056] Where x = x p -x e describes the position difference between two satellites, and g = k. Considering a special cost function:

[0057]

[0058] Where Q(x)=x T Q x and is a positive definite diagonal matrix, and

[0059]

[0060] Among them, λ1>λ2>0, l c =(1,1,1) T, R1 and R2 are positive definite diagonal matrices, and and is a row vector consisting of the diagonal elements of R1 and R2.

[0061] Differentiating formula (4) yields the Hamilton-Jacobi-Isaacs (HJI) equation where r(x,u,w)=Q(x)+U(u)-W(w) and

[0062] Solving the HJI equation The optimal value function V can be obtained * The Nash equilibrium solution is:

[0063]

[0064] In order to approach the optimal solution, we can define an optimal value function V based on a neural network * (x) = W T σ(x)+ε, where represents the network weight, is the activation function, and ε is the approximation error. Therefore, the HJI equation can be rewritten as:

[0065]

[0066] Among them, D1, D2 are and

[0067] It is challenging to directly obtain the optimal control in Equation (7), considering that the predicted state generated by the generative prediction model and Design the predictive control of the chasing spacecraft and the approximate optimal control of the escaping spacecraft as:

[0068]

[0069] Among them, the adaptive learning rate of network weights is designed as:

[0070]

[0071] in, ε x is the double prediction error, whose boundedness can be guaranteed by the generative prediction model, l c is the learning rate, choose Satisfy and

[0072] The approximation error of the neural network can be defined as Combining formulas (8) and (11), we can obtain The derivative of the approximation error is:

[0073]

[0074] in

[0075] Consider the Lyapunov function L=L1+L2+L3, L1=V, L3=l s J s , and its derivative can be expressed as:

[0076]

[0077]

[0078] in,

[0079]

[0080] According to Lagrange mean theorem, tanh(a)-tanh(b)=(1-tanh 2 (c))(ab), and b≤c≤a, we get

[0081]

[0082] Because the activation function of the neural network The activation function can be chosen to be continuously differentiable so that is bounded, and |ε W |<b W Therefore, formula (16) can be rewritten as

[0083]

[0084] The differential of L3 is:

[0085]

[0086] in

[0087] Combining formulas (13), (14) and (18), we get The differential of is:

[0088] (19)

[0089] and

[0090] You can choose get:

[0091]

[0092] and It shows that the state is asymptotically stable.

[0093] In a specific embodiment, the algorithm block diagram of the rolling prediction non-cooperative track pursuit and escape game maneuver control method is as follows:

[0094]

[0095] In a specific embodiment, it is assumed that the escaping spacecraft and the pursuing spacecraft are both point masses, and the influence of their own volumes is ignored during the orbital game. The orbital altitude is 600 km, and the initial separation between the two spacecraft is (5, 5, 5) km. The speeds are (-0.003, 0.008, 0.005) km / s respectively. The sampling interval is set to 1 second, and the delay is set to 4 seconds. The generative prediction model receives a 400-second sequence as input and predicts data for the next 24 seconds. The first 4 seconds of the prediction results are selected as input to the policy model, and control instructions are solved and implemented to achieve dynamic closed-loop control. The weight model of the generative prediction model is updated every 4 seconds, as the sequence data used for training is updated. This process is iterated in a rolling manner.

[0096] The prediction effect indicates the robustness of the rolling prediction method to the spacecraft trajectory prediction, such as Figures 2 to 5 As shown. Among them, Figure 2 In, x p1 ,x p2 ,x p3 It means that in the absence of time delay, the chasing spacecraft adopts the adaptive dynamic programming method (Critic) to solve the control strategy and updates the dynamics to obtain the position on the x-axis, y-axis and z-axis. It means that the pursuit spacecraft solves the control strategy according to the position predicted by the generative model in the case of time delay, and updates the dynamics to obtain the position on the x-axis, y-axis and z-axis. Figure 3 The values ​​in the figure represent the error between the x-, y-, and z-axis positions of the chasing spacecraft obtained using the rolling prediction iteration with a time delay and the x-, y-, and z-axis positions obtained using the pure strategy solution without a time delay. This error is within a reasonable range, indicating a good prediction performance. Note that before 400 seconds, there was no time delay and the pure strategy solution was used, so the error was zero. Figure 4 In, x e1 ,x e2 ,x e3 It means that in the absence of time delay, the escaping spacecraft adopts the adaptive dynamic programming method (Critic) to solve the control strategy and updates the dynamics to obtain the positions on the x-axis, y-axis and z-axis. It means that in the case of a time delay, the escape spacecraft solves the control strategy based on the position predicted by the generative model and updates the dynamics to obtain the position on the x-axis, y-axis and z-axis. Figure 5 The values ​​in the figure represent the error between the x-, y-, and z-axis positions of the escaping spacecraft obtained using the rolling prediction iteration with a time delay and the x-, y-, and z-axis positions obtained using the pure strategy solution without a time delay. This error is within a reasonable range, indicating a good prediction performance. It should be noted that before 400 seconds, there was no time delay and the pure strategy solution was used, so the error was zero.

[0097] It can be found that all error trajectories converge to a bounded region, which is consistent with the design requirements of stability analysis and generative prediction models.

[0098] The predicted tracking error represents the tracking effect of the chasing spacecraft on the escaping spacecraft. The chasing spacecraft uses the trajectory obtained by rolling prediction, while the escaping spacecraft uses the trajectory obtained by the simple strategy solution. The error is obtained by subtracting the position of the escaping spacecraft on each axis from the position of the chasing spacecraft on each axis. 100 experiments were conducted, and the errors obtained on the x-axis, y-axis, and z-axis were statistically processed. The mean and variance of the errors were calculated, as shown in the following figure. Figure 6 、 Figure 7 、 Figure 8 The colored line in the middle represents the mean, and the shaded area represents the calculated variance. The results (with the error gradually approaching 0) show that the pursuing spacecraft consistently approaches the escaping spacecraft and maintains a dominant position in the pursuit-escape game. Furthermore, when tracking the escaping spacecraft with a time delay but without prediction, the pursuing spacecraft's tracking of the escaping spacecraft diverges. Simulation results demonstrate that the proposed rolling prediction method is effective in compensating for the error introduced by the time delay.

[0099] Using the aforementioned rolling prediction non-cooperative orbital pursuit-escape game maneuvering control method, a generative prediction module uses the past positions and control commands of the pursuing and escaping spacecraft as input to predict the position of the pursuing spacecraft within a certain period of time. The adaptive dynamic programming-based control strategy solution module uses the prediction results of the generative prediction model as input to construct a critic network to solve for the approximate optimal strategy of the pursuing spacecraft, which is then used as the control command for the pursuing spacecraft. Furthermore, the positions of the pursuing and escaping spacecraft are predicted based on the generative prediction model. The predicted results are then used to solve for the approximate optimal control strategy of the pursuing spacecraft during the game using an adaptive dynamic programming method. This method compensates for the dynamic control command errors caused by time delays and ensures that the pursuing spacecraft maintains a dominant position in the game.

[0100] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood to indicate or imply relative importance or implicitly indicate the number of the technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of the embodiments of the present disclosure, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0101] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present disclosure. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art can combine and combine different embodiments or examples described in this specification.

[0102] Those skilled in the art will readily appreciate other embodiments of the present disclosure after considering the specification and practicing the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of the present disclosure that follow the general principles of the present disclosure and include common knowledge or customary techniques in the art not disclosed herein. The description and examples are to be considered as exemplary only, with the true scope and spirit of the present disclosure being indicated by the appended claims.

Claims

1. A rolling prediction non-cooperative track pursuit and escape game maneuver control method, characterized by: The method includes: Obtaining the position and control instructions of the chasing spacecraft and the position and control instructions of the escaping spacecraft within a preset time period, and processing the position and control instructions of all the chasing spacecraft and the position and control instructions of the escaping spacecraft to obtain a sequence matrix; The sequence matrix is ​​processed using a generative prediction model to obtain the encoder input, decoder input, and target sequence; Obtaining the current estimated position of the chasing spacecraft and the current estimated position of the escaping spacecraft according to the target sequence; Based on adaptive dynamic programming, a critic network is constructed and used to solve the control strategy for the current estimated position of the chasing spacecraft and the current estimated position of the escaping spacecraft during the current delay period, so as to obtain the predicted control instructions for the chasing spacecraft and the escaping spacecraft at each moment during the current delay period. According to the predicted control instructions of the chasing spacecraft and the predicted control instructions of the escaping spacecraft at each moment during the time delay duration period, a predicted position of the chasing spacecraft and a predicted position of the escaping spacecraft at each moment during the next time delay duration period are obtained; The step of obtaining the position and control instructions of the chasing spacecraft and the position and control instructions of the escaping spacecraft within a preset time period, and processing the position and control instructions of all the chasing spacecraft and the position and control instructions of the escaping spacecraft to obtain a sequence matrix includes: Get 0~ l tra Tracking the spacecraft's position within seconds and control instructions , the location of the escaping spacecraft and control instructions , and the position of each spacecraft at each time point and control instructions , the location of the escaping spacecraft and control instructions Form a state row vector ; All state row vectors Stack by time series to get a sequence matrix ; The steps of processing the sequence matrix using the generative prediction model to obtain the encoder input, the decoder input, and the target sequence include: Generative prediction model from sequence matrix Select t ~ The state row vector between composes the input of the encoder ; The generative prediction model takes the input from the encoder Select the last state row vectors, and ~ The state row vectors between are stacked to form the input of the decoder , and encode the predicted target in this data as 0, predicting from The starting length is Target sequence ; The step of obtaining the current estimated position of the chasing spacecraft and the current estimated position of the escaping spacecraft according to the target sequence includes: from Before removal The state row vectors form the current estimated position of the chasing spacecraft during the current delay period and the current estimated position of the escaping spacecraft ; Based on adaptive dynamic programming, a critic network is constructed, and the critic network is used to solve the control strategy for the current estimated positions of the chasing spacecraft and the escaping spacecraft during the current delay duration period, so as to obtain the predicted control instructions for the chasing spacecraft and the escaping spacecraft at each moment during the current delay duration period, including the following steps: The current estimated position of the chasing spacecraft at the current moment obtained by the chasing spacecraft and the current estimated position of the escaping spacecraft , solve the control strategy based on the Critic network and obtain the control instructions for the chasing spacecraft at each moment during the current delay duration and control instructions for the escape spacecraft .

2. The non-cooperative track pursuit and escape game maneuvering control method of rolling prediction according to claim 1 is characterized in that: The step of obtaining the predicted position of the chasing spacecraft and the predicted position of the escaping spacecraft at each moment in the next delay duration period according to the predicted control instruction of the chasing spacecraft and the predicted control instruction of the escaping spacecraft at each moment in the delay duration period includes: The chasing spacecraft and the escaping spacecraft respectively execute the control instructions of the chasing spacecraft and control instructions for the escape spacecraft ; Based on the control instructions of the chasing spacecraft and control instructions for the escape spacecraft , using the dynamic model to solve the predicted position of the chasing spacecraft at the next moment and the predicted position of the escaping spacecraft .

3. The non-cooperative track pursuit and escape game maneuvering control method of rolling prediction according to claim 2 is characterized in that: The method further includes: The above length is The predicted control instructions and predicted estimated positions are added to the sequence matrix, and the front of the sequence matrix is ​​removed. row vectors to update the sequence matrix; Perform rolling forecast iterations based on the updated sequence matrix.

Images