Total error iterative learning control method for variable tracking task
The full-error iterative learning control method addresses sensor noise and path length changes in observation satellites by using a random noise model and learning gain matrix, ensuring precise path tracking and enhanced robustness.
Patent Information
- Application Number
- CN202411331449.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-24
- Publication Date
- 2024-12-27
AI Technical Summary
Existing observation satellite tracking methods face challenges due to sensor noise, non-zero Markov parameter matrices not meeting full rank conditions, and iterative changes in path length, making precise path tracking difficult.
A full-error iterative learning control method is developed, incorporating a random noise model, learning gain matrix, and variable path length model to enhance control performance and robustness, using a multiple-input multiple-output time-varying satellite system with a finite-state Markov chain representation of path length changes and martingale difference sequences for noise modeling.
The method achieves high-precision path tracking under random interference and path length variations, improving the control performance and reliability of observation satellites.
Smart Images
Figure CN119200405A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of iterative learning control technology, and in particular to a full-error iterative learning control method for variable tracking tasks, and in particular to a full-error iterative learning control method for variable tracking tasks to achieve variable path target position tracking. Background Art
[0002] In recent years, with the increasing demand for earth observation data in fields such as global climate change, natural disaster monitoring, and agricultural management, research on repeated missions for observation satellites has become a hot topic. The advancement of satellite technology, especially the development of small satellites and satellite constellations, has significantly improved the observation frequency and resolution, and the demand for repeated mission scheduling and optimization has become more urgent. Through optimization methods such as mixed integer programming and heuristic algorithms, the observation coverage and data quality can be maximized while minimizing the waste of satellite resources. In addition, international cooperation and data sharing play an important role in the global observation system, and more efficient repeated observation missions can be achieved by coordinating satellite resources from different countries and organizations. Therefore, the research on repeated tracking missions of observation satellites has important practical significance and broad application prospects.
[0003] Existing observation satellite tracking mission control methods include traditional ground control, autonomous control and joint control. Traditional ground control methods have the advantages of high precision and high reliability, but have the disadvantages of time delay and poor flexibility. Autonomous control methods can quickly respond to environmental changes and reduce dependence on ground control, but their implementation is complex and reliability needs to be improved. Joint control methods combine the advantages of ground control and autonomous control, taking into account both timeliness and precision, but the system complexity and cost are relatively high.
[0004] The solution ideas of the above control methods are all based on time-domain feedback control. In this case, it is difficult to achieve complete path tracking within a limited time. In practical applications, when iterative learning control is applied to repeated observation tasks of observation satellites, there are the following difficulties: (1) Due to the existence of sensor measurement noise and environmental interference, the observation satellite will inevitably be affected by random noise; (2) The non-zero Markov parameter matrix corresponding to the observation satellite system usually does not meet the full rank condition; (3) Due to the existence of output constraints, the path length of the observation satellite in each iteration is different. Figure 1 For example, the path length is 8 in the initial iteration, 10 in the first iteration, and 6 in the third iteration.
[0005] Therefore, the full error information stochastic iterative learning control method for variable path length tracking tasks is a key and difficult problem that needs to be solved urgently. Summary of the invention
[0006] In view of the above problems, the present invention proposes a full-error iterative learning control method for variable tracking tasks. The present invention takes into account the fact that the observation satellite system will inevitably be affected by random noise during its operation, the problem of iterative changes in path length due to output limitations, and the problem that the non-zero Markov parameter matrix of the observation satellite system usually does not satisfy the full rank condition. In view of the above problems, a random noise model, a learning gain matrix, and a variable path length model of the full tracking state of the observation satellite system are designed, and a full-error iterative learning control model for variable tracking tasks is further obtained, which solves the problem of high-precision full path tracking control of the observation satellite under the influence of random interference and iterative changes in path length, thereby achieving the purpose of improving the control performance of the observation satellite and enhancing the robustness and reliability of the system.
[0007] The present invention provides a full-error iterative learning control method for variable tracking tasks, comprising:
[0008] Step S1, presetting the target position of the observation satellite system at each operating time based on the variable path length model;
[0009] Step S2, determining the input of the observation satellite system at each operating time and obtaining the random noise at the corresponding time;
[0010] Step S3, establishing an original full-error iterative learning control model for variable tracking tasks; using the input of the observation satellite system at each operating moment and the random noise at the corresponding moment as a training set;
[0011] The original full-error iterative learning control model for variable tracking tasks is trained based on the training set to obtain a full-error iterative learning control model for variable tracking tasks;
[0012] Based on the full-error iterative learning control model for the variable tracking task, the position at each running time is obtained;
[0013] Step S4, correcting the position of each operating moment based on the target position of the observation satellite system at each operating moment to obtain an updated position at each operating moment, and obtaining a tracking error at each operating moment based on the updated position at each operating moment;
[0014] Step S5, establishing a learning gain matrix; updating the tracking error at each running moment based on the learning gain matrix to obtain an updated tracking error at each running moment;
[0015] Step S6, combining the input of each operating time of the observation satellite system with the updated tracking error of each operating time, to obtain the updated input of each operating time of the observation satellite system;
[0016] Step S7: Based on the update input of the observation satellite system at each operating time, the updated position of the observation satellite system at each operating time is obtained, and the observation satellite system completes tracking of the target position at each operating time.
[0017] Preferably, the observation satellite system is a multiple-input multiple-output time-varying observation satellite system;
[0018] Furthermore, the expression of the observation satellite system is:
[0019]
[0020] Among them, x k (t) is the state of the observation satellite system at time t in the kth iteration, t∈{0,1,2,…,T} is the operating time of the observation satellite system, k=1,2,3…N k ,N k is the iteration cycle of the observation satellite system during operation. is the number of iterations, represents the set of non-negative integers {0,1,2,…}; u k (t) is the input of the observation satellite system at time t in the kth iteration, y k (t) is the position of the observation satellite system at time t in the kth iteration; A(t) is the first time-varying system matrix corresponding to the observation satellite system at time t, B(t) is the second time-varying system matrix corresponding to the observation satellite system at time t, and C(t) is the third time-varying system matrix corresponding to the observation satellite system at time t; v k (t) is the random noise of the observation satellite system at time t in the kth iteration.
[0021] Preferably, step S1 further includes the specific steps of obtaining the variable path length model in step S1, including:
[0022] Obtaining scenarios of observation satellite systems when performing repetitive path tracking missions;
[0023] The iteratively changing path length in the scene is obtained, and the iteratively changing path length in the scene is set to be tracked by the observation satellite system, so as to obtain the path length of multiple iteratively changing complete tracking states;
[0024] The path length of multiple iteratively changing full tracking states is modeled as a finite-state Markov chain to obtain a variable path length model;
[0025] Preferably, the full tracking state of each path length is returned normally.
[0026] Preferably, the scenarios in which the observation satellite system performs a repeated path tracking task include:
[0027] In high-resolution observation missions, due to the change in terrain height, the path length of the satellite must be dynamically adjusted to maintain a stable observation trajectory;
[0028] For dynamic target monitoring tasks, such as tracking ships or vehicles, the target's moving speed and direction changes will cause the satellite to constantly adjust its path length to maintain continuous observation of the target;
[0029] Environmental interference, such as weather changes, affects the quality of observations, requiring satellites to adjust path lengths to avoid or reduce interference effects; when satellites switch between different missions, such as switching from large-scale terrain scanning to high-precision local observations, path lengths need to be adjusted according to mission requirements.
[0030] The present invention models the iteratively changing path length as a finite-state Markov chain. The modeling method can use the modeling method based on probability distribution as a special case, eliminating the need for prior statistical information.
[0031] Preferably, the random noise in step S2 includes sensor noise, atmospheric disturbance, and noise caused by manufacturing and assembly errors;
[0032] Exemplarily, the sensor noise is generated by measurement errors of gyroscopes, accelerometers and star sensors; the atmospheric disturbances are caused by uneven atmospheric density and short-term changes in atmospheric density caused by the Earth's orbit; and the manufacturing and assembly errors are generated during the satellite manufacturing and assembly process and accumulated over a long period of time during orbital operation.
[0033] Preferably, the specific steps of obtaining the random noise at the corresponding moment in step S2 include:
[0034] Acquire random noise generated by the observation satellite system when performing a task; establish a random noise flow domain; and establish a martingale difference sequence of random noise based on the random noise flow domain;
[0035] Obtaining the random noise at the corresponding moment based on the martingale difference sequence of the random noise;
[0036] Furthermore, the random noise watershed expression is:
[0037]
[0038] Among them, A k is the watershed of random noise at the kth iteration; represents the element in the flow domain of the kth iteration, σ is algebraic, v k (t) is the noise of the observation satellite system at the kth iteration at time t.
[0039] Furthermore, the martingale difference sequence expression of the random noise is:
[0040]
[0041] Among them, v k+1 (t) represents the random noise of the observation satellite system at the k+1th iteration at time t, is the random noise of the observation satellite system at the k+1th iteration at time t relative to the basin conditional expectation.
[0042] Preferably, the specific steps of obtaining the full-error iterative learning control model for the variable tracking task in step S3 include:
[0043] Establish the original full-error iterative learning control model for variable tracking tasks;
[0044] Determine the learning gain matrix related parameters; obtain the learning gain matrix based on the learning gain matrix related parameters;
[0045] Obtaining inputs of the observation satellite system at each operating time; obtaining random noises of the observation satellite system at each operating time based on a martingale difference sequence of a random noise watershed; and using the inputs of the observation satellite system at each operating time and the random noise as a training set;
[0046] Based on the training set and the learning gain matrix, the original full-error iterative learning control model for the variable tracking task is trained to obtain the full-error iterative learning control model for the variable tracking task;
[0047] Furthermore, the specific steps of obtaining the position at each running time based on the full error iterative learning control model for the variable tracking task in step S3 include:
[0048] S31, let t = 1, when t = 1, it is the initial operation time of the observation satellite system; let k = 1, when k = 1, it is the first iteration of the operation time t;
[0049] Preset the target position of the observation satellite system at time t based on the variable path length model;
[0050] S32, obtaining the random noise of the kth iteration at the operating time t of the observation satellite system based on the martingale difference sequence of the random noise;
[0051] S33, determining the input of the k-th iteration of the observation satellite system at the operation time t; inputting the input of the k-th iteration of the observation satellite system at the operation time t and random noise into the full error iterative learning control model for the variable tracking task at the operation time t k, and obtaining the position of the observation satellite system at the operation time t k;
[0052] S34, based on the target position of the observation satellite system at operating time t, the position of the k-th iteration at operating time t is corrected to obtain an updated position of the k-th iteration at operating time t, and based on the updated position of the k-th iteration at operating time t, a tracking error of the k-th iteration at operating time t is obtained;
[0053] S35, updating the k-th iteration tracking error at running time t based on the learning gain matrix to obtain the k-th iteration updated tracking error at running time t;
[0054] S36, combining the input of the k-th iteration of the observation satellite system at the operation time t with the tracking error updated at the k-th iteration of the operation time t, to obtain the updated input of the k-th iteration of the observation satellite system at the operation time t; and using the updated input of the k-th iteration of the operation time t as the input of the next iteration;
[0055] S37, determine whether k is greater than or equal to N k , N k is the total number of iterations within the preset running time. If yes, stop the calculation and obtain the updated input of the observation satellite system running time t. Based on the updated input of the observation satellite system running time t, obtain the updated position of the observation satellite system running time t and record it, complete the tracking of the target position at the running time t, and enter the next step; if not, set k=k+1 and return to step S32.
[0056] S38. Determine whether t is greater than or equal to T, where T represents the total operating time of the observation satellite system. If so, stop the calculation to obtain an updated full-error iterative learning control model for the variable tracking task; if not, set t=t+1 and return to step S32.
[0057] Furthermore, the learning gain matrix related parameters include a random noise attenuation factor, a first learning parameter and a second learning parameter; the second learning parameter is used to solve the traceability problem of the observation satellite system;
[0058] The first learning parameter is used to ensure the convergence of the full error iterative learning control model for variable tracking tasks;
[0059] Furthermore, the update position expression of the kth iteration at the running time t is:
[0060]
[0061] in, is the updated position of the kth iteration of the observation satellite system at time t, N k is the path length of the observation satellite system at the kth iteration, N H The path with the longest length is set to the path full tracking state of the observation satellite, that is, the full tracking state y d (t) is the target path of the observation satellite system at operation time t.
[0062] Furthermore, the tracking error expression is:
[0063]
[0064] in, is the tracking error of the observed satellite system at the kth iteration at the running time t.
[0065] Furthermore, the expression of the full error iterative learning control model for the variable tracking task is:
[0066]
[0067] Among them, u k+1 (t) is the observation satellite system input at the k+1th iteration of the running time t, u k (t) is the observation satellite system input at the kth iteration of the running time t, T′ is the time corresponding to the longest path, is the tracking error at the kth iteration at runtime t+1, is the learning gain matrix of the kth iteration of path i corresponding to the observation satellite system at the running time t, i = 1, 2, 3...I, I is the total number of paths, when i = I, it means the longest path N H , a k is the random noise attenuation factor of the observation satellite system at the kth iteration, which is used to overcome the influence of random noise, Γ t is the first learning parameter of the observation satellite at the operating time t, which is used to ensure the convergence of the full error iterative learning control model for variable tracking tasks. t,i It is the second learning parameter of the path length i corresponding to the observation satellite at the operating time t, which is used to overcome the traceability problem of the observation satellite system.
[0068] Furthermore, step S3 also includes learning gain matrix related parameters and learning gain matrix related parameter constraints;
[0069] The learning gain matrix related parameters include a random noise attenuation factor, a first learning parameter, and a second learning parameter;
[0070] The learning gain matrix related parameter constraints include a random noise attenuation factor constraint and a first learning parameter constraint;
[0071] Furthermore, the constraint condition of the random noise attenuation factor is expressed as:
[0072]
[0073] The expression of the first learning parameter constraint condition is:
[0074] Re(λ k (Γ t G t,i ))>0
[0075] Among them, Re(·) is the eigenvalue function, Γ t G t,i is the algorithm condition matrix, Re(λ k (Γ t G t,i ))>0, indicating that the eigenvalues of the algorithm condition matrix all have positive real parts, λ k is the parameter of the kth iteration,
[0076] G t,i is the matrix determined by the observation satellite system and the second learning parameter,
[0077] A(t) is the first time-varying system matrix of the observation satellite system at the operating time t, C(t) is the second time-varying system matrix of the observation satellite at the operating time t, B(t-1) is the third time-varying system matrix of the observation satellite at the operating time t-1, η∈T, J t,i is the second learning parameter of the path length i corresponding to the observation satellite at the operating time t.
[0078] The present invention also includes step S8, using a full-error iterative learning control model for variable tracking tasks to ensure the precise tracking effect of the observation satellite system.
[0079] Specifically, the specific steps to ensure the precise tracking effect of the observation satellite system include:
[0080] Using Lyapunov function and stochastic approximation method, the convergence of the total error iterative learning control model for variable tracking task is analyzed in the mean square sense and in the probability 1 sense.
[0081] Based on the convergence in the mean square sense and in the probability 1 sense, the stability and accuracy of the full error iterative learning control model for variable tracking tasks under random noise are obtained;
[0082] Based on the stability and accuracy under random noise, the precise tracking effect of the observation satellite system is guaranteed.
[0083] Compared with the prior art, the present invention has at least the following beneficial effects:
[0084] (1) The full-error iterative learning control method for variable tracking tasks of the present invention can effectively overcome the influence of random noise on the tracking effect and realize the precise tracking of the specified path by the observation satellite;
[0085] (2) The full error iterative learning control method for variable tracking tasks of the present invention is applicable to the scenario where the non-zero Markov parameter matrix of the system does not satisfy the full rank condition, thereby improving the applicability of the observation satellite system;
[0086] (3) The full-error iterative learning control method for variable tracking tasks of the present invention is applicable to the scenario where the path length changes iteratively, thereby improving the practicality of random iterative learning control;
[0087] (4) The full-error iterative learning control method for variable tracking tasks of the present invention ensures that the observation satellite system fully tracks the specified path from beginning to end within a period of time. BRIEF DESCRIPTION OF THE DRAWINGS
[0088] The drawings are only for the purpose of illustrating particular embodiments and are not to be construed as limiting the invention.
[0089] Figure 1 A schematic diagram of the iteratively changing path length of the present invention;
[0090] Figure 2 It is a schematic diagram of a full error information random iterative learning control block diagram for variable path length tracking tasks of the present invention;
[0091] Figure 3 It is a schematic diagram of a flow chart of a full error information stochastic iterative learning control algorithm for variable path length tracking tasks in an embodiment of the present invention;
[0092] Figure 4 It is a schematic diagram of a simulation diagram of observing satellite path tracking in an embodiment of the present invention. DETAILED DESCRIPTION
[0093] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments can be combined with each other. In addition, the present invention can also be implemented in other ways different from those described herein, and therefore, the protection scope of the present invention is not limited by the specific embodiments disclosed below.
[0094] A specific embodiment of the present invention, as Figure 1-4 , discloses a full error iterative learning control method for variable tracking tasks. In order to illustrate the effectiveness of the method proposed in the present invention, the above technical solution of the present invention is described in detail through a specific embodiment below. The specific implementation steps are as follows:
[0095] The present invention provides a full-error iterative learning control method for variable tracking tasks, comprising:
[0096] Step S1, presetting the target position of the observation satellite system at each operating time based on the variable path length model;
[0097] Step S2, determining the input of the observation satellite system at each operating time and obtaining the random noise at the corresponding time;
[0098] Step S3, establishing an original full-error iterative learning control model for variable tracking tasks; using the input of the observation satellite system at each operating moment and the random noise at the corresponding moment as a training set;
[0099] The original full-error iterative learning control model for variable tracking tasks is trained based on the training set to obtain a full-error iterative learning control model for variable tracking tasks;
[0100] Based on the full-error iterative learning control model for the variable tracking task, the position at each running time is obtained;
[0101] Step S4, correcting the position of each operating moment based on the target position of the observation satellite system at each operating moment to obtain an updated position at each operating moment, and obtaining a tracking error at each operating moment based on the updated position at each operating moment;
[0102] Step S5, establishing a learning gain matrix; updating the tracking error at each running moment based on the learning gain matrix to obtain an updated tracking error at each running moment;
[0103] Step S6, combining the input of each operating time of the observation satellite system with the updated tracking error of each operating time, to obtain the updated input of each operating time of the observation satellite system;
[0104] Step S7: Based on the update input of the observation satellite system at each operating time, the updated position of the observation satellite system at each operating time is obtained, and the observation satellite system completes tracking of the target position at each operating time.
[0105] Preferably, the observation satellite system is a multiple-input multiple-output time-varying observation satellite system;
[0106] Furthermore, the expression of the observation satellite system is:
[0107]
[0108] Among them, x k(t) is the state of the observation satellite system at time t in the kth iteration, t∈{0,1,2,…,T} is the operating time of the observation satellite system, k=1,2,3…N k ,N k is the iteration cycle of the observation satellite system during operation. is the number of iterations, represents the set of non-negative integers {0,1,2,…}; u k (t) is the input of the observation satellite system at time t in the kth iteration, y k (t) is the position of the observation satellite system at time t in the kth iteration; A(t) is the first time-varying system matrix corresponding to the observation satellite system at time t, B(t) is the second time-varying system matrix corresponding to the observation satellite system at time t, and C(t) is the third time-varying system matrix corresponding to the observation satellite system at time t; v k (t) is the random noise of the observation satellite system at time t in the kth iteration.
[0109] Preferably, step S1 further includes the specific steps of obtaining the variable path length model in step S1, including:
[0110] Obtaining scenarios of observation satellite systems when performing repetitive path tracking missions;
[0111] The iteratively changing path length in the scene is obtained, and the iteratively changing path length in the scene is set to be tracked by the observation satellite system, so as to obtain the path length of multiple iteratively changing complete tracking states;
[0112] The path length of multiple iteratively changing full tracking states is modeled as a finite-state Markov chain to obtain a variable path length model;
[0113] Preferably, the full tracking state of each path length is returned normally.
[0114] Preferably, the scenarios in which the observation satellite system performs a repeated path tracking task include:
[0115] In high-resolution observation missions, due to the change in terrain height, the path length of the satellite must be dynamically adjusted to maintain a stable observation trajectory;
[0116] For dynamic target monitoring tasks, such as tracking ships or vehicles, the target's moving speed and direction changes will cause the satellite to constantly adjust its path length to maintain continuous observation of the target;
[0117] Environmental interference, such as weather changes, affects the quality of observations, requiring satellites to adjust path lengths to avoid or reduce interference effects; when satellites switch between different missions, such as switching from large-scale terrain scanning to high-precision local observations, path lengths need to be adjusted according to mission requirements.
[0118] The present invention models the iteratively changing path length as a finite-state Markov chain. The modeling method can use the modeling method based on probability distribution as a special case, eliminating the need for prior statistical information.
[0119] Preferably, the random noise in step S2 includes sensor noise, atmospheric disturbance, and noise caused by manufacturing and assembly errors;
[0120] Exemplarily, the sensor noise is generated by measurement errors of gyroscopes, accelerometers and star sensors; the atmospheric disturbances are caused by uneven atmospheric density and short-term changes in atmospheric density caused by the Earth's orbit; and the manufacturing and assembly errors are generated during the satellite manufacturing and assembly process and accumulated over a long period of time during orbital operation.
[0121] Preferably, the specific steps of obtaining the random noise at the corresponding moment in step S2 include:
[0122] Acquire random noise generated by the observation satellite system when performing a task; establish a random noise flow domain; and establish a martingale difference sequence of random noise based on the random noise flow domain;
[0123] Obtaining the random noise at the corresponding moment based on the martingale difference sequence of the random noise;
[0124] Furthermore, the random noise watershed expression is:
[0125]
[0126] Among them, A k is the watershed of random noise at the kth iteration; represents the element in the flow domain of the kth iteration, σ is algebraic, v k (t) is the noise of the observation satellite system at the kth iteration at time t.
[0127] Furthermore, the martingale difference sequence expression of the random noise is:
[0128]
[0129] Among them, v k+1 (t) represents the random noise of the observation satellite system at the k+1th iteration at time t, is the random noise of the observation satellite system at the k+1th iteration at time t relative to the basin conditional expectation.
[0130] Preferably, the specific steps of obtaining the full-error iterative learning control model for the variable tracking task in step S3 include:
[0131] Establish the original full-error iterative learning control model for variable tracking tasks;
[0132] Determine the learning gain matrix related parameters; obtain the learning gain matrix based on the learning gain matrix related parameters;
[0133] Obtaining inputs of the observation satellite system at each operating time; obtaining random noises of the observation satellite system at each operating time based on a martingale difference sequence of a random noise watershed; and using the inputs of the observation satellite system at each operating time and the random noise as a training set;
[0134] Based on the training set and the learning gain matrix, the original full-error iterative learning control model for the variable tracking task is trained to obtain the full-error iterative learning control model for the variable tracking task;
[0135] Furthermore, the specific steps of obtaining the position at each running time based on the full error iterative learning control model for the variable tracking task in step S3 include:
[0136] S31, let t = 1, when t = 1, it is the initial operation time of the observation satellite system; let k = 1, when k = 1, it is the first iteration of the operation time t;
[0137] Preset the target position of the observation satellite system at time t based on the variable path length model;
[0138] S32, obtaining the random noise of the kth iteration at the operating time t of the observation satellite system based on the martingale difference sequence of the random noise;
[0139] S33, determining the input of the k-th iteration of the observation satellite system at the operation time t; inputting the input of the k-th iteration of the observation satellite system at the operation time t and random noise into the full error iterative learning control model for the variable tracking task at the operation time t k, and obtaining the position of the observation satellite system at the operation time t k;
[0140] S34, based on the target position of the observation satellite system at operating time t, the position of the k-th iteration at operating time t is corrected to obtain an updated position of the k-th iteration at operating time t, and based on the updated position of the k-th iteration at operating time t, a tracking error of the k-th iteration at operating time t is obtained;
[0141] S35, updating the k-th iteration tracking error at running time t based on the learning gain matrix to obtain the k-th iteration updated tracking error at running time t;
[0142] S36, combining the input of the k-th iteration of the observation satellite system at the operation time t with the tracking error updated at the k-th iteration of the operation time t, to obtain the updated input of the k-th iteration of the observation satellite system at the operation time t; and using the updated input of the k-th iteration of the operation time t as the input of the next iteration;
[0143] S37, determine whether k is greater than or equal to N k , N k is the total number of iterations within the preset running time. If yes, stop the calculation and obtain the updated input of the observation satellite system running time t. Based on the updated input of the observation satellite system running time t, obtain the updated position of the observation satellite system running time t and record it, complete the tracking of the target position at the running time t, and enter the next step; if not, set k=k+1 and return to step S32.
[0144] S38. Determine whether t is greater than or equal to T, where T represents the total operating time of the observation satellite system. If so, stop the calculation to obtain an updated full-error iterative learning control model for the variable tracking task; if not, set t=t+1 and return to step S32.
[0145] Furthermore, the learning gain matrix related parameters include a random noise attenuation factor, a first learning parameter and a second learning parameter; the second learning parameter is used to solve the traceability problem of the observation satellite system;
[0146] The first learning parameter is used to ensure the convergence of the full error iterative learning control model for variable tracking tasks;
[0147] Furthermore, the update position expression of the kth iteration at the running time t is:
[0148]
[0149] in, is the updated position of the kth iteration of the observation satellite system at time t, N k is the path length of the observation satellite system at the kth iteration, N H The path with the longest length is set to the path full tracking state of the observation satellite, that is, the full tracking state y d (t) is the target path of the observation satellite system at operation time t.
[0150] Furthermore, the tracking error expression is:
[0151]
[0152] in, is the tracking error of the observed satellite system at the kth iteration at the running time t.
[0153] Furthermore, the expression of the full error iterative learning control model for the variable tracking task is:
[0154]
[0155] Among them, u k+1 (t) is the observation satellite system input at the k+1th iteration of the running time t, u k (t) is the observation satellite system input at the kth iteration of the running time t, T′ is the time corresponding to the longest path, is the tracking error at the kth iteration at runtime t+1, is the learning gain matrix of the kth iteration of path i corresponding to the observation satellite system at the running time t, i = 1, 2, 3...I, I is the total number of paths, when i = I, it means the longest path N H , a k is the random noise attenuation factor of the observation satellite system at the kth iteration, which is used to overcome the influence of random noise, Γ t is the first learning parameter of the observation satellite at the operating time t, which is used to ensure the convergence of the full error iterative learning control model for variable tracking tasks. t,i It is the second learning parameter of the path length i corresponding to the observation satellite at the operating time t, which is used to overcome the traceability problem of the observation satellite system.
[0156] Furthermore, step S3 also includes learning gain matrix related parameters and learning gain matrix related parameter constraints;
[0157] The learning gain matrix related parameters include a random noise attenuation factor, a first learning parameter, and a second learning parameter;
[0158] The learning gain matrix related parameter constraints include a random noise attenuation factor constraint and a first learning parameter constraint;
[0159] Furthermore, the constraint condition of the random noise attenuation factor is expressed as:
[0160]
[0161] The expression of the first learning parameter constraint condition is:
[0162] Re(λ k (Γ t G t,i ))>0
[0163] Among them, Re(·) is the eigenvalue function, Γ t Gt,i is the algorithm condition matrix, Re(λ k (Γ t G t,i ))>0, indicating that the eigenvalues of the algorithm condition matrix all have positive real parts, λ k is the parameter of the kth iteration,
[0164] G t,i is the matrix determined by the observation satellite system and the second learning parameter,
[0165] A(t) is the first time-varying system matrix of the observation satellite system at the operating time t, C(t) is the second time-varying system matrix of the observation satellite at the operating time t, B(t-1) is the third time-varying system matrix of the observation satellite at the operating time t-1, η∈T, J t,i is the second learning parameter of the path length i corresponding to the observation satellite at the operating time t.
[0166] The present invention also includes step S8, using a full-error iterative learning control model for variable tracking tasks to ensure the precise tracking effect of the observation satellite system.
[0167] Specifically, the specific steps to ensure the precise tracking effect of the observation satellite system include:
[0168] Using Lyapunov function and stochastic approximation method, the convergence of the total error iterative learning control model for variable tracking task is analyzed in the mean square sense and in the probability 1 sense.
[0169] Based on the convergence in the mean square sense and in the probability 1 sense, the stability and accuracy of the full error iterative learning control model for variable tracking tasks under random noise are obtained;
[0170] Based on the stability and accuracy under random noise, the precise tracking effect of the observation satellite system is guaranteed.
[0171] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by any technician familiar with the technical field within the technical scope disclosed by the present invention should be covered within the protection scope of the present invention.
Claims
1. A full error iterative learning control method for variable tracking tasks, characterized in that: include: Step S1, presetting the target position of the observation satellite system at each operating time based on the variable path length model; Step S2, determining the input of the observation satellite system at each operating time and obtaining the random noise at the corresponding time; Step S3, establishing an original full-error iterative learning control model for variable tracking tasks; using the input of the observation satellite system at each operating moment and the random noise at the corresponding moment as a training set; The original full-error iterative learning control model for variable tracking tasks is trained based on the training set to obtain a full-error iterative learning control model for variable tracking tasks; Based on the full-error iterative learning control model for the variable tracking task, the position at each running time is obtained; Step S4, correcting the position of each operating moment based on the target position of the observation satellite at each operating moment to obtain an updated position at each operating moment, and obtaining a tracking error at each operating moment based on the updated position at each operating moment; Step S5, establishing a learning gain matrix; The tracking error at each running moment is updated based on the learning gain matrix to obtain an updated tracking error at each running moment; Step S6, combining the input of each operating time of the observation satellite system with the updated tracking error of each operating time, to obtain the updated input of each operating time of the observation satellite system; Step S7: Based on the updated input of the observation satellite system at each operating time, the updated position of the observation satellite at each operating time is obtained, and the observation satellite tracks the target position at each operating time.
2. The total error iterative learning control method according to claim 1, characterized in that: Step S1 also includes the specific steps of obtaining the variable path length model described in step S1: Obtaining scenarios of observation satellite systems when performing repetitive path tracking missions; The iteratively changing path length in the scene is obtained, and the iteratively changing path length in the scene is set to be tracked by the observation satellite system, so as to obtain the path length of multiple iteratively changing complete tracking states; The path length of the fully tracked state with multiple iterative changes is modeled as a finite-state Markov chain, resulting in a variable path length model.
3. The total error iterative learning control method according to claim 1, characterized in that: The specific steps of obtaining the random noise at the corresponding moment in step S2 include: Acquire random noise generated by the observation satellite system when performing a task; establish a random noise flow domain; and establish a martingale difference sequence of random noise based on the random noise flow domain; The random noise at the corresponding moment is obtained based on the martingale difference sequence of the random noise.
4. The total error iterative learning control method according to claim 1, characterized in that: The specific steps of obtaining the full error iterative learning control model for the variable tracking task in step S3 include: Establish the original full-error iterative learning control model for variable tracking tasks; Determine the learning gain matrix related parameters; obtain the learning gain matrix based on the learning gain matrix related parameters; Obtaining inputs of the observation satellite system at each operating time; obtaining random noises of the observation satellite system at each operating time based on a martingale difference sequence of a random noise watershed; and using the inputs of the observation satellite system at each operating time and the random noise as a training set; The original full-error iterative learning control model for variable tracking tasks is trained based on the training set and the learning gain matrix to obtain a full-error iterative learning control model for variable tracking tasks.
5. The total error iterative learning control method according to claim 1, characterized in that: The update position expression at each running time is: in, is the updated position of the kth iteration of the observation satellite system at time t, N k is the path length of the observation satellite system in the kth iteration, N H is the longest path, y d (t) is the target path of the observation satellite system at operation time t.
6. The total error iterative learning control method according to claim 1, characterized in that: The tracking error expression is: in, is the tracking error of the observed satellite system at the kth iteration at the running time t.
7. The total error iterative learning control method according to claim 1, characterized in that: The expression of the full error iterative learning control model for the variable tracking task is: Among them, u k+1 (t) is the observation satellite system input at the k+1th iteration of the running time t, u k (t) is the observation satellite system input at the kth iteration of the running time t, T′ is the time corresponding to the longest path, is the tracking error at the kth iteration at runtime t+1, is the learning gain matrix of the kth iteration of path i corresponding to the observation satellite system at the running time t, i = 1, 2, 3...I, I is the total number of paths.
8. The total error iterative learning control method according to claim 1, characterized in that: Step S3 also includes learning gain matrix related parameters and learning gain matrix related parameter constraints.
9. The total error iterative learning control method according to claim 8, characterized in that: The learning gain matrix related parameters include a random noise attenuation factor, a first learning parameter and a second learning parameter.
10. The total error iterative learning control method according to claim 8, characterized in that: The learning gain matrix related parameter constraints include a random noise attenuation factor constraint and a first learning parameter constraint.