State estimation method based on block Cholesky decomposition cross direction multiplier method
By using the cross-direction multiplier method of blocked Cholesky decomposition in state estimation, the problems of degradation in performance and low computational efficiency of traditional Bayesian filtering and Lagrangian methods in constrained state estimation are solved, and higher accuracy is achieved and faster state estimates.
Patent Information
- Application Number
- CN202411048412.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-01
- Publication Date
- 2024-11-15
AI Technical Summary
The traditional Bayesian filtering method has deteriorated performance when processing constraint state estimation, especially in big data situations, and it is difficult to achieve real-time performance. Although the Lagrangian method is effective, iterative solution multiple times leads to low computational efficiency.
The cross-direction multipliers method based on Lagrangians is used, and the large-scale matrix operation is converted into multiple sub-matrix solutions through blocked Cholesky decomposition during each iteration, and the upper and lower triangle matrix characteristics are simplified into sequential iteration. equation.
Improved accuracy and calculation speed of state estimation, especially in dealing with long-term step problems, achieving higher accuracy and faster calculation speed.
Smart Images

Figure CN118964822A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of data analysis, and more specifically, to a state estimation method based on a block Cholesky decomposition cross-direction multiplier method. Background Art
[0002] Constrained state estimation is the process of estimating the state variables of a dynamic system by taking into account constraints that must be satisfied during the estimation process. These constraints involve using indirect measurements or observations to estimate the current or future state of the system, and are often applied in control systems, robotics, signal processing, and autonomous systems.
[0003] Most of the traditional methods used to solve the state estimation problem are based on Bayesian filtering. These methods obtain the posterior probability distribution of the system state by fusing prior information and measurement data. Specifically, common algorithms include Kalman filtering, particle filtering, extended Kalman filtering, etc. However, the Bayesian filtering method relies on the mathematical modeling of the state equation and the observation equation, and integrating constraints into the model may lead to a degradation of the filter performance. For some problem domains, it is feasible to directly apply Bayesian filtering and smoothing to the optimization objective function. This approach involves formulating the joint probability distribution of the model and then maximizing the posterior probability. Although conceptually simple, the actual implementation may encounter challenges in solving high-dimensional optimization problems and computationally intensive tasks.
[0004] The Lagrangian method is an effective method for dealing with constrained state estimation problems. It uses Lagrangian multipliers to introduce constraints and transforms the original problem into an equivalent optimization problem. In particular, the augmented Lagrangian method and the cross-direction multiplier method, by constructing an augmented Lagrangian function containing constraints and using an optimization algorithm to solve the function, can provide more accurate and reliable state estimation results, especially in practical applications with complex constraints. However, this method involves multiple iterative solutions, so it is difficult to meet real-time requirements when facing big data.
[0005] In view of this, the present invention provides a state estimation method based on block Cholesky decomposition cross-direction multiplier method to solve the above problem. Summary of the invention
[0006] In order to overcome the problems in the prior art, the present invention proposes a state estimation method based on a block Cholesky decomposition cross-direction multiplier method. The cross-direction multiplier method based on Lagrangian is adopted to achieve higher accuracy. At the same time, in each Lagrangian iteration process, a block Cholesky decomposition is proposed, and large-scale matrix operations are converted into multiple sub-matrices in the form of blocks for solution. The upper triangular matrix and the lower triangular matrix form characteristics after Cholesky decomposition are used to simplify the problem into two sets of sequential iterative equations, which can achieve faster calculation speed when dealing with long time step problems.
[0007] In a first aspect, the present invention provides a state estimation method based on a block Cholesky decomposition cross-direction multiplier method, comprising the following steps:
[0008] S1, target state data corresponding to the unmanned ship state estimation;
[0009] S2, abstract the complex ocean environment into constraint equations based on Gaussian affine and establish a state estimation model under constraint conditions;
[0010] S3, constructing the Lagrangian equation and solving the Lagrangian equation iteratively;
[0011] S4, proposed a cross-direction multiplier method based on block Cholesky decomposition;
[0012] S5. Decompose the large Lagrange equation matrix into multiple small matrices, and convert the matrix calculation into a sequential solution to obtain the optimal state estimation value corresponding to the unmanned ship.
[0013] As a preferred technical solution of the present invention, the target state data includes position, heading, speed and angular velocity; the target state data is obtained through measurement information of the sensor and prior information of the dynamic system.
[0014] As a preferred technical solution of the present invention, the state estimation model is:
[0015] x t =A t x t-1 +b t +q t
[0016] y t =H t x t +e t +r t
[0017] The constraint equation is:
[0018] C t x t +dt ≤0
[0019] Among them: A t , H t and C t are the state transfer matrix, observation matrix and constraint matrix of the state estimation model respectively; b t 、e t , and d t are the deviation terms corresponding to the state transfer matrix, observation matrix and constraint matrix respectively; the initial state x 1 Assume that it follows a Gaussian distribution with mean m 1 and covariance P 1 .
[0020] As a preferred technical solution of the present invention, it is characterized in that the Lagrange iterative sub-equation:
[0021]
[0022]
[0023]
[0024] In the formula is the barrier variable for constraint processing, is the Langrange multiplier, ρ is the penalty parameter, and l is the number of iterations of the sub-equation.
[0025] As a preferred technical solution of the present invention, the large matrix in the Lagrangian equation of step S3 is decomposed into multiple small matrices, and iteratively solved by the cross-direction multiplication method to obtain the state estimation matrix corresponding to the state estimation of the unmanned ship.
[0026] As a preferred technical solution of the present invention, a block Cholesky decomposition method is used to divide the matrix into T sub-matrices and perform block sequential solution;
[0027] By defining the upper triangular matrix and vector, the form of the corresponding solution can be obtained, and the updates are performed in sequence based on the form of the corresponding solution, and the sub-matrices are solved in sequence.
[0028] As a preferred technical solution of the present invention, the matrix calculation is converted into a sequential solution, and the linear equations of T sub-matrices are operated to obtain the state estimation value.
[0029] In a second aspect, the present invention further provides a state estimation device based on block Cholesky decomposition cross-direction multiplier method, which is a device for executing the method described in the first aspect above.
[0030] In a third aspect, the present invention further provides an electronic device comprising: a processor and a memory, wherein the memory stores a computer program that can be called by the processor;
[0031] The processor executes the method described in the first aspect by calling the computer program stored in the memory.
[0032] In a fourth aspect, the present invention provides a computer-readable storage medium storing instructions, which, when executed on a computer, enable the computer to execute the method described in the first aspect.
[0033] The present invention provides a state estimation method based on a block Cholesky decomposition cross-direction multiplier method. The cross-direction multiplier method based on Lagrangian can achieve higher accuracy. At the same time, in each Lagrangian iteration process, a block Cholesky decomposition is proposed to convert large-scale matrix operations into multiple sub-matrices in a block form for solution, and the upper triangular matrix and lower triangular matrix form characteristics after Cholesky decomposition are used to simplify the problem into two sets of sequential iterative equations. When dealing with long time step problems, the calculation speed is faster. The specific advantages of the present invention are as follows:
[0034] 1. High calculation accuracy: By using the cross-direction multiplier method to deal with the state estimation problem, higher accuracy can be achieved compared to the Bayesian filtering-based method, especially for handling constraint problems.
[0035] 2. Fast calculation speed: Large-scale matrix operations are divided into blocks and converted into multiple sub-matrices for solution. The characteristics of the upper triangular matrix and lower triangular matrix after Cholesky decomposition are used to simplify the problem into two sets of sequential iterative equations, which can achieve faster calculation speed when dealing with long time step problems.
[0036] 3. Wide application: With the development of artificial intelligence, state estimation plays a key role in the fields of unmanned ships and unmanned vehicles. This method has strong generalization and can effectively deal with practical problems. BRIEF DESCRIPTION OF THE DRAWINGS
[0037] Figure 1 It is a schematic diagram of the process of the present invention;
[0038] Figure 2 This is a comparison chart of the actual data of the unmanned ship and the trajectory prediction results. DETAILED DESCRIPTION
[0039] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0040] Example 1
[0041] See also Figure 1 This embodiment provides a state estimation method based on block Cholesky decomposition cross-direction multiplier method, and the specific method is as follows:
[0042] S1. Obtaining target state data corresponding to the state estimation of the unmanned ship based on the measurement information of the sensor and the prior information of the dynamic system;
[0043] It should be noted that the state estimation problem of unmanned ships refers to the process of continuously and accurately restoring target state data, such as position, heading, speed, angular velocity, etc., by using the measurement information of sensors and the prior information of dynamic systems, using certain optimal criteria and modern computing methods. This ensures that the unmanned ship can accurately perceive its own state in various complex environments, providing a reliable basis for subsequent navigation, control, decision-making, etc.
[0044] S2, abstract the complex ocean environment into constraint equations based on Gaussian affine and establish a state estimation model under constraint conditions;
[0045] It should be noted that the Gaussian affine is to apply affine transformation under the Gaussian distribution framework to abstract the constraints of the ocean environment and establish a corresponding state estimation model.
[0046] More specifically, based on Gaussian affine, the complex ocean constraints are abstracted into corresponding mathematical models, and the established state estimation model is:
[0047] x t =A t x t-1 +b t +q t ;
[0048] y t =H t x t +e t +r t ;
[0049] The constraint equation is:
[0050] C t x t +d t ≤0
[0051] Among them: A t , H t and C t are the state transfer matrix, observation matrix and constraint matrix of the state estimation model respectively; b t 、e t , and d t are the deviation terms corresponding to the state transfer matrix, observation matrix and constraint matrix respectively; the initial state x 1 Assume that it follows a Gaussian distribution with mean m 1 and covariance P 1 .
[0052] Error term q t and r t are assumed to be independent zero-mean Gaussian random variables with a known positive definite covariance matrix Q t and R t .
[0053] S3, constructing the Lagrangian equation, iteratively solving the dynamic behavior and constraint conditions in step S2 based on the Lagrangian equation, and obtaining the state estimation value corresponding to the state estimation of the unmanned ship;
[0054] It should be noted that the Lagrange equation is used to describe the dynamic behavior and constraints of the system; the state estimation problem of the unmanned ship is transformed into a problem of solving the Lagrange equation. Since the solution process is usually iterative, the objective function is optimized by continuously adjusting the state estimation value of the unmanned ship.
[0055] Specifically: the Lagrange iterator equation can be obtained:
[0056]
[0057]
[0058]
[0059] in: is the barrier variable for constraint processing, is the Langrange multiplier, ρ is the penalty parameter, and l is the number of iterations of the sub-equation. Next, the gradient of the above equation is calculated to be 0 to obtain the corresponding solution.
[0060] S4. A cross-direction multiplier method based on block Cholesky decomposition is proposed to decompose the large matrix in the Lagrangian equation in step S3 into multiple small matrices, and iteratively solve them through the cross-direction multiplier method to obtain the state estimation matrix corresponding to the state estimation of the unmanned ship;
[0061] It should be noted that block Cholesky decomposition is a method of decomposing a large matrix into multiple small matrices, which is suitable for processing large-scale, sparse or symmetric positive definite matrices. By processing in blocks, we can reduce the computational complexity and improve the computational efficiency.
[0062] The cross-directional multiplier method is an iterative algorithm for solving optimization problems, which is suitable for distributed computing and large-scale data sets. The original problem is decomposed into multiple sub-problems, and the solution of the original problem is approximated by alternately solving these sub-problems and updating the multipliers.
[0063] Specifically: Solve When faced with big data, it involves the calculation of multi-dimensional matrices. Therefore, the block Cholesky decomposition method is used to divide it into T sub-matrices and solve them in block order. By defining the upper triangular matrix L and the vector ω, the corresponding solution can be obtained in the form of:
[0064] LL T x (l+1) =ω;
[0065] Where: x (l+1) represents the state information obtained after (l+1) iterations; L T represents the lower triangular matrix after the decomposition of the state-space equation; L represents the lower triangular matrix after the decomposition of the state-space equation.
[0066] Definition D t and E t They are the diagonal submatrices and sub-diagonal submatrices in the L matrix, respectively. 1 →E 2 →D 2 →E 3 →…→D T The order of updates is as follows:
[0067]
[0068] Chol represents the Cholesky decomposition operation. After updating according to the above steps, the upper triangular matrix L can be obtained. Next, the lower triangular equation Ly is solved by the previous generation method. (l+1) =ω, use the descendant method to solve the upper triangular equation L T x (l+1) =y. Based on the special form of matrix L, the following recursive equation can be derived:
[0069]
[0070]
[0071] By utilizing the formal characteristics of the upper triangular matrix and the lower triangular matrix after Cholesky decomposition, the problem is simplified into two sets of sequential iterative equations. When dealing with long time step problems, large matrix operations are avoided and the calculation speed is faster.
[0072] S5. Decompose the large Lagrange equation matrix into multiple small matrices, and convert the matrix calculation into a sequential solution to obtain the optimal state estimation value corresponding to the unmanned ship.
[0073] It should be noted that: by converting matrix calculations into sequential solutions, large-scale matrix calculations are converted into linear equation operations with T sub-matrices, so that the solution of y can be obtained. Similarly, by calculating the linear equations of the upper triangular system, the value of x can be obtained.
[0074]
[0075] At this point, the large-scale matrix operations in the iterative process of the Lagrange day problem have been converted into a solution method with two iterative calculations.
[0076] Example 2
[0077] The parts not described in detail in this embodiment are as shown in Embodiment 1. This embodiment provides a specific example of a state estimation method based on a block Cholesky decomposition cross-direction multiplier method.
[0078] When an unmanned ship is sailing in the ocean, its position and speed are estimated in real time. The ocean environment is complex and changeable, including factors such as waves, ocean currents, and wind direction. These factors are quantified through Gaussian distribution or other statistical models, and the corresponding constraint equations are constructed.
[0079] The Lagrangian equation is constructed to describe the dynamic behavior and constraints of the unmanned ship. The large matrix in the Lagrangian equation is decomposed into multiple small matrices using block Cholesky decomposition, and the cross-direction multiplier method is used to iteratively solve it.
[0080] In the iterative process, we solve the sub-problems corresponding to each small matrix in a certain order and update the multiplier to approximate the solution to the original problem. In this way, we can efficiently estimate the state information of the unmanned ship and adjust its navigation trajectory in real time to cope with the complex changes in the marine environment.
[0081] Example 3
[0082] The part not described in detail in this embodiment is as shown in Embodiment 1. This embodiment provides an electronic device including: a processor and a memory, wherein the memory stores a computer program that can be called by the processor;
[0083] The processor executes the above-mentioned state estimation method based on block Cholesky decomposition cross-direction multiplier method by calling the computer program stored in the memory. Currently, it has been tested on unmanned ships. Figure 2 To make trajectory prediction results based on the real data of the unmanned ship, it can be seen that the prediction results are almost consistent with the actual track trajectory. Therefore, this method has a high practical application value.
[0084] Example 4
[0085] The part not described in detail in this embodiment is as shown in Example 1. This embodiment provides a computer-readable storage medium storing instructions. When the instructions are executed on a computer, the computer executes the above-mentioned state estimation method based on the block Cholesky decomposition cross-direction multiplier method.
[0086] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.
[0087] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.
Claims
1. A state estimation method based on block Cholesky decomposition cross-direction multiplier method, characterized in that: The following steps are involved: S1, target state data corresponding to the unmanned ship state estimation; S2, abstract the complex ocean environment into constraint equations based on Gaussian affine and establish a state estimation model under constraint conditions; S3, constructing the Lagrangian equation and solving the Lagrangian equation iteratively; S4, proposed a cross-direction multiplier method based on block Cholesky decomposition; S5. Decompose the large Lagrange equation matrix into multiple small matrices, and convert the matrix calculation into a sequential solution to obtain the optimal state estimation value corresponding to the unmanned ship.
2. The state estimation method based on block Cholesky decomposition cross-direction multiplier method according to claim 1, characterized in that: The target state data includes position, heading, speed and angular velocity; The target state data is obtained through measurement information of the sensor and prior information of the dynamic system.
3. The state estimation method based on block Cholesky decomposition cross-direction multiplier method according to claim 1, characterized in that: The state estimation model is: x t =A t x t-1 +b t +q t y t =H t x t +e t +r t The constraint equation is: C t x t +d t ≤0 Among them: A t , H t and C t are the state transfer matrix, observation matrix and constraint matrix of the state estimation model respectively; b t 、e t , and d t They are the deviation terms corresponding to the state transfer matrix, observation matrix and constraint matrix respectively; the initial state x1 is assumed to obey a Gaussian distribution with mean m1 and covariance P1.
4. The state estimation method based on block Cholesky decomposition cross-direction multiplier method according to claim 3 is characterized in that: Lagrange iterator equation: In the formula is the barrier variable for constraint processing, is the Langrange multiplier, ρ is the penalty parameter, and l is the number of iterations of the sub-equation.
5. According to the state estimation method based on block Cholesky decomposition and cross-direction multiplier method described in claim 4, the large matrix in the Lagrangian equation of step S3 is decomposed into multiple small matrices, and the cross-direction multiplier method is used to iteratively solve the state estimation matrix corresponding to the state estimation of the unmanned ship.
6. According to the state estimation method based on block Cholesky decomposition cross-direction multiplier method in claim 5, the block Cholesky decomposition method is used to divide it into T sub-matrices and perform block sequential solution; By defining the upper triangular matrix and vector, the form of the corresponding solution can be obtained, and the updates are performed in sequence based on the form of the corresponding solution, and the sub-matrices are solved in sequence.
7. According to the state estimation method based on block Cholesky decomposition cross-direction multiplier method described in claim 6, matrix calculation is converted into sequential solution, and linear equations of T sub-matrices are operated to obtain state estimation values.
8. A state estimation device based on block Cholesky decomposition cross-direction multiplier method, characterized in that: A device for executing a state estimation method based on a block Cholesky decomposition cross-direction multiplier method as described in any one of claims 1 to 7.
9. An electronic device comprising: include: A processor and a memory, wherein the memory stores a computer program that can be called by the processor; The processor executes the state estimation method based on block Cholesky decomposition cross-direction multiplier method described in any one of claims 1 to 7 by calling the computer program stored in the memory.
10. A computer-readable storage medium storing instructions, which, when executed on a computer, enable the computer to execute the state estimation method based on block Cholesky decomposition cross-direction multiplier method as described in any one of claims 1 to 7.
Citation Information
Cited By
Gas film cooling measuring point layout optimization method and system based on expected information gain
CN122389659A
