A Distributed Fusion Method for Sensor Systems Based on Set Theory

Through the distributed fusion method of sensor system based on ensemble theory, the difficulty of solving multi-set intersection in multi-sensor fusion is solved, which improves the fusion performance and provides a higher robust state estimation As a result, it is suitable for complex systems.

CN115114985BActive Publication Date: 2024-11-26ROCKET FORCE UNIV OF ENG
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210767714.X
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-07-01
Publication Date
2024-11-26
Estimated Expiration
2042-07-01

AI Technical Summary

Technical Problem

The existing multi-sensor fusion method has defects in dealing with uncertainty, strong correlation, strong nonlinearity, and high-dimensional problems in complex systems. In particular, the premise of probability assumptions leads to strict noise distribution requirements and is difficult to adapt to multi-set intersections. Solved problem.

Method used

The distributed fusion method of sensor system based on ensemble theory is adopted. By obtaining the system measurement information of each local sensor, the state estimation method is used to estimate the state, the ellipsoid intersecting method is used to fuse, and the parameters of the central estimation ellipsoid are performed. Optimize to obtain the final fusion estimation result.

Benefits of technology

Overcoming the shortcomings of traditional estimation methods, it can better handle uncertainty and strong relevant information, provide a more robust state feasible set description, and is suitable for multi-sensor fusion of complex systems, avoiding dependence on statistical information.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN115114985B_ABST
    Figure CN115114985B_ABST
Patent Text Reader

Abstract

The present invention discloses a distributed fusion method for sensor systems based on set theory. For sensor systems under bounded interference, each local processor first processes the measurement information of the corresponding sensor through a set membership estimation method to obtain a local estimated state feasible set, and then sends the result to a fusion center for fusion processing; according to set operation theory, the fusion between local estimates is achieved by intersecting multiple ellipsoids; in order to deal with the situation where the intersection of multiple ellipsoids may be empty, the concept of q-relaxed intersection is introduced; the algorithm is guaranteed to obtain a robust state feasible set through parameter selection, while reducing the influence of outliers or sensor damage on global estimation. Simulation examples show that the distributed fusion method proposed by the present invention effectively improves the estimation accuracy of the system.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of information fusion, and in particular to a distributed fusion method of a sensor system based on set theory. Background Art

[0002] In recent years, the functional requirements of large and complex systems are increasing rapidly, especially the requirements for high-performance estimation of system states. Due to the obvious defects in measurement accuracy, range, stability and reliability when using a single sensor, multi-sensor systems and related data fusion technologies have received increasing attention in recent years. Estimation fusion algorithms can be divided into centralized fusion and distributed fusion according to their structure. Among them, distributed fusion can reduce channel bandwidth, has strong survivability, and is easy to implement in engineering. Therefore, it has received more attention and has gradually become the focus of research and application in the field of information fusion. Distributed fusion is widely used in civil and defense fields, such as multi-sensor fusion positioning, artificial intelligence, multi-robot collaboration, intelligent connected vehicles, sensor networks, etc.

[0003] At present, estimation fusion is usually solved by estimation methods based on random noise assumptions, such as traditional estimation methods such as Kalman filtering and its related extended algorithms. These methods are difficult to solve new problems such as uncertainty, strong correlation, strong nonlinearity, and high dimensionality brought about by complex systems. The fundamental reason lies in the premise of probabilistic assumptions. These probabilistic methods have strict requirements on the distribution of noise and require its statistical characteristics to be known, which leads to certain defects in such methods.

[0004] In recent years, estimation methods based on set theory have gradually gained attention. This type of method estimates the feasible set of parameters or states through set membership estimation theory. They are based on the assumption of unknown but bounded noise, that is, they only require the noise to be bounded, but do not require the specific distribution of the noise within the boundary. They can overcome the defects of traditional estimation methods and are well applied to situations where traditional estimation methods cannot adapt, and this simple condition is easier to meet.

[0005] However, the application of set theory in multi-sensor fusion still faces great challenges. The fundamental reason is that a new form of estimation result is adopted. Unlike the point estimation method, the set membership estimation method uses the state feasible set as the state estimation result. The representation form and optimization process of the intersection of multiple feasible sets are the difficulties of fusion technology.

[0006] Therefore, it is necessary to provide a new fusion method that uses set theory to overcome the defects of traditional fusion methods, break through the difficulty of solving the intersection of multiple sets, and further improve the performance of multi-sensor fusion. Summary of the invention

[0007] In view of the above-mentioned problems, the present invention aims to provide a distributed fusion method for sensor systems based on set theory, which utilizes set theory to overcome the defects of traditional fusion methods, breaks through the difficulty of solving the intersection of multiple sets, and can improve the performance of multi-sensor fusion.

[0008] In order to achieve the above object, the technical solution adopted by the present invention is as follows:

[0009] A distributed fusion method of a sensor system based on set theory, characterized by comprising the following steps:

[0010] S1: Obtain system measurement information of each local sensor and establish a sensor system model;

[0011] S2: Based on the system measurement information of each local sensor, the set membership estimation method is used to perform state estimation to obtain the local estimation result;

[0012] S3: The local estimation results are fused using the ellipsoid intersection method to obtain the central estimation ellipsoid;

[0013] S4: Optimize the parameters of the central estimation ellipsoid to obtain the final fusion estimation result.

[0014] Furthermore, the specific operation of step S1 includes the following steps:

[0015] S101: Assume that N sensors independently measure the same moving target. is the observation vector of the ith sensor at time k, Represents the m-dimensional vector in real number space; the corresponding state equation and measurement equation are established respectively

[0016] x k =F k-1 x k-1 +G k-1 w k-1 (1)

[0017] z i,k =H i,k x k +v i,k , i = 1, 2, ..., N (2)

[0018] In the formula, represents the state at time k, which is an n-dimensional state vector, x k-1 is the state vector at time k-1; F k-1 is the non-singular state transfer matrix, G k-1 is the process noise input matrix;

[0019] is the k-1 moment process noise, Represents the l-dimensional model of real number space; H i,k is the observation matrix of the i-th sensor at time k, is the observation noise of the i-th sensor at time k;

[0020] S102: Assume that the model process noise and observation noise belong to the ellipsoid set

[0021]

[0022]

[0023] In the formula, Q k and R i,k is a known positive definite matrix, Q k represents the k-time process noise ellipsoid shape matrix, R i,k represents the measurement noise ellipsoid matrix of the i-th sensor at time k;

[0024] Then the initial value of the system state to be estimated belongs to the ellipsoid described by the following formula

[0025]

[0026] Where x is an arbitrary point in the ellipsoid and is an m-dimensional vector. is the center of the ellipsoid, also an m-dimensional vector, P 0 is a positive definite matrix that defines the shape of the ellipsoid, σ 0 is a real number greater than 0, and the subscript 0 indicates the initial time.

[0027] Furthermore, the state estimation using the set membership estimation method in step S2 is divided into a state update phase and a measurement update phase.

[0028] Furthermore, the specific operations in the status update phase include the following steps:

[0029] In the state update phase, update the ellipsoid

[0030]

[0031] The calculation method is:

[0032]

[0033]

[0034] σ k|k-1 =σ k-1 ;

[0035] Among them, P k-1 is the ellipsoid shape matrix at time k-1, Q k-1is the process noise shape matrix at time k-1, σ k-1 As the upper bound of the system's Lyapunov function, and P k-1 Together they determine the shape of the state ellipsoid at time k-1, p k ∈(0, +∞) is used to optimize the ellipsoid ε k|k-1 The size of p k The optimal value is Here, tr represents the trace of the matrix.

[0036] Furthermore, the specific operations of the measurement update phase include the following steps:

[0037] During the measurement update phase, the ellipsoid is updated

[0038]

[0039] The calculation method is:

[0040] P i,k =(I n -K i,k H i,k ) k|k-1 ;

[0041]

[0042]

[0043] Among them, I n is the n-dimensional unit matrix;

[0044] Filter gain

[0045] Residual α i,k ∈[0, +∞) represents the parameters of the i-th sensor at time k, which is used to optimize the ellipsoid ε i,k The size of is the optimal value of which is the solution of the following formula

[0046]

[0047] when When , equation (6) has no solution, and 0 is taken as the optimal value of the parameter.

[0048] Furthermore, the specific operation of step S3 includes the following steps:

[0049] S301: Assume that the center estimated ellipsoid is

[0050]

[0051] For the ellipsoid to contain the intersection of all local estimated ellipsoids, the elements in it must satisfy

[0052]

[0053] Where β i,k ∈[0, 1], and

[0054] S302: Transform the above formula (8) to obtain

[0055]

[0056] According to formula (9), we can take

[0057]

[0058]

[0059] Then there is

[0060]

[0061] S303: Substituting equations (10) and (12) into equation (9), we can obtain

[0062]

[0063] S304: Substituting equations (10), (11) and (13) into equation (7), the center estimated ellipsoid can be obtained.

[0064] Furthermore, the specific operation of step S4 includes the following steps:

[0065] S401: Use the minimum bounding box of the center estimation ellipsoid to calculate the relaxed intersection center Used to replace the true intersection center to calculate β i,k , the minimum enclosing box of the center estimated ellipsoid is represented by the interval vector:

[0066] In the formula, the superscript j represents the jth state variable, the subscripts + and - represent the maximum and minimum values, respectively, and P i,j represents the element in the i-th row and j-th column of the matrix P;

[0067] S402: Get the relaxed intersection center Afterwards, β i,k Calculate according to the following formula

[0068]

[0069] in,

[0070]

[0071] The norm in the formula is defined as

[0072] Furthermore, in step S401, the q-relaxed intersection method is used to calculate The specific steps include:

[0073] S4011: Project the vertices of each box onto the coordinate axis;

[0074] S4012: The points projected on the coordinate axis can form (2λ-1) in the plane 2 sub-boxes, where λ is the number of boxes;

[0075] S4013: Calculate the center point of each sub-box;

[0076] S4014: Check whether each center point belongs to each initial box, and find the center points belonging to λ-q boxes The subbox containing this center point is the q-relaxed intersection that needs to be solved; at the same time, the initial box that does not contain the q-relaxed intersection is regarded as an outlier; in this process, first set q = 0, and then gradually increase it until a center point that meets the conditions is found.

[0077] The beneficial effects of the present invention are:

[0078] 1. The distributed fusion method of the sensor system based on set theory in the present invention can overcome the defects of the traditional estimation method, and is well applied to the occasions that the traditional estimation method cannot adapt to, and this simple condition is easier to be satisfied.

[0079] 2. The sensor system distributed fusion method based on set theory in the present invention can include uncertainty in the set for calculation, and then obtain a deterministic description of the state feasible set and the boundary. This makes the method more robust than traditional estimation methods.

[0080] 3. The sensor system distributed fusion method based on set theory in the present invention can conveniently handle the fusion estimation problem with strong correlation information. Because it only needs to know the error boundary information and does not need to know the statistical information including the mutual covariance, the operation process does not involve the role of correlation. Therefore, for multi-sensor fusion of complex systems, set membership estimation is more advantageous than probabilistic estimation. BRIEF DESCRIPTION OF THE DRAWINGS

[0081] Figure 1 is a process noise distribution diagram in an embodiment of the present invention;

[0082] Figure 2 is a measurement noise distribution diagram of the sensor 1 in the embodiment of the present invention;

[0083] Figure 3 is a measurement noise distribution diagram of the sensor 2 in an embodiment of the present invention;

[0084] Figure 4 is the average mean square error of state 1 in a simulation in an embodiment of the present invention;

[0085] Figure 5 is the average mean square error of state 2 in a simulation in an embodiment of the present invention;

[0086] Figure 6 It is the mean mean square error of state 3 in a simulation in an embodiment of the present invention. DETAILED DESCRIPTION

[0087] In order to enable those skilled in the art to better understand the technical solution of the present invention, the technical solution of the present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0088] A distributed fusion method of sensor system based on set theory includes the following steps:

[0089] S1: Obtain system measurement information of each local sensor and establish a sensor system model;

[0090] Specifically, S101: Consider a general linear discrete system, assuming that N sensors independently measure the same moving target. is the observation vector of the ith sensor at time k, Represents the m-dimensional vector in real number space; the corresponding state equation and measurement equation are established respectively

[0091] x k =F k-1 x k-1 +G k-1 w k-1 (1)

[0092] z i,k =H i,k x k +v i,k , i = 1, 2, ..., N (2)

[0093] In the formula, represents the state at time k, which is an n-dimensional state vector, x k-1 is the state vector at time k-1; F k-1 is the non-singular state transfer matrix, G k-1 is the process noise input matrix; is the k-1 moment process noise, Represents the l-dimensional model of real number space; H i,k is the observation matrix of the i-th sensor at time k, is the observation noise of the i-th sensor at time k;

[0094] S102: Assume that the model process noise and observation noise belong to the ellipsoid set

[0095]

[0096]

[0097] In the formula, Q k and R i,k is a known positive definite matrix, Q k represents the k-time process noise ellipsoid shape matrix, R i,k represents the measurement noise ellipsoid matrix of the i-th sensor at time k;

[0098] Then the initial value of the system state to be estimated belongs to the ellipsoid described by the following formula

[0099]

[0100] Where x is an arbitrary point in the ellipsoid and is an m-dimensional vector. is the center of the ellipsoid, also an m-dimensional vector, P 0 is a positive definite matrix that defines the shape of the ellipsoid, σ 0 is a real number greater than 0, and the subscript 0 indicates the initial time.

[0101] Further, S2: based on the system measurement information of each local sensor, a set membership estimation method is used to perform state estimation to obtain a local estimation result;

[0102] Specifically, state estimation using the set membership estimation method is divided into a state update phase and a measurement update phase.

[0103] In the state update phase, update the ellipsoid

[0104]

[0105] The calculation method is:

[0106]

[0107]

[0108] σ k|k-1 =σ k-1 ;

[0109] Among them, P k-1 is the ellipsoid shape matrix at time k-1, Q k-1 is the process noise shape matrix at time k-1, σ k-1As the upper bound of the system's Lyapunov function, and P k-1 Together they determine the shape of the state ellipsoid at time k-1, p k ∈(0, +∞) is used to optimize the ellipsoid ε k|k-1 The size of p k The optimal value is Here, tr represents the trace of the matrix.

[0110] During the measurement update phase, the ellipsoid is updated

[0111]

[0112] The calculation method is:

[0113] P i,k =(I n -K i,k H i,k ) k|k-1 ;

[0114]

[0115]

[0116] Among them, I n is the n-dimensional unit matrix;

[0117] Filter gain

[0118] Residual α i,k ∈[0, +∞) represents the parameters of the i-th sensor at time k, which is used to optimize the ellipsoid ε i,k The size of is the optimal value of which is the solution of the following formula

[0119]

[0120] when When , equation (6) has no solution, and 0 is taken as the optimal value of the parameter.

[0121] Further, S3: each local estimation result is fused using the ellipsoid intersection method to obtain a central estimation ellipsoid;

[0122] Specifically, S301: Assume that the center estimated ellipsoid is

[0123]

[0124] For the ellipsoid to contain the intersection of all local estimated ellipsoids, the elements in it must satisfy

[0125]

[0126] Where βi,k ∈[0, 1], and

[0127] S302: Transform the above formula (8) to obtain

[0128]

[0129] According to formula (9), we can take

[0130]

[0131]

[0132] Then there is

[0133]

[0134] S303: Substituting equations (10) and (12) into equation (9), we can obtain

[0135]

[0136] S304: Substituting equations (10), (11) and (13) into equation (7), the center estimated ellipsoid can be obtained.

[0137] Further, S4: Parameters β of the estimated ellipsoid at the center i,k Optimize and obtain the final fusion estimation result.

[0138] Specifically, the relaxed intersection center is calculated using the minimum bounding box of the center estimation ellipsoid Used to replace the true intersection center to calculate β i,k , the minimum enclosing box of the center estimated ellipsoid is represented by the interval vector:

[0139] In the formula, the superscript j represents the jth state variable, the subscripts + and - represent the maximum and minimum values, respectively, and P i,j represents the element in the i-th row and j-th column of the matrix P;

[0140] S402: Get the relaxed intersection center Afterwards, β i,k Calculate according to the following formula

[0141]

[0142] in,

[0143]

[0144] The norm in the formula is defined as

[0145] In step S401, the q-relaxed intersection and Calculation method:

[0146] S4011: Project the vertices of each box onto the coordinate axis;

[0147] S4012: The points projected on the coordinate axis can form (2λ-1) in the plane 2 = 25 sub-boxes, where λ is the number of boxes;

[0148] S4013: Calculate the center point of each sub-box;

[0149] S4014: Check whether each center point belongs to each initial box, find the center point λ belonging to mq boxes, and the subbox containing this center point is the q-relaxed intersection that needs to be solved; at the same time, the initial box that does not contain the q-relaxed intersection is regarded as an outlier; in this process, first set q=0, and then gradually increase it until a center point that meets the conditions is found.

[0150] Example:

[0151] Consider a target tracking system with three sensors. The sensor system is given by equations (1) and (2), and the corresponding matrix is

[0152]

[0153] H 3,k =[0 0 1],

[0154] Among them, T 0 = 0.1 is the sampling period, and the state vector at time k is x k =[x 1,k x 2,k x 3,k ] T , x 1,k 、x 2,k and x 3,k Respectively refer to the position, velocity and acceleration of the maneuvering target. i,k is the detection signal of the i-th sensor to the target. The initial state ellipsoid conforms to formula (5), and the corresponding parameter is P 0 =I 3 , σ 0 =1, the actual initial state value in the simulation is [0 0 0] T The matrix that determines the noise boundary ellipsoid is Q k =diag(10,10,10),R1,k =diag(0.2,0.2),R 2,k =diag(0.8,0.6),R 3,k =0.7. In the simulation, the process noise and the measurement noise of sensor 1 and sensor 2 are uniformly distributed on the ellipsoid ε(0, Q k ),ε(0,R 1,k ),ε(0,R 2,k ), the measurement noise of sensor 3 is evenly distributed in the interval In the simulation, the process noise and observation noise are uniformly distributed inside the ellipsoid, as shown in the attached figure. Figure 1-3 As shown, Figure 1 is the process noise distribution diagram; Figure 2 is the measurement noise distribution diagram of the sensor 1; Figure 3 is the measurement noise distribution diagram of sensor 2. Then, the distributed fusion method based on set theory proposed in the present invention is used to fuse the sensor information to obtain the fusion estimation result. In the simulation, the set membership estimation fusion algorithm proposed by Becis-Aubry is used as a comparison.

[0155] The simulation was run 100 times in Matlab R2014a on an Intel Core i5 PC (3.2GHz, 4G RAM), and each simulation contained 1000 samples. The mean mean square error (MSE) of each state variable of the global state estimation and the average volume of the estimated ellipsoid were used as evaluation indicators, as shown in Table 1.

[0156] Table 1 Statistical comparison of estimation results

[0157]

[0158] Table Refers to the MSE of states i=1, 2, 3.

[0159] It should be noted that the MSE and volume are averaged from step 100 to step 1000 to exclude the effect of the initial phase, and the center of the ellipsoid is considered as a point estimate in the simulation.

[0160] Attached Figure 4 To the attached Figure 6 The MSE in a simulation changes over time to show the results more intuitively. Figure 4 is the mean square error of state 1, Figure 5 is the mean square error of state 2, Figure 6 is the mean square error of state 3. From Table 1 and the attached Figure 4-6 The results in show that the MSE obtained by the fusion method proposed in the present invention is smaller than that of the comparison algorithm, which indicates that the fusion method proposed in the present invention is superior to the comparison algorithm in terms of point estimation accuracy.

[0161] The above shows and describes the basic principles, main features and advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments, and the above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A distributed fusion method for sensor systems based on set theory, characterized in that: The following steps are included: S1: Obtain system measurement information of each local sensor and establish a sensor system model; S2: Based on the system measurement information of each local sensor, the set membership estimation method is used to perform state estimation to obtain the local estimation result; S3: The local estimation results are fused using the ellipsoid intersection method to obtain the central estimation ellipsoid; S4: Optimize the parameters of the central estimation ellipsoid to obtain the final fusion estimation result; The specific operation of step S1 includes the following steps: S101: Assume that N sensors independently measure the same moving target. is the observation vector of the ith sensor at time k, Represents the m-dimensional vector in real number space; the corresponding state equation and measurement equation are established respectively x k =F k-1 x k-1 +G k-1 w k-1 (1) z i,k =H i,k x k +v i,k ,i=1,2,…,N (2) In the formula, represents the state at time k, which is an n-dimensional state vector, x k-1 is the state vector at time k-1; F k-1 is the non-singular state transfer matrix, G k-1 is the process noise input matrix; is the k-1 moment process noise, Represents the l-dimensional model of real number space; H i,k is the observation matrix of the i-th sensor at time k, is the observation noise of the i-th sensor at time k; S102: Assume that the model process noise and observation noise belong to the ellipsoid set In the formula, Q k and R i,k is a known positive definite matrix, Q k represents the k-time process noise ellipsoid shape matrix, R i,k represents the measurement noise ellipsoid matrix of the i-th sensor at time k; Then the initial value of the system state to be estimated belongs to the ellipsoid described by the following formula Where x is an arbitrary point in the ellipsoid and is an m-dimensional vector. is the center of the ellipsoid, also an m-dimensional vector, P0 is a positive definite matrix, which defines the shape of the ellipsoid, σ0 is a real number greater than 0, and the subscript 0 indicates the initial moment; In step S2, the state estimation using the set membership estimation method is divided into a state update phase and a measurement update phase; The specific operations in the status update phase include the following steps: In the state update phase, update the ellipsoid The calculation method is: s k|k-1 =s k-1 ; Among them, P k-1 is the ellipsoid shape matrix at time k-1, Q k-1 is the process noise shape matrix at time k-1, σ k-1 As the upper bound of the system's Lyapunov function, and P k-1 Together they determine the shape of the state ellipsoid at time k-1, p k ∈(0, +∞) is used to optimize the ellipsoid ε k|k-1 The size of p k The optimal value is Where tr represents the trace of the matrix; The specific operations of the measurement update phase include the following steps: During the measurement update phase, the ellipsoid is updated The calculation method is: P i,k =(I n -K i,k H i,k )P k|k -1; Among them, I n is the n-dimensional unit matrix; Filter gain Residual α i,k ∈[0, +∞) represents the parameters of the i-th sensor at time k, which is used to optimize the ellipsoid ε i,k The optimal value is the solution of the following formula when When , equation (6) has no solution, and 0 is taken as the optimal value of the parameter; The specific operation of step S3 includes the following steps: S301: Assume that the center estimated ellipsoid is For the ellipsoid to contain the intersection of all local estimated ellipsoids, the elements in it must satisfy Where β i,k ∈[0, 1], and S302: Transform the above formula (8) to obtain According to formula (9), take Then there is S303: Substituting equations (10) and (12) into equation (9), we can obtain S304: Substituting equations (10), (11) and (13) into equation (7), the center estimated ellipsoid can be obtained.

2. A sensor system distributed fusion method based on set theory according to claim 1, characterized in that: The specific operation of step S4 includes the following steps: S401: Use the minimum bounding box of the center estimation ellipsoid to calculate the relaxed intersection center Used to replace the true intersection center to calculate β i,k , the minimum enclosing box of the center estimated ellipsoid is represented by the interval vector: In the formula, the superscript j represents the jth state variable, the subscripts + and - represent the maximum and minimum values ​​respectively, and P i,j represents the element in the i-th row and j-th column of the matrix P; S402: Get the relaxed intersection center Afterwards, β i,k Calculate according to the following formula in, The norm in the formula is defined as 3. The sensor system distributed fusion method based on set theory according to claim 1, characterized in that: In step S401, the q-relaxed intersection method is used to calculate The specific steps include: S4011: Project the vertices of each box onto the coordinate axis; S4012: The points projected on the coordinate axis can form (2λ-1) in the plane 2 sub-boxes, where λ is the number of boxes; S4013: Calculate the center point of each sub-box; S4014: Check whether each center point belongs to each initial box, and find the center point belonging to λ-q boxes The subbox containing this center point is the q-relaxed intersection that needs to be solved; at the same time, the initial box that does not contain the q-relaxed intersection is regarded as an outlier; in this process, first set q = 0, and then gradually increase it until a center point that meets the conditions is found.