Underground space unmanned system positioning method and system under multi-source interference

Abstract

This invention discloses a method and system for locating unmanned systems in underground space under multi-source interference, relating to the field of unmanned system positioning technology. The invention first collects various motion data of the unmanned system at different acquisition times and constructs a system model. Then, within a statistical similarity framework, a cost function is defined, and the mean of the state vector and the error covariance matrix at each acquisition time are iterated. Next, variational Bayesian learning is used to jointly infer the noise covariance matrix within a sliding window. Finally, the position estimation result and error covariance matrix of the unmanned system are output. This invention solves the problem of poor integrated navigation and positioning performance of unmanned systems caused by non-Gaussian noise from coupled sensors in complex interference environments in underground space, improving the positioning accuracy and anti-interference capability of unmanned systems in underground space.

Description

Technical Field

[0001] This invention relates to the field of unmanned system positioning technology, specifically to a method and system for locating unmanned systems in underground space under multi-source interference. Background Technology

[0002] When unmanned systems perform positioning in underground spaces with multi-source interference, they face various interference factors such as electromagnetic interference, impact vibration, dust and water mist, and cyberattacks. These interferences significantly affect the wireless positioning signals and data acquisition of micro-inertial sensors, resulting in a complex non-Gaussian distribution of positioning sensor error characteristics. The positioning sensors of unmanned systems are affected by non-Gaussian noise and unknown interference from different sources, which may prevent them from accurately obtaining the position estimation information of the unmanned system, thus increasing the positioning error and affecting the system's operation. Therefore, in complex environments with multi-source interference, designing a highly robust combined positioning method and system is crucial to improving the autonomous operation capability of unmanned systems in underground spaces.

[0003] Existing technologies, such as the invention patent application with publication number CN116734846A, disclose a seamless positioning method for mine vehicles based on GNSS / UWB fusion IMU. This method uses GNSS fusion IMU positioning above ground and UWB fusion IMU positioning underground. The above-ground and underground interactive areas are positioned by confidence fusion. The position information of GNSS or UWB and IMU is input into a Kalman filter for calibration to correct the original solution position of the system.

[0004] Existing technologies, such as the invention patent application with publication number CN104156564A, disclose a method for simplifying downhole equipment detection rules based on discrete Kalman filtering. This method uses GPS equipment to determine the position and speed of the inspection personnel and uses the discrete Kalman filtering algorithm to predict the next state, thereby selecting the corresponding roadway and equipment.

[0005] Regarding the above solution, the applicant of this invention has found that the above technology has at least the following technical problems: 1. This method only considers the combined positioning under the ideal Gaussian distribution error characteristics. When the sensor error characteristics exhibit significant non-Gaussian characteristics or other unknown interference effects, the positioning accuracy will be significantly reduced, and the positioning accuracy cannot be guaranteed.

[0006] 2. The data after Kalman filter calibration has a certain error. The above scheme does not estimate and update the corrected data after Kalman filter calibration, and cannot guarantee the accuracy of the corrected data. Summary of the Invention

[0007] To address the aforementioned technical shortcomings, the present invention aims to provide a method and system for locating unmanned systems in underground space under multi-source interference.

[0008] To solve the above-mentioned technical problems, the present invention adopts the following technical solution: In the first aspect, the present invention provides a positioning method for an unmanned system in underground space under multi-source interference, including the following steps: Step 1, data acquisition: In underground space, the navigation information of the unmanned system is obtained by using a combined navigation system of UWB tags and micro inertial measurement units mounted on the unmanned system and a preset UWB ranging base station.

[0009] Step 2: Construct a system model: Obtain various motion data of the unmanned system and establish a system model.

[0010] Step 3: Filter parameter update: Obtain the prior state vector and covariance matrix of the system, define the cost function within the framework of statistical similarity, and iterate the mean of the state vector and the error covariance matrix at each acquisition time to obtain the posterior state vector and error covariance matrix.

[0011] Step 4: Jointly infer the noise covariance matrix: Based on the obtained posterior state estimate, variational Bayesian learning is used within a sliding window to jointly infer the noise covariance matrix.

[0012] Step 5, Output: After the loop iteration is complete, output the position estimation result and covariance matrix of the unmanned system.

[0013] Preferably, the process of constructing the system model is as follows: S11, constructing the state transition matrix and the measurement matrix, while using sensors to acquire measurement noise.

[0014] S12. Obtain the target values ​​of various motion data of the unmanned system at each acquisition time, and call them the state vector, denoted as x. Then x k This represents the state vector of the unmanned system at the k-th data acquisition time.

[0015] S13. Obtain the measured values ​​of various motion data of the unmanned system at each acquisition time, and call them the measurement vector, denoted as y. Then y k This represents the measurement variable of the unmanned system at the k-th data acquisition time.

[0016] S14. Construct a linear state-space model with control inputs containing process noise, measurement noise, and non-Gaussian noise:

[0017] x k =F k x k-1 +η k β k +w k ,

[0018] y k =H k x k +vk +ψ k τ k ,

[0019] In the formula, F k This represents the state transition matrix of the unmanned system at the k-th data acquisition time. Let n be the measurement vector, and n be the dimension of the measurement vector. For n-dimensional real numbers, For m×m dimensional real numbers, For n×m dimensional real numbers, Let m be the system state vector, and m be the dimension of the system state vector. w is an m-dimensional real number k The process noise of the unmanned system at the k-th data acquisition time is represented by [the noise level]. v k The measurement noise of the unmanned system at the k-th data acquisition time is represented by . β k and τ k Both represent the non-Gaussian interference noise of the unmanned system at the k-th acquisition time, ψ k η represents the probability that the unmanned system is affected by non-Gaussian noise at the k-th data acquisition time. k This represents the probability that the measured variable is affected by non-Gaussian noise at the k-th acquisition time.

[0020] Preferably, the specific process for constructing the state transition matrix and the measurement matrix is ​​as follows:

[0021]

[0022] H k =[I n 0],

[0023] In the formula, ΔT represents the preset time interval duration, ΔT = 1s, I n It represents an n-dimensional identity matrix.

[0024] Preferably, the specific process of constructing the measurement update equation is as follows: S21, Solve for the prior state vector of the system using the Kalman filter method. and covariance matrix P k|k-1 .

[0025] S22. Within the statistical similarity framework, the cost function q(x) is defined by maximizing the use of system equations and covariance information. k ), q(x k The distribution is approximately Gaussian, i.e., q(x) k )≈N(x k μ k ,Σk ), where μ k Σ represents the mean of the state vector at the k-th acquisition time. k This represents the estimation error covariance matrix of the unmanned system at the k-th data acquisition time.

[0026] S23. Iterate over the mean state vector and the estimation error covariance matrix in the cost function at each acquisition time, and determine whether the result after each iteration meets the iteration termination condition. If the result after a certain iteration does not meet the condition, continue iterating until the result of the iteration meets the condition. Then, take the mean state vector after that iteration as the state vector estimate, and output the state vector estimate and the estimation error covariance matrix after that iteration.

[0027] Preferably, the iterative process of the mean state vector and the estimated error covariance matrix in the cost function at each acquisition time is as follows: S31, obtain the measurement variables of the unmanned system at each acquisition time and the measurement noise covariance matrix at each acquisition time.

[0028] S32, q(x) k The maximization problem of μ is transformed into a problem concerning μ. k and Σ k The maximization problem and Let represent the optimal posterior probability density function obtained through i fixed-point iterations. According to the maximum criterion, we can obtain:

[0029]

[0030] In the formula, This represents the prior state vector of the unmanned system at the k-th data acquisition time. The Kalman filter gain at the k-th acquisition time in the i-th iteration is calculated as follows:

[0031]

[0032] In the formula, The covariance matrix representing the one-step prediction error estimate of the unmanned system in the i-th iteration at the k-th data acquisition time is denoted as . This represents the one-step predictive measurement noise covariance matrix corrected by the unmanned system in the i-th iteration at the k-th acquisition time.

[0033]

[0034] In the formula, R k The measurement noise covariance matrix represents the measurement noise at the k-th acquisition time.

[0035] The one-step prediction noise covariance matrix corrected in the i-th iteration at the k-th acquisition time and The calculation is as follows:

[0036]

[0037] In the formula, P k|k-1 This represents the covariance matrix of the one-step prediction error estimation at the k-th acquisition time. The auxiliary variable representing the i-th iteration at the k-th acquisition time is calculated as follows:

[0038]

[0039] In the formula, v represents the degree of freedom parameter, and σ represents the kernel bandwidth. and The auxiliary variable representing the i-th iteration at the k-th acquisition time is calculated as follows:

[0040]

[0041] Preferably, the process of determining whether the result after each iteration satisfies the iteration termination condition is as follows: At each acquisition time, during each iteration, the number of iterations and the result after each iteration are recorded, and it is determined whether the iteration termination condition is met, i.e.:

[0042]

[0043] In the formula, ò represents the termination threshold, which is reached when the above inequality holds or after N... m -1 iterations represent satisfying the iteration termination condition, when the above inequality does not hold and N iterations have not been completed. m -1 iterations indicate that the iteration termination condition is not met.

[0044] Preferably, the joint inference of the noise covariance matrix is ​​carried out as follows: S41, Set a time interval, denoted as l, where l∈[k-L+1,k]. Within the time interval, define the state transition probability density function and the measurement likelihood probability density function:

[0045] p(x l |x l-1 Q (k) )=N(x l ;F l x l-1 Q (k) ), p(z l |x l ,R (k) )=N(z l H l x l ,R (k) ),

[0046] In the formula, p(x) l |x l-1 Q (k) p(z) represents the state transition probability density function of the unmanned system during the acquisition time [k-L+1,k]. l |x l ,R (k) Q represents the measurement likelihood probability density function of the unmanned system during the acquisition time [k-L+1,k]. (k) R (k) Both represent the noise covariance matrix.

[0047] S42. Construct the noise covariance matrix, the noise covariance matrix Q. (k) and R (k) The prior distribution is modeled using the inverse Wissaud distribution as follows:

[0048]

[0049] In the formula y 1:k-L This represents the set of measurements taken at time 1:kL. and p(Q) (k) |y 1:k-L ) and p(R (k) |y 1:k-L The degrees of freedom parameters and inverse scaling matrix of ) and Let be the posterior variable at time kL, and ρ be the forgetting factor.

[0050] S43. Correct the scaling matrix to obtain the corrected scaling matrix R. (k),(i+1) and Q (k),(i+1) They can be represented as follows:

[0051]

[0052] In the formula, These represent the posterior variables of the unmanned system at the k-th acquisition time during the i-th iteration.

[0053] S44. According to the RTS smoothing algorithm, smooth the estimated vector. covariance matrix and its smoothing gain

[0054]

[0055] And for posterior variables and Update.

[0056] S45. Determine whether the results of each iteration at each sample collection time meet the iteration termination condition. If the result of an iteration at a certain sample collection time does not meet the iteration termination condition, continue iterating until the iteration result meets the iteration termination condition, and then output the obtained state vector estimate. and its corresponding estimation error covariance matrix

[0057] Preferably, the posterior variable is... and The update process is as follows:

[0058]

[0059] Preferably, the process of determining whether the results of each iteration at each sample collection time meet the iteration termination condition is as follows: During each iteration at each sample collection time, the number of iterations and the results after each iteration are recorded, and it is determined whether the iteration termination condition is met, i.e.:

[0060]

[0061] When the above inequality holds or passes through N vb -1 iterations represent satisfying the iteration termination condition, when the above inequality does not hold and N iterations have not been completed. vb -1 iterations indicate that the iteration termination condition is not met.

[0062] Secondly, the present invention provides a positioning system for an unmanned system in underground space under multi-source interference, comprising the following modules: a data acquisition module for acquiring navigation information of the unmanned system in underground space using a combined navigation system of UWB tags and micro inertial measurement units mounted on the unmanned system and a preset UWB ranging base station.

[0063] The system model building module is used to acquire various motion data of the unmanned system and build a system model.

[0064] The filter parameter update module is used to obtain the prior state vector and covariance matrix of the system, and defines the cost function within the framework of statistical similarity. At the same time, it iterates the mean of the state vector and the error covariance matrix at each acquisition time to obtain the posterior state vector and error covariance matrix.

[0065] The joint inference noise covariance matrix module uses variational Bayesian learning within a sliding window to jointly infer the noise covariance matrix based on the obtained posterior state estimate.

[0066] The output module is used for iterative looping to output the position estimation results and covariance matrix of the unmanned system.

[0067] The beneficial effects of this invention are as follows: This invention provides a method and system for locating unmanned systems in underground spaces under multi-source interference. First, various motion data of the unmanned system are collected at different acquisition times, and a system model is constructed. To address the impact of non-Gaussian noise in the multi-source interference environment of underground spaces on the unmanned system's location, a Student's t kernel function is used for robust processing of abnormal data. Multiple measures of a sliding window are used, and the noise-contaminated covariance matrix is ​​adaptively adjusted using a variational Bayesian method. The posterior distributions of the system state, unknown noise parameters, and auxiliary variables are jointly solved using the variational Bayesian method, ultimately obtaining the optimal estimated position of the unmanned system. This invention offers high estimation accuracy, significantly improving the positioning accuracy of unmanned systems in multi-source non-Gaussian noise environments in underground spaces, ensuring the system's mission execution capability in complex environments, and achieving high-precision positioning. Attached Figure Description

[0068] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0069] Figure 1 This is a schematic diagram of the implementation steps of the method of the present invention.

[0070] Figure 2 This is a schematic diagram of the system structure connection of the present invention.

[0071] Figure 3 This is a physical image of the present invention.

[0072] Figure 4 , Figure 5 The root mean square error values ​​for position and velocity are respectively compared between the method proposed in this invention and existing methods such as maximum entropy Kalman filtering, variational Bayesian learning filtering, statistical similarity filtering, and sliding window filtering based on Student t-distribution variational learning. Detailed Implementation

[0073] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0074] Please see Figure 1As shown, in a first aspect, the present invention provides a method for locating an unmanned system in underground space under multi-source interference, comprising the following steps: Step 1, data acquisition: In underground space, navigation information of the unmanned system is acquired using a combined navigation system of UWB tags and micro inertial measurement units mounted on the unmanned system and a preset UWB ranging base station.

[0075] It should be noted that navigation information includes speed, position, and angular velocity.

[0076] Step 2: Construct a system model: Obtain various motion data of the unmanned system and construct a system model.

[0077] In a specific embodiment, the process of constructing the system model is as follows: S11, constructing the state transition matrix and the measurement matrix, while using sensors to acquire measurement noise.

[0078] S12. Obtain the target values ​​of various motion data of the unmanned system at each acquisition time, and call them the state vector, denoted as x. Then x k This represents the state vector of the unmanned system at the k-th data acquisition time.

[0079] It should be noted that the target values ​​for each sports data point were set by the staff.

[0080] S13. Obtain the measured values ​​of various motion data of the unmanned system at each acquisition time, and call them the measurement vector, denoted as y. Then y k This represents the measurement variable of the unmanned system at the k-th data acquisition time.

[0081] S14. Construct a linear state-space model with control inputs containing process noise, measurement noise, and non-Gaussian noise:

[0082] x k =F k x k-1 +η k β k +w k ,

[0083] y k =H k x k +v k +ψ k τ k ,

[0084] In the formula, F k This represents the state transition matrix of the unmanned system at the k-th data acquisition time. Let n be the measurement vector, and n be the dimension of the measurement vector. For n-dimensional real numbers, For m×m dimensional real numbers, For n×m dimensional real numbers, Let m be the system state vector, and m be the dimension of the system state vector. w is an m-dimensional real number k The process noise of the unmanned system at the k-th data acquisition time is represented by [the noise level]. v k The measurement noise of the unmanned system at the k-th data acquisition time is represented by . β k and τ k Both represent the non-Gaussian interference noise of the unmanned system at the k-th acquisition time, ψ k η represents the probability that the unmanned system is affected by non-Gaussian noise at the k-th data acquisition time. k This represents the probability that the measured variable is affected by non-Gaussian noise at the k-th acquisition time.

[0085] The specific process for constructing the state transition matrix and measurement matrix described above is as follows:

[0086]

[0087] H k =[I n 0],

[0088] In the formula, ΔT represents the preset time interval duration, ΔT = 1s, I n It represents an n-dimensional identity matrix.

[0089] Step 3: Filter parameter update: Obtain the prior state vector and covariance matrix of the system, define the cost function within the framework of statistical similarity, and iterate the mean of the state vector and the error covariance matrix at each acquisition time to obtain the posterior state vector and error covariance matrix.

[0090] In a specific embodiment, the filter parameter update process is as follows: S21, Solve for the prior state vector of the system using the Kalman filter method. and covariance matrix P k|k-1 .

[0091] S22. Within the statistical similarity framework, the cost function q(x) is defined by maximizing the use of system equations and covariance information. k ), q(x k The distribution is approximately Gaussian, i.e., q(x) k )≈N(x k μ k ,Σ k ), where μ k Σ represents the mean of the state variable at the k-th acquisition time. kThis represents the estimation error covariance matrix of the unmanned system at the k-th data acquisition time.

[0092] S23. Iterate over the mean state vector and the estimation error covariance matrix in the cost function at each acquisition time, and determine whether the result after each iteration meets the iteration termination condition. If the result after a certain iteration does not meet the condition, continue iterating until the result of the iteration meets the condition. Then, take the mean state vector after that iteration as the state vector estimate, and output the state vector estimate and the estimation error covariance matrix after that iteration.

[0093] In the above, the iterative process of the mean state vector and the estimated error covariance matrix in the cost function at each acquisition time is as follows: S31, obtain the measurement variables of the unmanned system at each acquisition time and the measurement noise covariance matrix at each acquisition time.

[0094] S32, q(x) k The maximization problem of μ is transformed into a problem concerning μ. k and Σ k The maximization problem and Let represent the optimal posterior probability density function obtained through i fixed-point iterations. According to the maximum criterion, we can obtain:

[0095]

[0096] In the formula, This represents the prior state vector of the unmanned system at the k-th data acquisition time. The Kalman filter gain at the k-th acquisition time in the i-th iteration is calculated as follows:

[0097]

[0098] In the formula, The covariance matrix representing the one-step prediction error estimate of the unmanned system in the i-th iteration at the k-th data acquisition time is denoted as . This represents the one-step predictive measurement noise covariance matrix corrected by the unmanned system in the i-th iteration at the k-th acquisition time.

[0099]

[0100] In the formula, R k The measurement noise covariance matrix represents the measurement noise at the k-th acquisition time.

[0101] The corrected one-step prediction noise covariance matrix in the i-th iteration at the k-th acquisition time and The calculation is as follows:

[0102]

[0103] In the formula, P k|k-1 This represents the covariance matrix of the one-step prediction error estimation at the k-th acquisition time. The auxiliary variable representing the i-th iteration at the k-th acquisition time is calculated as follows:

[0104]

[0105] In the formula, v represents the degree of freedom parameter, and σ represents the kernel bandwidth. and The auxiliary variable representing the i-th iteration at the k-th acquisition time is calculated as follows:

[0106]

[0107] The above process of determining whether the result after each iteration satisfies the iteration termination condition is as follows: During each iteration at each acquisition time, the number of iterations and the result after each iteration are recorded, and it is determined whether the iteration termination condition is met, i.e.:

[0108]

[0109] In the formula, ò represents the termination threshold, which is reached when the above inequality holds or after N... m -1 iterations represent satisfying the iteration termination condition, when the above inequality does not hold and N iterations have not been completed. m -1 iterations indicate that the iteration termination condition is not met.

[0110] It should be noted that the termination threshold was obtained by the staff based on experiments, and the preset iteration number threshold was set by the staff.

[0111] Step 4: Jointly infer the noise covariance matrix: Based on the obtained posterior state estimate, variational Bayesian learning is used within a sliding window to jointly infer the noise covariance matrix.

[0112] In a specific embodiment, the joint inference of the noise covariance matrix is ​​characterized by the following process: S41, setting a time interval, denoted as l, where l∈[k-L+1,k], and defining the state transition probability density function and the measurement likelihood probability density function within the time interval:

[0113] p(x l |x l-1 Q (k) )=N(x l ;F l x l-1 Q (k) ), p(z l |xl ,R (k) )=N(z l H l x l ,R (k) ),

[0114] In the formula p(x l |x l-1 Q (k) p(z) represents the state transition probability density function of the unmanned system during the acquisition time [k-L+1,k]. l |x l ,R (k) Q represents the measurement likelihood probability density function of the unmanned system during the acquisition time [k-L+1,k]. (k) R (k) Both represent the noise covariance matrix.

[0115] S42. Construct the noise covariance matrix, the noise covariance matrix Q. (k) and R (k) The prior distribution is modeled using the inverse Wissaud distribution as follows:

[0116]

[0117] In the formula, y 1:k-L This represents the set of measurements taken at time 1:kL. and p(Q) (k) |y 1:k-L ) and p(R (k )|y 1:k-L The degrees of freedom parameters and inverse scaling matrix of ) and Let be the posterior variable at time kL, and ρ be the forgetting factor.

[0118] S43. Correct the scaling matrix to obtain the corrected scaling matrix R. (k),(i+1) and Q (k),(i+1) They can be represented as follows:

[0119]

[0120] In the formula, These represent the posterior variables of the unmanned system at the k-th acquisition time during the i-th iteration.

[0121] S44. According to the RTS smoothing algorithm, smooth the estimated vector. covariance matrix and its smoothing gain

[0122]

[0123] And for posterior variables and Update.

[0124] S45. Determine whether the results of each iteration at each sample collection time meet the iteration termination condition. If the result of an iteration at a certain sample collection time does not meet the iteration termination condition, continue iterating until the iteration result meets the iteration termination condition, and then output the obtained state vector estimate. and its corresponding estimation error covariance matrix

[0125] In the above, the posterior variable is described. and The update process is as follows:

[0126]

[0127] The above process of determining whether the results of each iteration at each sample collection time meet the iteration termination condition is as follows: During each iteration at each sample collection time, the number of iterations and the results after each iteration are recorded, and it is determined whether the iteration termination condition is met.

[0128]

[0129] When the above inequality holds or passes through N vb -1 iterations represent satisfying the iteration termination condition, when the above inequality does not hold and N iterations have not been completed. vb -1 iterations indicate that the iteration termination condition is not met.

[0130] Step 5, Output: After the loop iteration is complete, output the position estimation result and covariance matrix of the unmanned system.

[0131] Please see Figure 2 As shown, in a second aspect, the present invention provides a positioning system for an unmanned underground space system under multi-source interference, including a data acquisition module, a system model construction module, a measurement update equation construction module, a noise covariance matrix inference module, an output module, and a database.

[0132] The data acquisition module is used in underground spaces to acquire navigation information of unmanned systems by using a combined navigation system of UWB tags and micro inertial measurement units mounted on the unmanned system, as well as a preset UWB ranging base station.

[0133] The system model building module is used to acquire various motion data of the unmanned system and build a system model.

[0134] The filter parameter update module is used to obtain the prior state vector and covariance matrix of the system, and defines the cost function within the framework of statistical similarity. At the same time, it iterates the mean of the state vector and the error covariance matrix at each acquisition time to obtain the posterior state vector and error covariance matrix.

[0135] The joint inference noise covariance matrix module uses variational Bayesian learning within a sliding window to jointly infer the noise covariance matrix based on the obtained posterior state estimate.

[0136] The output module is used for iterative looping to output the position estimation results and covariance matrix of the unmanned system.

[0137] In this embodiment of the invention, motion data of the unmanned system are first collected at each acquisition time, and a system model is constructed. Then, based on the cost function defined within the framework of statistical similarity, the mean of the state vector and the error covariance matrix at each acquisition time are iterated. Afterwards, variational Bayesian learning is used to jointly infer the noise covariance matrix. Finally, the position estimation result and covariance matrix of the unmanned system are output, ensuring the accuracy of the corrected data and the precision of the unmanned system's positioning.

[0138] The above description is merely an example and illustration of the concept of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described or use similar methods to replace them, as long as they do not deviate from the concept of the invention or exceed the scope defined in this specification, they should all fall within the protection scope of the present invention.

Claims

1. A method for locating unmanned systems in underground space under multi-source interference, characterized in that, Includes the following steps: Step 1: Data Acquisition: In underground space, the navigation information of the unmanned system is acquired using a combined navigation system of UWB tags and micro inertial measurement units mounted on the unmanned system, along with a pre-set UWB ranging base station. Step 2: Constructing the system model: Acquire various motion data of the unmanned system and establish a system model. The specific process is as follows: S11: Construct the state transition matrix and measurement matrix, and at the same time use sensors to acquire measurement noise; S12. Obtain the target values ​​of each motion data of the unmanned system at each acquisition time, and call them the state vector, denoted as... ,but Representing unmanned systems in the The state vector at each acquisition moment; S13. Acquire the measurement values ​​of each motion data of the unmanned system at each acquisition time, and call them the measurement vector, denoted as . ,but Representing unmanned systems in the Measurement vectors at each acquisition time; S14. Construct a linear state-space model with control inputs containing process noise, measurement noise, and non-Gaussian noise: ， ， In the formula, Representing unmanned systems in the State transition matrix at each acquisition time, For measurement vectors, Let be the dimension of the measurement vector. for dimensional real number, , for dimensional real number, , for dimensional real number, Let be the system state vector. Let be the dimension of the system state vector. for dimensional real number, Representing unmanned systems in the Process noise at each acquisition moment , Representing unmanned systems in the Measurement noise at each acquisition time, , and All represent unmanned systems in the first Non-Gaussian interference noise at each acquisition time Representing unmanned systems in the The probability that a data acquisition moment is affected by non-Gaussian noise. The representative measurement variable is in the th The probability that a data acquisition moment is affected by non-Gaussian noise; Step 3, Filter Parameter Update: Obtain the prior state vector and covariance matrix of the system, define the cost function within the framework of statistical similarity, and iterate the mean state vector and error covariance matrix at each acquisition time to obtain the posterior state vector and error covariance matrix. Step 4: Jointly infer the noise covariance matrix: Based on the obtained posterior state estimate, variational Bayesian learning is used within a sliding window to jointly infer the noise covariance matrix; Step 5, Output: After the loop iteration is complete, output the position estimation result and covariance matrix of the unmanned system.

2. The method for locating an unmanned system in underground space under multi-source interference as described in claim 1, characterized in that, The specific process for constructing the state transition matrix and measurement matrix is ​​as follows: ， ， In the formula This represents the preset time interval duration. , represent 3D identity matrix.

3. The method for locating an unmanned system in underground space under multi-source interference as described in claim 2, characterized in that, The filter parameter update process is as follows: S21. Solve for the prior state vector of the system using the Kalman filter method. and covariance matrix ; S22. Within the statistical similarity framework, a cost function is defined by maximizing the use of system equations and covariance information. ,Will It is approximately a Gaussian distribution, i.e. In the formula Represents the state variable in the th... The average value at each collection time point. Representing unmanned systems in the The estimation error covariance matrix at each acquisition time; S23. Iterate over the mean state vector and the estimation error covariance matrix in the cost function at each acquisition time, and determine whether the result after each iteration meets the iteration termination condition. If the result after a certain iteration does not meet the condition, continue iterating until the result of the iteration meets the iteration termination condition. Then, take the mean state vector after that iteration as the state vector estimate, and output the state vector estimate and the estimation error covariance matrix after that iteration.

4. The method for locating an unmanned system in underground space under multi-source interference as described in claim 3, characterized in that, The iterative process for the mean state vector and the estimation error covariance matrix in the cost function at each acquisition time is as follows: S31. Obtain the measurement variables and measurement noise covariance matrix of the unmanned system at each acquisition time. S32, will The maximization problem is transformed into a problem about and The maximization problem and Indicates passage The optimal posterior probability density function obtained by the fixed-point iterative approximation can be obtained according to the maximum value criterion: ， In the formula, Representing unmanned systems in the The prior state vector at each acquisition time. Representing the The first data collection time The Kalman filter gain in the next iteration is calculated as follows: ， In the formula, Representing unmanned systems in the The first data collection time The corrected one-step prediction error estimate covariance matrix in the next iteration Representing unmanned systems in the The first data collection time The corrected one-step predictive measurement noise covariance matrix in the next iteration; ， In the formula, Representing the Measurement noise covariance matrix at each acquisition time; In the The first data collection time The corrected one-step prediction noise covariance matrix in the next iteration and The calculation is as follows: ， In the formula, Representing the The covariance matrix of the one-step prediction error estimation at each acquisition time. , Representing the The first data collection time The auxiliary variables used in this iteration are calculated as follows: ， In the formula, Represents the degree of freedom parameter. Represents kernel bandwidth. and Representing the The first data collection time The auxiliary variables used in this iteration are calculated as follows: ， 。 5. The method for locating an unmanned system in underground space under multi-source interference as described in claim 3, characterized in that, The specific process for determining whether the result after each iteration satisfies the iteration termination condition is as follows: At each acquisition time, during each iteration, the number of iterations and the result after each iteration are recorded. Then: ， In the formula This represents the termination threshold, which is reached when the above inequality is true or has passed. In the second iteration, it represents that the iteration termination condition is met, when the above inequality does not hold and no iteration has been performed. In the second iteration, it means that the iteration termination condition is not met.

6. The method for locating an unmanned system in underground space under multi-source interference as described in claim 4, characterized in that, The joint inference of the noise covariance matrix is ​​carried out as follows: S41. Set the time interval, denoted as... ,but Within the time interval, define the state transition probability density function and the measurement likelihood probability density function: ， In the formula, Representing unmanned systems The state transition probability density function within the acquisition time. Representing unmanned systems The measurement likelihood probability density function at the acquisition time. , Both represent the noise covariance matrix; S42. Construct the noise covariance matrix. and The prior distribution is modeled using the inverse Wissaud distribution as follows: ， In the formula, express A collection of time-based measurements. , , and for and The degrees of freedom parameters and the inverse scaling matrix, , , and Its corresponding The posterior variable at time 1, Forgetting factor, , , , ; S43. Correct the scaling matrix to obtain the corrected scaling matrix. and They can be represented as follows: ， In the formula, , , , These represent the unmanned systems in the [number]th [year]. The first data collection time The posterior variable of the next iteration; S44. According to the RTS smoothing algorithm, smooth the estimated vector. Covariance matrix and its smoothing gain : ， ， ， And for posterior variables , , and Update; S45. Determine whether the results of each iteration at each sample collection time meet the iteration termination condition. If the result of an iteration at a certain sample collection time does not meet the iteration termination condition, continue iterating until the iteration result meets the iteration termination condition, and then output the obtained state vector estimate. and its corresponding estimation error covariance matrix .

7. The method for locating an unmanned system in underground space under multi-source interference as described in claim 6, characterized in that, The above and the posterior variables , , and The update process is as follows: ， ， ， 。 8. The method for locating an unmanned system in underground space under multi-source interference as described in claim 6, characterized in that, The process of determining whether the results of each iteration at each sample collection time meet the iteration termination condition is as follows: During each iteration at each sample collection time, record the number of iterations and the result after each iteration. ， When the above inequality holds or passes In the second iteration, it represents that the iteration termination condition is met, when the above inequality does not hold and no iteration has been performed. In the second iteration, it means that the iteration termination condition is not met.

9. A positioning system for an unmanned underground system implementing the multi-source interference positioning method according to any one of claims 1-8, characterized in that, include: The data acquisition module is used in underground spaces to acquire navigation information of unmanned systems by using a combined navigation system of UWB tags and micro inertial measurement units mounted on the unmanned system and a preset UWB ranging base station. The system model building module is used to acquire various motion data of the unmanned system and build a system model. The specific process is as follows: S11, construct the state transition matrix and measurement matrix, and at the same time use sensors to acquire measurement noise; S12. Obtain the target values ​​of each motion data of the unmanned system at each acquisition time, and call them the state vector, denoted as... ,but Representing unmanned systems in the The state vector at each acquisition moment; S13. Acquire the measurement values ​​of each motion data of the unmanned system at each acquisition time, and call them the measurement vector, denoted as . ,but Representing unmanned systems in the Measurement vectors at each acquisition time; S14. Construct a linear state-space model with control inputs containing process noise, measurement noise, and non-Gaussian noise: ， ， In the formula, Representing unmanned systems in the State transition matrix at each acquisition time, For measurement vectors, Let be the dimension of the measurement vector. for dimensional real number, , for dimensional real number, , for dimensional real number, Let be the system state vector. Let be the dimension of the system state vector. for dimensional real number, Representing unmanned systems in the Process noise at each acquisition moment , Representing unmanned systems in the Measurement noise at each acquisition time, , and All represent unmanned systems in the first Non-Gaussian interference noise at each acquisition time Representing unmanned systems in the The probability that a data acquisition moment is affected by non-Gaussian noise. The representative measurement variable is in the th The probability that a data acquisition moment is affected by non-Gaussian noise; The filter parameter update module is used to obtain the prior state vector and covariance matrix of the system, and define the cost function within the framework of statistical similarity. At the same time, it iterates the mean of the state vector and the error covariance matrix at each acquisition time to obtain the posterior state vector and error covariance matrix. The joint inference noise covariance matrix module uses variational Bayesian learning within a sliding window to jointly infer the noise covariance matrix based on the obtained posterior state estimate. The output module is used for iterative looping to output the position estimation results and covariance matrix of the unmanned system.

Images