An unmanned aerial vehicle ultra-wideband fine positioning method based on artificial noise excitation

Abstract

This invention belongs to the field of UAV positioning, specifically relating to a UAV ultra-wideband fine positioning method based on artificial noise excitation. First, for the UAV ultra-wideband and dynamic information fusion positioning problem, a nonlinear positioning system is established, incorporating a UAV motion state model and an ultra-wideband measurement model, while introducing a slanted t-distribution to describe the statistical characteristics of ultra-wideband measurement errors. Second, a filtering algorithm is designed to introduce artificial noise into the measurement residuals during the measurement update process, constructing residual neighborhood samples, and averaging the robust scale weights calculated from the residual samples to obtain noise-enhanced scale weights, which are used to adaptively correct the measurement update process. Finally, the noise-enhanced scale weights are used to complete the recursive estimation of the UAV state, iteratively executing the measurement update steps, and outputting the UAV position estimation results in real time. This method is applicable to UAV ultra-wideband positioning systems under complex interference conditions such as non-line-of-sight errors, multipath effects, and external disturbances.

Description

Technical Field

[0001] This invention belongs to the field of UAV positioning, specifically a method for UAV ultra-wideband fine positioning based on artificial noise excitation, which is applicable to UAV ultra-wideband positioning systems under complex interference conditions such as non-line-of-sight errors, multipath effects and external disturbances. Background Technology

[0002] In recent years, drones have been widely used in indoor inspection, warehousing and logistics, emergency rescue, and intelligent manufacturing. High-precision and high-reliability positioning technology is the foundation for the safe and stable operation of drones. In indoor and semi-enclosed environments where satellite signals are blocked, ultra-wideband technology has become one of the important technical means for indoor drone positioning due to its advantages such as high positioning accuracy, strong multipath resistance, and low power consumption.

[0003] However, in practical applications, ultra-wideband positioning systems still face a variety of complex interference factors. On the one hand, UAVs are subject to external dynamic disturbances such as wind disturbances during flight, causing their motion state to change rapidly. On the other hand, non-line-of-sight propagation and multipath effects in complex environments introduce measurement errors with heavy tails and skewed characteristics, making it difficult for traditional filtering algorithms based on Gaussian noise assumptions to accurately model the characteristics of measurement errors, thus causing a decrease in positioning accuracy or even filter divergence.

[0004] To improve the accuracy and robustness of UAV ultra-wideband (UWB) positioning, existing technologies typically integrate UWB measurements and UAV dynamics for indoor UAV positioning. However, traditional filtering methods often rely on single-residual weight calculations when handling anomalous measurements, making them susceptible to the significant impact of occasional anomalous measurements and unable to maintain stable performance in complex interference environments. Furthermore, most existing methods focus on suppressing single interference sources, lacking a collaborative processing mechanism for dynamic interference and non-Gaussian measurement noise. Therefore, developing filtering methods with noise enhancement mechanisms to fully exploit the inherent information of measurement residuals and their statistical characteristics, addressing the complex multi-source interference present in UAV UWB positioning, including dynamic disturbances, non-line-of-sight propagation, and multipath effects, has become an urgent need.

[0005] Chinese invention patent application CN108520640B proposes a UAV navigation method based on ultra-wideband (UWB). However, this method focuses on acquiring relative position and navigation direction, without combining UAV dynamics models for state recursive estimation, and does not robustly handle the uncertainty of UWB measurement noise in complex environments. Chinese invention patent application CN114993299B proposes a UAV localization method based on UWB, but this method still relies on the Gaussian noise assumption, making it difficult to effectively handle anomalous measurements with skewed and heavy-tailed characteristics in non-line-of-sight and multipath environments. Furthermore, it does not introduce a noise enhancement mechanism to suppress the impact of single anomalous measurements on filter updates. Summary of the Invention

[0006] To overcome the shortcomings of existing technologies, this invention proposes an ultra-wideband fine-tuning localization method for UAVs based on artificial noise excitation, aiming to improve localization accuracy and robustness. This method introduces a noise enhancement mechanism based on the UAV dynamics model and the ultra-wideband measurement model. By perturbing the measurement residuals and statistically averaging the robustness scale weights, the impact of single abnormal measurements on the state estimation results is effectively reduced. Simultaneously, by incorporating a composite interference filtering framework, it achieves simultaneous cancellation and suppression of external dynamic disturbances and skewed, heavy-tailed measurement noise caused by non-line-of-sight and multipath interference.

[0007] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0008] A method for ultra-wideband fine positioning of unmanned aerial vehicles (UAVs) based on artificial noise excitation includes the following steps:

[0009] Step 1: Establish a nonlinear positioning system that includes a UAV motion state model and an ultra-wideband measurement model, and introduce a slant-t distribution to describe the statistical characteristics of the ultra-wideband measurement error.

[0010] Step 2: Design a variational Bayesian filtering algorithm to introduce artificial noise into the measurement residuals during the measurement update process, construct residual neighborhood samples, and average the robust scale weights calculated from the residual samples to obtain noise-enhanced scale weights, which are used to adaptively correct the measurement update process.

[0011] Step 3: Use noise-enhanced scale weights to perform recursive estimation of the UAV state, iterate through the measurement update steps, and output the UAV position estimation results in real time.

[0012] Step 1 includes:

[0013] The nonlinear positioning system, which includes a UAV motion state model and an ultra-wideband measurement model, is as follows:

[0014] ;

[0015] in, It is the system status; and It is a control input; It is external wind disturbance; These are ultra-wideband measurements; It is process noise and follows a Gaussian distribution. ; It is measurement noise, which follows a slanted t-distribution; This is the state transition matrix; and To control the input matrix; The interference input matrix; Assign a matrix to the process noise; Measurement matrix; ultra-wideband non-Gaussian measurement noise The amount Follows a slanted t-distribution:

[0016] ;

[0017] in, Here, is a location parameter, representing the mean. is a scale parameter, representing the standard deviation; Here, represents the shape parameter, indicating the degree of skewness; This is a parameter representing the degrees of freedom, reflecting kurtosis characteristics.

[0018] The positioning accuracy of drones can also be affected by wind disturbance; wind disturbance Modeled as:

[0019] ;

[0020] From time At the time Increase in wind disturbance satisfy ;in, It is the sampling time interval between two adjacent moments. It indicates time. The parameter corresponding to the upper bound of the rate of change of disturbance.

[0021] Step 2 includes:

[0022] for The filter algorithm consists of three steps: interference estimation, state prediction, and measurement update.

[0023] The interference estimation steps are as follows:

[0024] Design the following observer:

[0025] ;

[0026] in, The observer state estimate, This is an estimate of the interference. Define the augmented state transition matrix as the predicted output of the system. Gain matrix Augmented output matrix as follows:

[0027] ;

[0028] State estimation gain matrix and interference estimation gain matrix The selection needs to ensure It is stable, i.e., the matrix All poles are located within the unit circle to achieve unbiased estimation of disturbances.

[0029] The state prediction steps are as follows:

[0030] The estimated values ​​of the system state and its covariance:

[0031] ;

[0032] The estimated value of the predicted state at this time and its estimation error covariance Current measurements have not yet been taken into account. .

[0033] The measurement update steps are as follows:

[0034] In the measurement update process, artificial noise is introduced in the calculation stage of weight expectation to reduce the impact of abnormal residuals;

[0035] (a) First, perform initialization:

[0036] , ;

[0037] in, It is a moment The expected scale weights It is a moment The mean of the perturbation residuals;

[0038] (b) Generate L artificial noise samples, where the l-th noise sample is:

[0039] ;

[0040] Among them, the variance of artificial noise Selected as:

[0041] ;

[0042] Where m is the measurement dimension; As a proportional coefficient, a limit can be set to ensure that the disturbance is controllable; Represents the trace operation of a matrix; Used to determine the variance of artificial noise, its value is ;

[0043] (c) Perform iteration:

[0044] Calculate the centralized residual :

[0045] ;

[0046] in, The state estimate is given; to improve the stability of the scale weight estimate, artificial noise is added to the centered residual to construct the perturbed residual; after adding artificial noise, the l-th perturbed residual is... for:

[0047] ;

[0048] For the l-th disturbance residual Constructing an auxiliary statistical matrix ,

[0049]

[0050] And based on this, the expected value of the scale weights is calculated; where, It is a latent variable The second-order moment statistics matrix;

[0051] For the l-th disturbance residual Calculate the expected value of its scale weights:

[0052] ;

[0053] The noise enhancement weights are obtained by averaging the expectations of the L scale weights. The value of the i-th diagonal element is:

[0054] ;

[0055] use Update hidden variables and its covariance ,

[0056] ;

[0057] in, It is a latent variable The auxiliary gain matrix; the hidden variables are obtained. posterior mean ,

[0058] ;

[0059] in, , , It is a normal distribution The probability density function, It is a normal distribution The cumulative distribution function;

[0060] state Covariance renew:

[0061] ;

[0062] in, This is the gain matrix for state updates; these update results can be used as input for the next iteration; finally, after... After the next iteration, the updated posterior state estimate can be obtained. and covariance matrix .

[0063] Step 3 is implemented as follows:

[0064] By recursively executing step 2, external wind disturbances can be estimated and compensated by the interference observer; unknown statistical information of the biased measurement noise caused by non-line-of-sight errors and multipath effects is identified using a variational Bayesian approximation; the UAV state is recursively estimated based on noise-enhanced scale weights, the measurement update step is executed iteratively, and the UAV position estimation result is output in real time; specifically, for UAV position state estimation at each moment and its measurement error covariance The output is:

[0065] ;

[0066] In this way, non-Gaussian noise and skewed interference in ultra-wideband measurements can be suppressed simultaneously, effectively reducing the impact of single abnormal measurements on positioning results and improving the numerical stability of the filtering process.

[0067] The advantages of this invention compared to existing technologies are as follows: This invention is applicable to UAV ultra-wideband positioning systems. By introducing a noise enhancement mechanism during the filtering measurement update process and averaging the robust scale weights within a neighborhood, the impact of a single abnormal measurement on the state estimation results is effectively reduced. Simultaneously, by combining the UAV dynamics model, it achieves simultaneous processing of external dynamic interference and non-Gaussian measurement noise, thereby improving the positioning accuracy and estimation stability of the UAV in complex environments without significantly increasing computational complexity. This method is particularly suitable for indoor and semi-enclosed scenarios where satellite signals are denied, possessing strong engineering practicality and broad application prospects. The main components include: establishing a nonlinear positioning system incorporating a UAV motion state model and an ultra-wideband measurement model, while introducing a slanted t-distribution to describe the statistical characteristics of ultra-wideband measurement errors; designing a variational Bayesian filtering algorithm to introduce artificial noise into the measurement residuals during the measurement update process, constructing residual neighborhood samples, and averaging the robust scale weights calculated from the residual samples to obtain noise-enhanced scale weights for adaptive correction of the measurement update process; using the noise-enhanced scale weights to complete the recursive estimation of the UAV state, iteratively executing the measurement update steps, and outputting the UAV position estimation results in real time. Attached Figure Description

[0068] Figure 1 This is a system block diagram of an ultra-wideband fine positioning method for unmanned aerial vehicles based on artificial noise excitation according to the present invention.

[0069] Figure 2 The diagram shows the UAV and its onboard ultra-wideband sensor hardware used for method verification in a specific embodiment.

[0070] Figure 3 The three-dimensional trajectory of the estimated position of the UAV is used to verify the method in a specific embodiment.

[0071] Figure 4 This is a comparison chart of UAV position estimation errors verified by the method in a specific embodiment.

[0072] Figure 5 This refers to the average scale weight value used in the UAV position estimation process during the method verification in a specific embodiment.

[0073] Figure 6 The values ​​of some diagonal elements of the scale weight matrix are used in the UAV position estimation process during the method verification in a specific embodiment. Detailed Implementation

[0074] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.

[0075] like Figure 1 As shown, a method for ultra-wideband fine positioning of unmanned aerial vehicles (UAVs) based on artificial noise excitation includes the following steps:

[0076] Step 1: Establish a nonlinear positioning system that includes a UAV motion state model and an ultra-wideband measurement model, and introduce a slant-t distribution to describe the statistical characteristics of the ultra-wideband measurement error.

[0077] Step 2: Design a variational Bayesian filtering algorithm to introduce artificial noise into the measurement residuals during the measurement update process, construct residual neighborhood samples, and average the robust scale weights calculated from the residual samples to obtain noise-enhanced scale weights, which are used to adaptively correct the measurement update process.

[0078] Step 3: Use noise-enhanced scale weights to perform recursive estimation of the UAV state, iterate through the measurement update steps, and output the UAV position estimation results in real time.

[0079] Step 1 includes:

[0080] The nonlinear positioning system, which includes a UAV motion state model and an ultra-wideband measurement model, is as follows:

[0081] ;

[0082] in, It is the system status; and It is a control input; It is external wind disturbance; These are ultra-wideband measurements; It is process noise and follows a Gaussian distribution. ; It is measurement noise, which follows a slanted t-distribution; This is the state transition matrix; and To control the input matrix; The interference input matrix; Assign a matrix to the process noise; Measurement matrix; ultra-wideband non-Gaussian measurement noise The amount Follows a slanted t-distribution:

[0083] ;

[0084] in, Here, is a location parameter, representing the mean. is a scale parameter, representing the standard deviation; Here, represents the shape parameter, indicating the degree of skewness; This is a parameter representing the degrees of freedom, reflecting kurtosis characteristics.

[0085] The positioning accuracy of drones can also be affected by wind disturbance; wind disturbance Modeled as:

[0086] ;

[0087] From time At the time Increase in wind disturbance satisfy ;in, It is the sampling time interval between two adjacent moments. It indicates time. The parameter corresponding to the upper bound of the rate of change of disturbance.

[0088] Step 2 includes:

[0089] for The filter algorithm consists of three steps: interference estimation, state prediction, and measurement update.

[0090] The interference estimation steps are as follows:

[0091] Design the following observer:

[0092] ;

[0093] in, The observer state estimate, This is an estimate of the interference. Define the augmented state transition matrix as the predicted output of the system. Gain matrix Augmented output matrix as follows:

[0094] ;

[0095] State estimation gain matrix and interference estimation gain matrix The selection needs to ensure It is stable, i.e., the matrix All poles are located within the unit circle to achieve unbiased estimation of disturbances.

[0096] The state prediction steps are as follows:

[0097] The estimated values ​​of the system state and its covariance:

[0098] ;

[0099] The estimated value of the predicted state at this time and its estimation error covariance Current measurements have not yet been taken into account. .

[0100] The measurement update steps are as follows:

[0101] In the measurement update process, artificial noise is introduced in the calculation stage of weight expectation to reduce the impact of abnormal residuals;

[0102] (a) First, perform initialization:

[0103] , ;

[0104] in, It is a moment The expected scale weights It is a moment The mean of the perturbation residuals;

[0105] (b) Generate L artificial noise samples, where the l-th noise sample is:

[0106] ;

[0107] Among them, the variance of artificial noise Selected as:

[0108] ;

[0109] Where m is the measurement dimension; As a proportional coefficient, a limit can be set to ensure that the disturbance is controllable; Represents the trace operation of a matrix; Used to determine the variance of artificial noise, its value is ;

[0110] (c) Perform iteration:

[0111] Calculate the centralized residual :

[0112] ;

[0113] in, The state estimate is given; to improve the stability of the scale weight estimate, artificial noise is added to the centered residual to construct the perturbed residual; after adding artificial noise, the l-th perturbed residual is... for:

[0114] ;

[0115] For the l-th disturbance residual Constructing an auxiliary statistical matrix ,

[0116]

[0117] And based on this, the expected value of the scale weights is calculated; where, It is a latent variable The second-order moment statistics matrix;

[0118] For the l-th disturbance residual Calculate the expected value of its scale weights:

[0119] ;

[0120] The noise enhancement weights are obtained by averaging the expectations of the L scale weights. The value of the i-th diagonal element is:

[0121] ;

[0122] use Update hidden variables and its covariance ,

[0123] ;

[0124] in, It is a latent variable The auxiliary gain matrix; the hidden variables are obtained. posterior mean ,

[0125] ;

[0126] in, , , It is a normal distribution The probability density function, It is a normal distribution The cumulative distribution function;

[0127] state Covariance renew:

[0128] ;

[0129] in, This is the gain matrix for state updates; these update results can be used as input for the next iteration; finally, after... After the next iteration, the updated posterior state estimate can be obtained. and covariance matrix .

[0130] Step 3 is implemented as follows:

[0131] By recursively executing step 2, external wind disturbances can be estimated and compensated by the interference observer; unknown statistical information of the biased measurement noise caused by non-line-of-sight errors and multipath effects is identified using a variational Bayesian approximation; the UAV state is recursively estimated based on noise-enhanced scale weights, the measurement update step is executed iteratively, and the UAV position estimation result is output in real time; specifically, for UAV position state estimation at each moment and its measurement error covariance The output is:

[0132] ;

[0133] In this way, non-Gaussian noise and skewed interference in ultra-wideband measurements can be suppressed simultaneously, effectively reducing the impact of single abnormal measurements on positioning results and improving the numerical stability of the filtering process.

[0134] The following specific embodiments illustrate the implementation principle of the present invention:

[0135] The experiment focuses on verifying the estimation of external wind interference, suppression of non-Gaussian measurement errors, and the impact of artificial noise excitation on the stability of scale weight estimation during the ultra-wideband fine positioning process of UAVs. The experimental object is a quadcopter UAV flying in three-dimensional space, and the ultra-wideband positioning system consists of 8 fixed ultra-wideband anchor points and tags on the UAV. The simulation time is 60s, and the sampling period is 0.05s.

[0136] Ultra-wideband anchor point setting: Ultra-wideband anchor points are arranged at the vertex positions of the 8m×8m×4m three-dimensional space, with a total of 8 anchor points, located at (0,0,0), (8,0,0), (0,8,0), (8,8,0), (0,0,4), (8,0,4), (0,8,4), (8,8,4).

[0137] Trajectory settings: The UAV reference trajectory adopts a three-dimensional circular trajectory located inside the ultra-wideband anchor point space, with the trajectory center at (4.0m, 4.0m, 2.0m) and the circular radius at 2.4m.

[0138] Non-line-of-sight offset setting: To simulate the positive offset ranging error caused by non-line-of-sight propagation and multipath propagation in complex indoor environments, a phased offset was added to some ultra-wideband anchor points in two time periods. In the 12–24s interval, positive offset ranging errors were added to the 2nd and 5th anchor points; in the 32–43s interval, positive offset ranging errors were added to the 3rd and 7th anchor points. This setting was used to verify the improvement in scale weight estimation stability and positioning accuracy of the proposed artificial noise excitation method when non-line-of-sight offset and heavy-tailed anomaly ranging coexist.

[0139] Artificial noise excitation settings: Artificial noise excitation only applies to the scale weight expectation calculation in the variational measurement update phase and does not directly change the actual ultra-wideband ranging value. L=100 artificial noise samples are generated at each time step, and the variational iteration count is 3. The method without artificial noise excitation directly uses the original centered residual to calculate the scale weight; the method with artificial noise excitation calculates the scale weight based on L perturbation residuals and then averages them to obtain the noise-enhanced scale weight.

[0140] Test results are as follows Figures 3 to 6 As shown.

[0141] Figure 3 The estimated trajectories of a UAV in 3D space without and with artificial noise excitation are presented. It can be seen that, after introducing artificial noise excitation, the estimated trajectory exhibits smoother convergence characteristics under non-line-of-sight bias intervals and heavy-tailed anomaly ranging. This result demonstrates that the proposed method can achieve stable recursive estimation of the UAV's 3D position under conditions of complex non-Gaussian measurement noise and external wind interference. Compared with the method without excitation, artificial noise excitation improves the scale weight estimation in the measurement update stage, making the impact of anomaly ranging on state correction more gradual, thereby enhancing the stability of the estimation.

[0142] Figure 4 The 3D position estimation errors are presented under two conditions: with and without artificial noise excitation. The gray area represents the artificially set non-line-of-sight offset ranging interval. It can be seen that during normal measurement, the error levels of the two methods are relatively similar; however, within the non-line-of-sight offset interval, the position estimation error fluctuates significantly due to the persistently large distances measured at some anchor points. Introducing artificial noise excitation reduces the peak error and suppresses abrupt changes in the error curve. These results indicate that the advantage of artificial noise excitation is mainly reflected in the intervals with strong abnormal residuals and skewed ranging errors.

[0143] Figure 5The average expected value of the scale weights for all ultra-wideband ranging channels is given. It can be seen that, compared with the unexcited method, the average scale weight change obtained by the artificial noise excitation method is generally smoother, indicating that calculating the noise-enhanced scale weights by artificially perturbed residuals can reduce the impact of a single anomalous measurement on the overall reliability assessment.

[0144] Figure 6 The scale weight variations of typical anchor point ranging channels affected by non-line-of-sight multipath bias are presented, including anchor points 2, 5, 3, and 7. (Compared to...) Figure 5 The average scale weights are different. Figure 6 Directly observing the weight changes of specific abnormal anchor points can avoid the dilution effect of normal anchor points on the average value. Results show that the proposed method can adaptively reduce the weights of ranging channels affected by non-line-of-sight conditions, and uses artificial noise excitation to make the channel-level scale weight changes more stable, thereby reducing the impact of abnormal anchor points on the positioning results.

[0145] Based on the above experimental results, it can be seen that in the range of non-line-of-sight bias and heavy-tailed anomaly ranging, the artificial noise excitation method usually exhibits more stable scale weights and more stable posterior position estimation.

[0146] The contents not described in detail in this specification are existing technologies known to those skilled in the art.

[0147] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for ultra-wideband fine positioning of unmanned aerial vehicles (UAVs) based on artificial noise excitation, characterized in that, Includes the following steps: Step 1: Establish a nonlinear positioning system that includes a UAV motion state model and an ultra-wideband measurement model, and introduce a slant-t distribution to describe the statistical characteristics of the ultra-wideband measurement error. Step 2: Design a variational Bayesian filtering algorithm to introduce artificial noise into the measurement residuals during the measurement update process, construct residual neighborhood samples, and average the expectation of the scale weight matrix calculated from the residual samples to obtain a noise-enhanced scale weight matrix, which is used to adaptively correct the measurement update process. Step 3: Use the noise-enhanced scale weight matrix to perform recursive estimation of the UAV state, cyclically execute the measurement update step, and output the UAV position estimation result in real time; The measurement update steps are as follows: During the measurement update process, L artificial noise samples are adaptively generated based on the current measurement residuals and residual covariance. The noise is: ； ； Where m is the measurement dimension; This is the proportionality coefficient; Represents the trace operation of a matrix; Used to determine the variance of artificial noise, its value is ; Calculate the centralized residual : ； in, This is the state estimate. The artificial noise samples are superimposed onto the centered measurement residual to obtain the first... The disturbance residual for: ； According to each disturbance residual Constructing auxiliary statistical matrix And calculate the first The scaling weight matrix corresponding to each perturbation residual Expectations Wherein, the scale weight matrix Used to adjust the degree of influence of each measurement dimension on the measurement update according to the magnitude of the perturbation residual; its expected value is... The i-th diagonal element is: ； The noise enhancement scale weight matrix is ​​obtained by averaging the expectations of the L scale weights. : ； The measurement noise covariance is corrected based on the noise enhancement scale weight matrix, and the measurement update of the UAV state estimate and covariance matrix is ​​completed.

2. The method for ultra-wideband fine positioning of unmanned aerial vehicles based on artificial noise excitation according to claim 1, characterized in that, Step 1 includes: The nonlinear positioning system, which includes a UAV motion state model and an ultra-wideband measurement model, is as follows: ； in, It is the system status; and It is a control input; It is external wind disturbance; These are ultra-wideband measurements; It is process noise and follows a Gaussian distribution. ; It is measurement noise, which follows a slanted t-distribution; This is the state transition matrix; and To control the input matrix; The interference input matrix; Assign a matrix to the process noise; Measurement matrix; ultra-wideband non-Gaussian measurement noise The amount Follows a slanted t-distribution: ； in, Here, is a location parameter, representing the mean. is a scale parameter, representing the standard deviation; Here, represents the shape parameter, indicating the degree of skewness; This is a parameter representing the degrees of freedom, reflecting kurtosis characteristics.

3. The method for ultra-wideband fine positioning of unmanned aerial vehicles based on artificial noise excitation according to claim 2, characterized in that, The positioning accuracy of drones can also be affected by wind disturbance; wind disturbance Modeled as: ； From time At the time Increase in wind disturbance satisfy ;in, It is the sampling time interval between two adjacent moments. It indicates time. The parameter corresponding to the upper bound of the rate of change of disturbance.

4. The UAV ultra-wideband fine positioning method based on artificial noise excitation according to claim 3, characterized in that, Step 2 includes: for The filter algorithm consists of three steps: interference estimation, state prediction, and measurement update.

5. The UAV ultra-wideband fine positioning method based on artificial noise excitation according to claim 4, characterized in that, The interference estimation steps are as follows: Design the following observer: ； in, The observer state estimate, This is an estimate of the interference. Define the augmented state transition matrix as the predicted output of the system. Gain matrix Augmented output matrix as follows: ； State estimation gain matrix and interference estimation gain matrix The selection needs to ensure It is stable, i.e., the matrix All poles are located within the unit circle to achieve unbiased estimation of disturbances.

6. The method for ultra-wideband fine positioning of unmanned aerial vehicles based on artificial noise excitation according to claim 5, characterized in that, The state prediction steps are as follows: The estimated values ​​of the system state and its covariance: ； The estimated value of the predicted state at this time and its estimation error covariance Current measurements have not yet been taken into account. .

7. The method for ultra-wideband fine positioning of unmanned aerial vehicles based on artificial noise excitation according to claim 6, characterized in that, Step 3 is implemented as follows: By recursively executing step 2, external wind disturbances are estimated and compensated by the disturbance observer; unknown statistics of the biased measurement noise caused by non-line-of-sight errors and multipath effects are identified using a variational Bayesian approximation. The UAV state is recursively estimated based on the noise-enhanced scale weight matrix, and the measurement update step is executed iteratively, with the UAV position estimation result output in real time; specifically, for UAV position state estimation at each moment and its measurement error covariance The output is: ； In this way, non-Gaussian noise and skewed interference in ultra-wideband measurements can be suppressed simultaneously, effectively reducing the impact of single abnormal measurements on positioning results and improving the numerical stability of the filtering process.

Images