Unmanned aerial vehicle high-precision positioning method and system

Abstract

The invention relates to the technical field of unmanned aerial vehicle positioning, and provides an unmanned aerial vehicle high-precision positioning method and system. The method comprises the following steps: judging that the unmanned aerial vehicle is in an industrial plume interference state by monitoring perception data quality information of a visual perception unit and a laser perception unit carried by the unmanned aerial vehicle, activating a UWB distance measurement unit, obtaining distance measurement information between the unmanned aerial vehicle and a plurality of preset UWB anchor points, and further determining UWB positioning information of the unmanned aerial vehicle; generating position information of the unmanned aerial vehicle according to the UWB positioning information and sensing data of the visual sensing unit and the laser sensing unit; and performing navigation control on the unmanned aerial vehicle according to the position information. Therefore, the unmanned aerial vehicle can still obtain stable and reliable high-precision positioning in a severe environment, and the risks that the flight attitude is out of control, the flight line is deviated and even equipment is impacted due to positioning misalignment are avoided.

Description

Technical Field

[0001] This application relates to the field of unmanned aerial vehicle (UAV) positioning technology, and more specifically, to a high-precision positioning method and system for UAVs. Background Technology

[0002] In modern industrial production, automated inspection using drones has become an important trend in equipment inspection at large chemical plants. Its core task is to accurately inspect and locate faults in pipelines, valves, and storage tanks within complex plant areas. In open areas, drones can achieve centimeter-level positioning via satellite navigation; however, inside plant areas, tall equipment and dense pipelines often block satellite signals and trigger multipath effects, leading to a significant decrease in positioning accuracy.

[0003] To address this challenge, drones typically employ multi-source information fusion positioning systems. When satellite signals are limited, the system automatically switches to a fusion scheme of visual inertial odometry (VIO) and lidar. VIO estimates its own motion using cameras and an inertial measurement unit (IMU), while lidar constructs a real-time point cloud through scanning and matches it with a pre-set high-precision 3D map to correct its position and suppress drift. However, in actual operations, when a drone approaches a large cooling tower, the periodically emitted dense water vapor forms a white plume with severe optical interference. The laser beam is heavily scattered and absorbed as it penetrates the steam, resulting in a weak true echo signal. Simultaneously, the radar receives numerous false echoes from water droplets, generating noisy ghost point clouds. This drastically reduces or even disables the matching accuracy between the real-time point cloud and the pre-set map.

[0004] Meanwhile, steam also severely interferes with the vision system: reduced visibility leads to blurred images and difficulty in feature point extraction, causing VIO motion estimation errors to accumulate rapidly and highlighting drift issues. When both the lidar and vision systems are interfered with, the fusion positioning system struggles to distinguish between real environmental features and false data introduced by steam. The system may provide completely off-target global corrections based on erroneous point clouds, causing drastic changes in UAV positioning, deviations from the flight path, and even collisions or crashes, seriously threatening inspection missions and equipment safety.

[0005] To address the aforementioned issues, existing technologies urgently need improvement. Summary of the Invention

[0006] This application discloses a high-precision positioning method and system for unmanned aerial vehicles (UAVs), which aims to solve the problem that the positioning accuracy of UAV multi-source information fusion positioning systems drops sharply or even fails due to the deterioration of sensor data quality and the introduction of false information in industrial plume interference environments.

[0007] The technical solution of this application is as follows: In a first aspect, this application discloses a high-precision positioning method for a UAV, wherein the UAV is equipped with a visual sensing unit, a laser sensing unit, and a UWB ranging unit; the method includes: Monitor the quality of perception data from the visual and laser perception units on the drone; based on the perception data quality information, determine that the drone is in a state of industrial plume interference and generate an interference confirmation signal; The UWB ranging unit is activated based on the interference confirmation signal to obtain ranging information between the UAV and multiple preset UWB anchor points. Based on the ranging information and the location information of the preset UWB anchor points, the UWB positioning information of the UAV is determined; The location information of the UAV is generated by fusing the UWB positioning information, the perception data of the visual perception unit and the laser perception unit; the UAV is then used for navigation and control based on the location information.

[0008] Furthermore, in this high-precision positioning method for unmanned aerial vehicles, the visual perception unit includes a visual inertial odometry (VIO), and the laser perception unit includes a lidar (LiDAR); the visual inertial odometry (VIO) includes a camera and an inertial measurement unit (IMU). The monitoring of the perception data quality information of the visual perception unit and laser perception unit carried by the UAV includes: determining the number of identifiable feature points in the visual image, the tracking stability of feature points between consecutive image frames, and the local contrast of the visual image based on the visual image acquired in real time by the camera; at the same time, acquiring the point cloud density of the laser perception unit, the matching residual between the point cloud and the pre-built map, and the proportion of false echoes. Based on the perceived data quality information, it is determined that the drone is in an industrial plume interference state, including: if the number of identifiable feature points, tracking stability, local contrast, and point cloud density are all lower than the corresponding thresholds; and if the matching residual and the proportion of false echoes are both higher than the corresponding thresholds, then the drone is determined to be in an industrial plume interference state.

[0009] More specifically, in some implementations, the tracking stability of feature points between consecutive image frames is determined as follows: the average pixel displacement of the tracked feature point between adjacent image frames is calculated; the standard deviation and / or variance of the pixel displacement of the feature point are calculated to quantify its jitter level; when the average pixel displacement exceeds a preset displacement threshold, and / or the standard deviation or variance exceeds a preset jitter threshold, the tracking stability of the feature point is determined to be lower than a preset stability threshold.

[0010] Preferably, in this high-precision UAV positioning method, the UAV's position information is fused and generated based on an extended Kalman filter. The fusion of UAV position information based on UWB positioning information, visual perception unit, and laser perception unit sensing data includes: in the extended Kalman filter, using UWB positioning information as position observations, the output information of the visual inertial odometry (VIO) as visual observations, the matching result of the LiDAR and a pre-built map as point cloud matching observations, and using IMU data for state prediction; in response to interference confirmation signals, dynamically increasing the measurement noise covariance corresponding to the visual observations and point cloud matching observations in the extended Kalman filter; setting the measurement noise covariance corresponding to the UWB positioning information to be lower than the increased measurement noise covariance of the visual observations and point cloud matching observations; calculating the Kalman gain based on the adjusted measurement noise covariances, updating the UAV's state vector using the UWB positioning information, and outputting the UAV's position information.

[0011] Building upon the above, this application further proposes a method for calculating the Kalman gain based on the adjusted measurement noise covariances and updating the UAV's state vector using UWB positioning information to output the UAV's position information. This includes: constructing a measurement matrix based on the correspondence between the UWB positioning information and the position state in the state vector; calculating the Kalman gain based on the measurement matrix, the predicted covariance matrix of the state vector, and the measurement noise covariance matrix corresponding to the UWB positioning information; correcting the state vector based on the difference between the Kalman gain and the predicted state value from the UWB positioning information to complete the measurement update; and extracting the position state information from the updated state vector and outputting it as the UAV's position information.

[0012] As an optional solution, in this high-precision positioning method for UAVs, after determining the UWB positioning information of the UAV, the method further includes: determining whether there is a systematic deviation in the UWB ranging unit; if so, performing a correction operation on the UWB positioning information; wherein, determining whether there is a systematic deviation in the UWB ranging unit includes: calculating the estimated position value of the UAV based on multiple preset UWB anchor point combinations; calculating the difference between each estimated position value; if the difference exceeds a preset difference threshold, then determining that there is a systematic deviation in the UWB ranging unit.

[0013] In one implementation, performing a UWB positioning information correction operation includes: predicting the relative displacement of the UAV within a preset time window based on the angular velocity and linear acceleration collected by the inertial measurement unit (IMU), and obtaining the inertial predicted displacement; acquiring the position change measured by the UWB ranging unit within the preset time window as the UWB measured displacement; comparing the inertial predicted displacement with the UWB measured displacement, and if the difference between the two continuously exceeds a preset error threshold within multiple consecutive time windows, then confirming that the UWB ranging unit has a systematic deviation; and adaptively correcting the ranging information based on the systematic deviation to obtain the corrected UWB positioning information.

[0014] In another implementation, adaptive correction of ranging information based on systematic bias includes: modeling the systematic bias as a time-varying bias vector; using an extended Kalman filter to estimate the bias vector as an extended component of the state variable in real time to obtain an estimated bias vector; and compensating the ranging information at the current moment based on the estimated bias vector to obtain the corrected ranging information.

[0015] To enhance functionality, this high-precision UAV positioning method determines the UWB positioning information of the UAV based on ranging information and the position information of preset UWB anchor points. This includes: using trilateration or least squares methods to solve the ranging information from at least three non-collinear preset UWB anchor points to obtain the preliminary UWB positioning coordinates of the UAV; and then filtering and smoothing the preliminary UWB positioning coordinates to obtain the UWB positioning information of the UAV.

[0016] Secondly, this application also discloses a high-precision positioning system for unmanned aerial vehicles (UAVs). The UAV is equipped with a visual sensing unit, a laser sensing unit, and a UWB ranging unit. The system includes: a monitoring module for monitoring the quality of sensing data from the visual and laser sensing units on the UAV; determining that the UAV is under industrial plume interference based on the sensing data quality information and generating an interference confirmation signal; a ranging module for activating the UWB ranging unit based on the interference confirmation signal and acquiring ranging information between the UAV and multiple preset UWB anchor points; a positioning module for determining the UWB positioning information of the UAV based on the ranging information and the position information of the preset UWB anchor points; and a fusion and control module for fusing and generating the UAV's position information based on the UWB positioning information and the sensing data from the visual and laser sensing units; and performing navigation control on the UAV based on the position information. Beneficial effects

[0017] The high-precision positioning method for unmanned aerial vehicles (UAVs) disclosed in this application innovatively introduces a real-time monitoring mechanism for the quality of visual and laser sensing data, based on the UAV's integration of a visual sensing unit, a laser sensing unit, and a UWB ranging unit. When the system determines that the UAV is under industrial plume interference, it can intelligently generate an interference confirmation signal and activate the UWB ranging unit accordingly. The UWB ranging unit has strong penetration and anti-interference capabilities under industrial plume interference, enabling it to acquire reliable ranging information and thus determine the UAV's UWB positioning information. Subsequently, the method fuses the UWB positioning information with the sensing data from the visual and laser sensing units, which are affected by interference but still have some reference value, to generate high-precision position information for the UAV, and uses this information for navigation control.

[0018] Compared to existing technologies, the solution presented in this application effectively addresses the technical challenge of drastically reduced positioning accuracy or even complete failure of traditional visual inertial odometry (VIO) and lidar (LiDAR) systems in complex industrial environments such as chemical plants when drones encounter interference from industrial plumes like cooling tower steam. This is caused by reduced visibility, loss of feature points, and false echoes. By introducing UWB ranging as a key supplement and dynamically switching or enhancing the weight of UWB when interference occurs, this application ensures that drones can still obtain stable, reliable, and high-precision positioning even in harsh environments. This avoids the risks of flight attitude loss, deviation from the flight path, or even collision with equipment due to inaccurate positioning, greatly improving the safety, reliability, and operational efficiency of drones in industrial inspection tasks. Attached Figure Description

[0019] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a flowchart illustrating the steps of the high-precision positioning method for unmanned aerial vehicles (UAVs) disclosed in an embodiment of the present invention. Figure 2 This is a schematic diagram of the structure of the high-precision positioning system for unmanned aerial vehicles disclosed in an embodiment of the present invention. Detailed Implementation

[0021] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which these embodiments belong; the terminology used herein and in the specification of the application is for the purpose of describing particular embodiments only and is not intended to limit these embodiments; the terms "comprising" and "having," and any variations thereof, in the specification of these embodiments and the foregoing drawings, are intended to cover non-exclusive inclusion. The terms "first," "second," etc., in the specification of these embodiments and the foregoing drawings are used to distinguish different objects, not to describe a particular order.

[0022] The implementation details of the technical solution in this embodiment are described in detail below: In modern industrial inspections, especially in complex environments such as chemical plants, high-precision positioning of drones faces severe challenges. Traditionally, drones rely on satellite navigation systems for positioning. However, in areas with towering reaction towers, huge storage tanks, and intricate pipeline networks, satellite signals are easily blocked or multipath effects occur, leading to a significant decrease in positioning accuracy and failing to meet the needs of refined inspections. To address this issue, existing technologies typically employ a positioning scheme that integrates Visual Inertial Odometry (VIO) with LiDAR. VIO estimates the drone's attitude and relative position changes, while LiDAR constructs a 3D point cloud map and matches it with a pre-built map to correct for accumulated drift in VIO. However, when drones enter areas with industrial plumes (such as water vapor emitted from cooling towers), the dense plumes severely interfere with LiDAR and visual sensors. This results in LiDAR receiving a large number of false echoes, noisy point cloud data, and map matching failure. Simultaneously, visual image contrast decreases, feature points are difficult to extract and track, and VIO system drift accumulates rapidly. Under such dual interference, the multi-source information fusion positioning system cannot effectively identify and filter out false measurement data, resulting in drastic instability in positioning results, which seriously threatens the flight safety and mission success of UAVs.

[0023] To address this, this application proposes a high-precision positioning method for unmanned aerial vehicles (UAVs), wherein the UAV is equipped with a visual sensing unit, a laser sensing unit, and a UWB ranging unit; for example... Figure 1 As shown, the method includes: S101, monitor the perception data quality information of the visual perception unit and laser perception unit carried by the UAV; based on the perception data quality information, determine that the UAV is in an industrial plume interference state, and generate an interference confirmation signal; S102, activate the UWB ranging unit according to the interference confirmation signal, and obtain the ranging information between the UAV and multiple preset UWB anchor points; S103, Based on the ranging information and the location information of the preset UWB anchor point, determine the UWB positioning information of the UAV; S104, Based on the UWB positioning information, the perception data of the visual perception unit and the laser perception unit, the location information of the UAV is generated by fusing them; and the UAV is navigated and controlled based on the location information.

[0024] This application aims to provide a UAV solution that can achieve high-precision positioning even in industrial plume interference environments. The visual sensing unit typically refers to a sensor capable of acquiring environmental images or video, such as a camera or infrared camera, which provides visual information for positioning and environmental perception. The laser sensing unit typically refers to a lidar system, which measures distance and constructs a 3D point cloud of the environment by emitting and receiving laser beams. The UWB ranging unit is a device that uses ultra-wideband technology for high-precision ranging, characterized by strong multipath resistance and good penetration, providing reliable ranging data even in complex environments. The pre-deployed UWB anchor point is a UWB base station pre-positioned at a specific location for communication and ranging with the UWB ranging unit on the UAV.

[0025] In practical implementation, the UAV first needs to monitor the quality of the perception data from its onboard visual and laser sensing units. For example, data quality can be assessed by analyzing indicators such as the sharpness of visual images, the number of feature points, and the density and noise level of the LiDAR point cloud. When these quality indicators significantly decrease, it indicates that the UAV may have entered an area affected by industrial plume interference. At this point, the system determines that the UAV is in an industrial plume interference state and generates an interference confirmation signal. Once the interference confirmation signal is generated, the system activates the UWB ranging unit. The UWB ranging unit then begins operating, acquiring ranging information between the UAV and multiple preset UWB anchor points. This ranging information represents the precise distance between the UAV and each anchor point. Next, based on this ranging information and the position information of the preset UWB anchor points, the UWB positioning information of the UAV can be determined. For example, trilateration or polygonal positioning methods can be used, calculating the precise coordinates of the UAV in space using distance information provided by at least three non-collinear UWB anchor points. Subsequently, the obtained UWB positioning information is fused with the perception data from the visual and laser sensing units to generate the UAV's position information. During the fusion process, UWB positioning information has higher reliability in industrial plume interference environments, therefore it can be given higher weight in the fusion algorithm. Finally, based on the fused position information, the UAV is used for navigation control. This includes adjusting the UAV's flight attitude, speed, and heading to enable it to fly stably along a preset path and perform inspection tasks.

[0026] This application presents a high-precision UAV positioning method that effectively solves the failure problem of traditional vision and laser positioning systems in such environments by introducing a UWB ranging unit as the primary positioning method in industrial plume interference environments. When the quality of the perception data from the visual and laser sensing units deteriorates, and the system determines that the UAV is under industrial plume interference, it can promptly activate the UWB ranging unit. Leveraging its strong anti-interference capability, the system acquires reliable ranging information and determines UWB positioning information. Subsequently, the UWB positioning information is fused with the interfered visual and laser sensing data. By rationally adjusting the weights of each sensor's data, high-precision UAV position information can still be output even in harsh environments. This method avoids the risks of flight instability, deviation from the flight path, or even crashes caused by positioning failure, significantly improving the operational safety and mission success rate of UAVs in complex industrial environments. Compared to existing technologies that rely solely on visual and laser sensors, this application demonstrates significant robustness and accuracy advantages in industrial plume interference scenarios, providing reliable technical support for the application of UAVs in special industrial environments.

[0027] Specifically, the above-mentioned high-precision positioning method for UAVs can be implemented in the following ways when monitoring the perception data quality information of the visual perception unit and laser perception unit carried by the UAV, and when determining whether the UAV is in an industrial plume interference state based on the perception data quality information.

[0028] The visual perception unit includes a visual inertial odometry (VIO), and the laser perception unit includes a lidar (LiDAR); the visual inertial odometry (VIO) includes a camera and an inertial measurement unit (IMU).

[0029] The monitoring drone carries a visual perception unit and a laser perception unit to obtain perception data quality information, including: determining the number of identifiable feature points in the visual image, the tracking stability of the identifiable feature points between consecutive image frames, and the local contrast of the visual image based on the visual image acquired in real time by the camera; and simultaneously, obtaining the point cloud density of the laser perception unit, the matching residual between the point cloud and the pre-built map, and the false echo ratio.

[0030] The step of determining that the UAV is in an industrial plume interference state based on the perceived data quality information includes: if the number of identifiable feature points, tracking stability, local contrast, and point cloud density are all lower than the corresponding thresholds; and the matching residual and false echo ratio are all higher than the corresponding thresholds, then the UAV is determined to be in an industrial plume interference state.

[0031] The visual sensing unit can be understood as a collection of sensors used to acquire visual information about the environment, such as a visual inertial odometry (VIO). VIO is a positioning technology that combines a camera and an inertial measurement unit (IMU) to estimate motion using visual features and inertial data. The camera captures visual images of the environment, while the IMU measures the drone's angular velocity and linear acceleration, providing attitude and motion information. The laser sensing unit can be understood as a sensor used to acquire three-dimensional point cloud information about the environment, such as a lidar (LiDAR). LiDAR measures distance by emitting laser beams and receiving reflected signals, thereby constructing a three-dimensional point cloud map of the environment.

[0032] Specifically, when monitoring the quality of perceived data, for visual perception units, quality indicators are analyzed based on the real-time visual images acquired by the camera. The number of identifiable feature points refers to the number of salient points in an image that can be detected by the algorithm and used for tracking or matching. A decrease in this number usually indicates a decline in image quality or blurred environmental features. The tracking stability of feature points across consecutive image frames refers to the degree to which these feature points are accurately tracked between different time frames. Instability may indicate image blurring, excessively fast motion, or drastic environmental changes. The local contrast of a visual image reflects the brightness differences between different areas of the image; low contrast may indicate image blurring or insufficient lighting. For laser perception units, point cloud density is acquired, i.e., the number of points per unit volume or area. Low density may indicate that the laser signal is absorbed or scattered. The matching residual between the point cloud and the pre-built map refers to the error when matching the currently acquired point cloud with a pre-built environmental map. A high residual may indicate that the current point cloud data is distorted or that the environment has changed. The false echo ratio refers to the proportion of signals reflected by non-real objects received by the lidar; a high ratio may be caused by interference such as smoke or water vapor.

[0033] When determining whether a drone is under industrial plume interference, the aforementioned sensor data quality information is considered comprehensively. If the number of identifiable feature points, tracking stability, local contrast, and point cloud density are all below their respective preset thresholds, and the matching residual and false echo ratios are all above their respective preset thresholds, then the drone is considered to be under industrial plume interference. These thresholds can be set based on actual application scenarios and experience to ensure the accuracy of the judgment.

[0034] This application's solution introduces Visual Inertial Odometry (VIO) and LiDAR as the main sensing units and refines the monitoring indicators for the data quality of these sensing units, thereby enabling more accurate identification of industrial plume interference. In industrial plume environments (such as smoke, steam, and dust), visual images typically become blurred, leading to a reduction in the number of identifiable feature points, decreased tracking stability, and reduced local contrast. Simultaneously, the LiDAR laser beam undergoes scattering and absorption when penetrating the plume, resulting in reduced point cloud density and potentially generating numerous false echoes, increasing the matching residual between the point cloud and the pre-built map. By simultaneously monitoring specific quality indicators of these multimodal sensors and setting corresponding thresholds, this application can effectively distinguish industrial plume interference from general environmental changes or sensor malfunctions, improving the accuracy and robustness of interference detection.

[0035] Through the above technical solution, this application provides a more refined and reliable mechanism for judging industrial plume interference. By comprehensively analyzing various perception data quality indicators from visual sensing units (such as the camera and inertial measurement unit (IMU) of visual inertial odometry (VIO) and laser sensing units (such as LiDAR), including the number of identifiable feature points, tracking stability, local contrast, point cloud density, the matching residual between the point cloud and the pre-built map, and the proportion of false echoes, the environmental state of the UAV can be assessed more comprehensively and accurately. This multi-dimensional, multi-modal quality assessment method significantly improves the ability to identify plume interference in complex industrial environments, avoids misjudgments or omissions that may be caused by a single sensor or a single indicator, and provides a solid foundation for subsequent positioning strategy adjustments, thereby ensuring high-precision positioning and safe operation of the UAV in interfered environments.

[0036] Specifically, the tracking stability of the aforementioned feature points across consecutive image frames can be determined as follows: The tracking stability of identifiable feature points across consecutive image frames is determined as follows: Calculate the average pixel displacement of the tracked feature point between adjacent image frames; calculate the standard deviation and / or variance of the average pixel displacement to quantify its jitter level; when the average pixel displacement exceeds a preset displacement threshold, and / or the standard deviation or variance exceeds a preset jitter threshold, it is determined that the tracking stability of the feature point is lower than a preset stability threshold.

[0037] The "tracked feature points" refer to image feature points that are successfully matched and tracked by a feature matching algorithm between consecutive visual image frames acquired by a visual sensing unit (such as a camera in a visual inertial odometry (VIO)). The "adjacent image frames" typically refer to two or more temporally consecutive image frames.

[0038] Specifically, the average pixel displacement of the feature points between adjacent image frames is calculated to measure the overall motion trend of the feature points on the image plane. For example, for each tracked feature point, the Euclidean distance of its pixel coordinates between the current frame and the previous frame can be calculated, and then the average of these distances for all tracked feature points is taken.

[0039] Furthermore, the standard deviation and / or variance of the average pixel displacement are calculated to quantify the jitter or instability during feature point tracking. A larger standard deviation or variance indicates greater displacement variation of the feature point across consecutive frames, and thus, more unstable tracking. For example, a pixel displacement sequence for each feature point across multiple consecutive frames can be calculated, and then the standard deviation or variance of these displacement sequences can be calculated.

[0040] In a preferred embodiment, when the average pixel displacement exceeds a preset displacement threshold, it indicates that the UAV may be undergoing violent movement or a significant change in the visual environment, resulting in a large overall drift of the feature points. When the standard deviation or variance exceeds a preset jitter threshold, it indicates that there is significant randomness or uncertainty in feature point tracking, i.e., poor tracking stability. When any one or both conditions are met, it can be determined that the tracking stability of the feature points is lower than a preset stability threshold. The preset displacement threshold, preset jitter threshold, and preset stability threshold can be set according to actual application scenarios and experience to balance positioning accuracy and interference detection sensitivity.

[0041] This application's solution, through quantitative analysis of the average pixel displacement, standard deviation, and / or variance of feature points across consecutive image frames, enables a more refined assessment of the tracking stability of visual sensing units. In industrial plume interference environments, smoke and steam can cause visual image blurring and reduced contrast, making feature point extraction difficult. Furthermore, the extracted feature points are prone to drift or jitter during tracking across consecutive frames. The aforementioned method can accurately capture this tracking instability caused by interference, thus providing a reliable basis for determining whether a UAV is under industrial plume interference. This quantitative evaluation method makes interference detection more objective and accurate.

[0042] The above technical solution enables precise quantification of the feature point tracking stability of the visual perception unit, avoiding the limitations of evaluation based solely on subjective judgment or a single indicator. This method effectively identifies feature point tracking drift and jitter caused by industrial plume interference, improving the accuracy of determining when the UAV is under industrial plume interference. Consequently, the UWB ranging unit can be activated more promptly and reliably, ensuring that the UAV maintains high-precision positioning capabilities even when the quality of visual and laser perception data deteriorates.

[0043] This application further proposes a method for fusing and generating the location information of the aforementioned UAV, which is based on an extended Kalman filter. The method involves fusing and generating the UAV's location information based on the aforementioned UWB positioning information, the perception data from the aforementioned visual sensing unit, and the perception data from the laser sensing unit, including: in the extended Kalman filter, using the aforementioned UWB positioning information as the location observation value, using the output information of the aforementioned visual inertial odometry (VIO) as the visual observation value, using the matching result of the aforementioned LiDAR and a pre-built map as the point cloud matching observation value, and using the data from the inertial measurement unit (IMU) for state prediction; in response to the aforementioned interference confirmation signal, dynamically increasing the measurement noise covariance corresponding to the aforementioned visual observation value and the aforementioned point cloud matching observation value in the extended Kalman filter; setting the measurement noise covariance corresponding to the aforementioned UWB positioning information to be lower than the increased measurement noise covariance of the aforementioned visual observation value and point cloud matching observation value; calculating the Kalman gain based on the adjusted measurement noise covariances, updating the UAV's state vector using the aforementioned UWB positioning information, and outputting the UAV's location information.

[0044] Specifically, the Extended Kalman Filter (EKF) is a nonlinear state estimation algorithm that approximates nonlinear problems by linearizing the nonlinear system model. It is widely used in multi-sensor data fusion to estimate system states. In this application, the EKF is used to fuse data from a UWB ranging unit, a visual sensing unit, a laser sensing unit, and an inertial measurement unit (IMU) to accurately estimate the UAV's position, velocity, and attitude. Position observations refer to the UAV's UWB positioning information provided by the UWB ranging unit, which has high reliability in industrial plume interference environments. Visual observations refer to the output information of the Visual Inertial Odometry (VIO), typically including the UAV's relative pose estimation. Point cloud matching observations refer to the pose information obtained after matching the LiDAR with a pre-built map. Data from the IMU, including angular velocity and linear acceleration, is used to predict the UAV's motion state, providing a state prediction model for the EKF. In response to interference confirmation signals, the measurement noise covariance of visual and point cloud matching observations in the extended Kalman filter (EKF) is dynamically increased. This means that when industrial plume interference is detected, the system reduces its confidence in the data from the visual and laser sensing units. The measurement noise covariance matrix reflects the uncertainty of sensor measurements. Increasing the diagonal elements of this covariance matrix indicates that the measurement noise of these sensors is considered to be greater, and their data reliability is reduced. Simultaneously, the measurement noise covariance corresponding to UWB positioning information is set to be lower than the increased measurement noise covariance of visual and point cloud matching observations, aiming to increase the weight of UWB positioning information in the fusion process. This means that in interference environments, UWB data is given higher confidence, and its impact on UAV state updates will be more significant. Kalman gain is a key parameter in the EKF, determining the degree to which observations correct for state estimation. By adjusting the various measurement noise covariances, the Kalman gain changes accordingly, allowing the system to dynamically adjust its fusion strategy based on the real-time reliability of the sensors. State vector update refers to the process of correcting the UAV's state vector (including position, velocity, attitude, etc.) using Kalman gain and current observations to obtain a more accurate estimate.

[0045] This application's solution effectively addresses the problem of compromised fusion positioning accuracy caused by degraded data quality from visual and laser sensing units under industrial plume interference by dynamically adjusting the measurement noise covariance of each sensor observation in the extended Kalman filter. Specifically, when an interference confirmation signal is generated, it indicates that visual and laser data are severely affected and their reliability is reduced. At this point, by increasing the measurement noise covariance between visual observations and point cloud matching observations, the extended Kalman filter reduces its trust in these data, thereby minimizing their impact on UAV state updates. Simultaneously, since the UWB ranging unit can still provide relatively reliable positioning information under industrial plume interference, its measurement noise covariance is set to a lower value, allowing UWB positioning information to dominate the fusion process and play a crucial role in updating the UAV state vector. This dynamic adjustment mechanism enables the fusion algorithm to adaptively allocate weights based on the real-time quality of the sensor data, ensuring high-precision positioning even in harsh environments.

[0046] Through the above technical solution, this application can significantly improve the positioning accuracy and robustness of UAVs in industrial plume interference environments. Compared with traditional fixed-weight fusion methods, the dynamic noise covariance adjustment mechanism of this application enables the system to intelligently identify and respond to changes in the quality of sensor data, effectively suppressing errors introduced by interference with visual and laser data. Therefore, even in industrial plume areas with low visibility and complex environments, UAVs can still rely on the dominant role of UWB positioning information, combined with other sensor data, to perform reliable fusion positioning, thereby ensuring the safe and stable operation and high-precision work of UAVs in complex industrial environments.

[0047] In some preferred embodiments, suppose a UAV, while performing an industrial inspection mission, enters the chimney emission area of ​​a chemical plant, where a dense industrial plume exists. At this time, the perceived data quality information from the UAV's onboard visual sensing unit (e.g., Visual Inertial Odometry, VIO) and laser sensing unit (e.g., LiDAR) is detected to have significantly decreased; for example, the number of identifiable feature points in the visual image decreases sharply, the laser point cloud density decreases, and the proportion of false echoes increases. Based on this, the system determines that the UAV is in an industrial plume interference state and generates an interference confirmation signal. In response to this interference confirmation signal, the extended Kalman filter inside the UAV immediately adjusts its parameters. Specifically, the diagonal elements of the measurement noise covariance matrix originally used for matching visual observations and point cloud observations are dynamically increased, for example, from 0.1 to 1.0, indicating a significant decrease in the system's trust in the visual and laser data. Simultaneously, the measurement noise covariance corresponding to the UWB positioning information is maintained at a low level, for example, 0.01, to ensure it has a higher weight in the fusion process. Based on these adjusted measurement noise covariances, the Extended Kalman Filter (EKF) recalculates the Kalman gain. Due to the increased relative weight of UWB positioning information, the Kalman gain will rely more heavily on UWB data to correct the UAV's state vector. For example, if UWB positioning information indicates the UAV is at (X, Y, Z) coordinates, while visual / laser data is significantly biased due to interference, the EKF will prioritize the UWB data and use it to update the UAV's position. Ultimately, the output UAV position information will be primarily driven by UWB positioning information, combined with predictions from the Inertial Measurement Unit (IMU), thus providing high-precision and robust positioning results even under industrial plume interference. This ensures the UAV can accurately complete inspection tasks and avoids collisions or mission failures caused by positioning errors.

[0048] In some embodiments described above in this application, the UAV's position information fusion is performed based on an extended Kalman filter. Specifically, under industrial plume interference, UWB positioning information is given higher weight to ensure robustness of positioning. To more clearly illustrate this crucial update process, this application further details the specific steps of calculating the Kalman gain based on the adjusted noise covariance of each measurement, updating the UAV's state vector using the UWB positioning information, and outputting the UAV's position information.

[0049] The above-mentioned calculation of Kalman gain based on the adjusted measurement noise covariances, and updating of the UAV's state vector using UWB positioning information to output the UAV's position information includes: constructing a measurement matrix based on the correspondence between UWB positioning information and position states in the state vector; calculating Kalman gain based on the measurement matrix, the predicted covariance matrix of the state vector, and the measurement noise covariance matrix corresponding to the UWB positioning information; correcting the state vector based on the difference between the Kalman gain and the predicted state value of the UWB positioning information to complete the measurement update; and extracting position state information from the updated state vector as the UAV's position information for output.

[0050] Specifically, when constructing the measurement matrix, it is designed to map the UAV's state vector (typically containing position, velocity, attitude, etc.) to the observation space of UWB positioning information. For example, if the state vector contains the UAV's position (x, y, z) in three-dimensional space, the measurement matrix will contain terms corresponding to these position components to reflect that UWB positioning information directly provides the UAV's position observation. The aim is to establish a mathematical bridge between the state space and the observation space, enabling the observations to be effectively used for state estimation. The prediction covariance matrix of the state vector is calculated by the extended Kalman filter in the prediction step based on the system dynamics model and the process noise covariance matrix; it quantifies the uncertainty of the predicted state. The measurement noise covariance matrix corresponding to the UWB positioning information reflects the noise level of the UWB measurement. Under industrial plume interference conditions, this matrix is ​​set to a low value to reflect the high reliability of the UWB positioning information. The calculation of the Kalman gain is a core component of the extended Kalman filter, determined based on the prediction covariance matrix, the measurement matrix, and the measurement noise covariance matrix. The role of Kalman gain is to balance the weights between predicted and observed values, ensuring that state updates fully utilize reliable observation information while avoiding over-reliance on noisy observations. In practical applications, state vector correction, i.e., measurement update, is accomplished by multiplying the Kalman gain by the residual between the UWB positioning information and the predicted state value, and then adding this product to the predicted state value. This process effectively integrates UWB positioning information into the UAV's state estimation, correcting for biases in the predicted state. Finally, position state information is extracted from the updated state vector, such as the components corresponding to the UAV's three-dimensional coordinates, as the precise position information of the UAV at the current moment, for subsequent navigation and control.

[0051] This application's solution ensures efficient and accurate integration of UWB positioning information into UAV state estimation by explicitly defining the measurement update step in the extended Kalman filter. Under industrial plume interference, the quality of the perception data from the visual and laser sensing units degrades, and their corresponding measurement noise covariance dynamically increases. However, due to its anti-interference characteristics, the positioning information from the UWB ranging unit is given higher confidence, meaning its measurement noise covariance is set to a lower value. By constructing a precise measurement matrix, a clear correspondence is established between the UWB positioning information and the UAV's state vector. Subsequently, based on the adjusted measurement noise covariances, the Kalman gain is calculated. This gain intelligently balances the difference between UWB observations and state predictions, thereby optimally correcting the state vector. Thus, even in harsh industrial plume interference environments, the UAV's position state can be updated promptly and accurately, avoiding positioning drift or distortion caused by degraded data quality from other sensors.

[0052] Through the above technical solution, this application ensures more accurate and stable UAV position information fusion in industrial plume interference environments. Specifically, by detailing the calculation of Kalman gain and the measurement and update mechanism of state vector, this application enables UWB positioning information to be optimally integrated into the extended Kalman filter. This not only improves positioning accuracy but also enhances the system's robustness in complex environments, effectively avoiding the accumulation of positioning errors that may occur with traditional methods when sensor data quality is compromised, thus providing reliable position assurance for the safe and efficient operation of UAVs.

[0053] Furthermore, this application proposes a scheme for judging and correcting the systematic deviation of the UWB ranging unit after determining the UWB positioning information of the UAV, so as to ensure the accuracy of the UWB positioning information. Specifically, after determining the UWB positioning information of the UAV, the method further includes: judging whether the UWB ranging unit has a systematic deviation; if so, performing a correction operation on the UWB positioning information; wherein, judging whether the UWB ranging unit has a systematic deviation includes: calculating the position estimate of the UAV according to a preset combination of multiple UWB anchor points; calculating the difference between the position estimates; if the differences all exceed a preset difference threshold, judging that the UWB ranging unit has a systematic deviation.

[0054] The UWB ranging unit is a device used to measure the distance between the UAV and preset UWB anchor points. Its ranging accuracy directly affects the positioning result. The systematic bias refers to a regular error exhibited by the UWB ranging unit over a long period or under specific conditions, rather than random noise. The preset multiple UWB anchor point combinations refer to pre-deployed UWB base stations with known location information within the UAV's operating area. These anchor points typically provide ranging data to the UAV using trilateration or polygonal positioning. By selecting different UWB anchor point combinations for positioning calculations, the performance of the UWB ranging unit can be evaluated from multiple perspectives. The position estimate is the UAV's coordinates at a certain moment calculated based on different anchor point combinations. The difference refers to the distance or vector difference between the position estimates calculated by different combinations. The preset difference threshold is an empirical value or an upper limit set according to system accuracy requirements, used to determine whether these position estimates are sufficiently consistent. If these differences all exceed this threshold, it indicates that the ranging data of the UWB ranging unit may have a general, non-random error, i.e., a systematic bias.

[0055] This application's solution introduces a mechanism to assess systematic biases in UWB ranging units, enabling proactive identification of potential error sources in UWB positioning information. When visual and laser sensing units fail due to industrial plume interference, UWB positioning becomes the primary auxiliary method. However, if the UWB ranging unit itself exhibits systematic biases, the positioning information it provides will inevitably contain errors, affecting the accuracy of subsequent fusion positioning. By calculating the estimated position of the UAV based on multiple preset UWB anchor point combinations and calculating the differences between these estimates, the internal consistency of the UWB ranging unit can be effectively detected. When these differences all exceed a preset difference threshold, it indicates that the ranging data from the UWB ranging unit may contain systematic, non-random errors, thus confirming the existence of systematic biases. Once systematic biases are confirmed, subsequent UWB positioning information correction operations are performed, thereby avoiding the direct use of UWB positioning information with systematic biases for fusion positioning and ensuring positioning accuracy and reliability even when visual and laser sensing units are limited.

[0056] Through the above technical solution, this application effectively solves the problem of systematic bias in UWB ranging units, significantly improving the robustness and accuracy of UAV positioning in industrial plume interference environments. In traditional solutions, the systematic bias of UWB ranging units may be ignored, resulting in inaccurate positioning information and affecting the overall fusion positioning performance. This application introduces a judgment mechanism for the systematic bias of UWB ranging units, enabling timely detection and confirmation of such biases, providing a basis for subsequent correction operations. Therefore, even when the visual sensing unit and laser sensing unit are interfered with, the reliability of UWB positioning information can be ensured, avoiding the decrease in positioning accuracy caused by UWB's own errors, thus guaranteeing high-precision navigation and control of UAVs in complex environments.

[0057] In some preferred embodiments, assuming a UAV is performing an inspection task in an industrial area with four UWB anchor points A, B, C, and D, whose locations are known, when the UAV enters an area affected by industrial plume interference, the quality of the perception data from the visual and laser sensing units degrades, and the system activates the UWB ranging unit for assisted positioning. At this time, the system can calculate the estimated position P1, P2, and P3 of the UAV at the current moment using different anchor point combinations (e.g., combination 1: A, B, C; combination 2: A, B, D; combination 3: B, C, D) based on the ranging information obtained by the UWB ranging unit. Subsequently, the differences between these position estimates are calculated, such as the distances between P1 and P2, P1 and P3, and P2 and P3. If these differences (e.g., root mean square error or maximum distance difference) all exceed a preset difference threshold (e.g., 0.5 meters), the system determines that the UWB ranging unit has a systematic bias. Based on this judgment, the system will trigger subsequent UWB positioning information correction operations to eliminate or reduce the impact of this systemic deviation on the positioning accuracy of the UAV.

[0058] Furthermore, this application also proposes a correction operation for the UWB positioning information, including: predicting the relative displacement of the UAV within a preset time window based on the angular velocity and linear acceleration collected by the inertial measurement unit (IMU), and obtaining the inertial predicted displacement; acquiring the position change measured by the UWB ranging unit within the preset time window as the UWB measured displacement; comparing the inertial predicted displacement with the UWB measured displacement, and if the difference between the two continuously exceeds a preset error threshold within multiple consecutive time windows, confirming that the UWB ranging unit has a systematic deviation; and adaptively correcting the ranging information based on the systematic deviation to obtain the corrected UWB positioning information.

[0059] Specifically, an inertial measurement unit (IMU) can be understood as a sensor capable of measuring the angular velocity and linear acceleration of a drone in three-dimensional space. Its data can be used to calculate the drone's motion state and relative displacement. Predicting the drone's relative displacement within a preset time window, thus obtaining the inertial predicted displacement, involves using the angular velocity and linear acceleration data collected by the IMU and performing integration calculations through a kinematic model to calculate the drone's displacement relative to its starting point within a short time. This preset time window can be flexibly set according to the drone's motion characteristics and the system's real-time response requirements to deviations; for example, it can be set to 0.1 seconds to 1 second.

[0060] Furthermore, the position change measured by the UWB ranging unit within the preset time window is obtained as the UWB measured displacement. This refers to the calculation of the actual position change of the UAV within the same preset time window based on the continuous positioning information obtained by the UWB ranging unit within the same preset time window.

[0061] In practical applications, comparing the predicted inertial displacement with the measured UWB displacement aims to cross-verify the high-precision relative positioning capability of the inertial measurement unit (IMU) with the absolute positioning information provided by the UWB ranging unit, utilizing the IMU's high-precision relative positioning capability within a short timeframe. If the difference between the two consistently exceeds a preset error threshold over multiple consecutive time windows, a systematic bias in the UWB ranging unit is confirmed. Here, "multiple consecutive time windows" and "preset error threshold" are key parameters; they together ensure sufficient robustness in judging systematic bias, effectively eliminating interference from instantaneous noise or random errors, and confirming the bias only when it exhibits persistent and systematic characteristics. The preset error threshold can be calibrated according to the positioning accuracy requirements of the actual application scenario; for example, it can be set to 0.1 meters to 0.5 meters. Therefore, adaptive correction of the ranging information based on the systematic bias aims to correct the UWB ranging information once a systematic bias is confirmed, thereby eliminating or reducing the impact of this bias on the final positioning result. Adaptive calibration means that the calibration parameters can be dynamically adjusted according to real-time changes in the deviation to adapt to different working environments and deviation patterns.

[0062] This application's solution overcomes the limitations of relying solely on UWB data to determine systematic biases by introducing a comparison mechanism between Inertial Measurement Unit (IMU) data and UWB-measured displacement. Specifically, the IMU provides high-frequency, high-precision relative motion information with relatively small integration errors over short periods, making it a reliable reference for judging UWB systematic biases. When the relative displacement predicted by the IMU and the relative displacement measured by UWB show a significant and continuous difference over multiple consecutive time windows, this strongly indicates a non-random, persistent systematic bias in the UWB ranging unit, rather than simple random noise. Through this cross-sensor data fusion and verification, the systematic bias of the UWB ranging unit can be more accurately identified and confirmed. Once the bias is confirmed, the original UWB ranging information can be adaptively corrected based on this bias, thereby improving the accuracy and reliability of UWB positioning.

[0063] Through the above technical solution, this application can more robustly and accurately identify systematic deviations of the UWB ranging unit, avoiding misjudgments caused by instantaneous errors or environmental fluctuations. By comparing the high-precision short-time relative displacement information of the inertial measurement unit (IMU) with the measured displacement of the UWB ranging unit, random errors and persistent systematic deviations can be effectively distinguished, thus ensuring that correction operations are triggered only when a truly systematic deviation exists. Furthermore, adaptive correction of the ranging information based on the confirmed systematic deviation can effectively eliminate or reduce long-term drift and cumulative errors in UWB positioning, significantly improving the positioning accuracy and stability of the UAV in complex industrial plume interference environments, and providing more reliable position information for subsequent navigation and control.

[0064] As a specific implementation, assuming the UAV is flying in an area affected by industrial plume interference, the UWB ranging unit may experience systematic biases due to environmental factors. In this case, the UAV's onboard Inertial Measurement Unit (IMU) continuously collects angular velocity and linear acceleration data. Within a preset time window (e.g., 0.5 seconds), the inertial predicted displacement of the UAV is obtained by integrating the IMU data. Simultaneously, the ranging information acquired by the UWB ranging unit within the same time window is used to calculate the UWB measured displacement of the UAV. The system continuously compares these two displacement values. If, within 10 consecutive time windows, the Euclidean distance between the inertial predicted displacement and the UWB measured displacement consistently exceeds a preset error threshold (e.g., 0.2 meters), the system confirms a systematic bias in the UWB ranging unit. Once confirmed, this systematic bias is modeled and used to compensate for subsequent UWB ranging information in real time. For example, the bias is estimated using a Kalman filter and subtracted from the original ranging value, thereby outputting corrected UWB positioning information.

[0065] This application further proposes a scheme for adaptive correction of the ranging information based on the systematic deviation, specifically including: modeling the systematic deviation as a deviation vector that varies with time; using the extended Kalman filter, estimating the deviation vector as an extended component of the state variable in real time to obtain an estimated deviation vector; and compensating the ranging information at the current moment based on the estimated deviation vector to obtain the corrected ranging information.

[0066] Specifically, modeling systematic bias as a time-varying bias vector means abstracting the inherent errors of the UWB ranging unit or the systematic offset caused by the environment into a mathematical quantity that evolves over time in a multi-dimensional space. This bias vector can include components such as distance offset, time synchronization error, or bias caused by multipath effects, and its value will dynamically adjust according to the UAV's operating status, environmental conditions, or the characteristics of the sensor itself. This modeling approach allows for a more accurate description and prediction of bias behavior.

[0067] Specifically, according to the Extended Kalman Filter (EKF), the deviation vector is used as an extended component of the state variable for real-time estimation. Obtaining the estimated deviation vector means that in addition to the basic state variables such as the UAV's position, velocity, and attitude, the EKF state vector also includes a deviation vector representing the systematic deviation of the UWB ranging unit. The EKF utilizes its prediction-update mechanism, combining IMU data for state prediction and updating measurements using UWB ranging information and observation data from other sensors (such as visual and laser sensing units), thereby estimating the complete state vector, including the deviation vector, in real time. In this way, the estimated value of the deviation vector can be continuously corrected and optimized with new observation data, thus achieving real-time tracking of dynamic deviations.

[0068] In practical applications, compensating for the current ranging information based on the estimated deviation vector to obtain corrected ranging information means applying the estimated deviation vector, obtained in real time from the Extended Kalman Filter (EKF), to the original UWB ranging information. Specifically, the corresponding components of the estimated deviation vector can be subtracted from the original ranging value or other forms of correction can be applied to eliminate or reduce the influence of systematic deviations on the ranging results. For example, if the deviation vector contains a distance offset, this offset is subtracted from each UWB ranging value to obtain corrected ranging information that is closer to the true distance.

[0069] This application's solution effectively addresses the problem of traditional static or simple adaptive correction methods struggling to handle dynamic biases by modeling the systematic bias of the UWB ranging unit as a time-varying bias vector and incorporating it into the state estimation framework of the Extended Kalman Filter (EKF). Specifically, the EKF leverages its powerful nonlinear state estimation capabilities, combined with the UAV's motion model and multi-source sensor observation data, to jointly estimate the complete system state in real time, including the UAV's own state and the UWB bias vector. This joint estimation allows the estimated bias vector to be dynamically updated as the UAV moves and the environment changes, ensuring accurate tracking of systematic biases. Therefore, once the accurately estimated bias vector is obtained, it can be used to compensate for the original UWB ranging information, providing more accurate UWB positioning information before data fusion, significantly improving the accuracy and robustness of subsequent fusion positioning.

[0070] Through the above technical solution, this application can achieve dynamic, real-time, and high-precision estimation and compensation for systematic biases in UWB ranging units. Compared to schemes that only determine the existence of biases and perform general corrections, this application significantly enhances the system's adaptability to complex and dynamically changing UWB ranging biases by modeling the biases as time-varying vectors and integrating them into an extended Kalman filter (EKF) for joint estimation. This not only improves the accuracy of UWB positioning information but also provides more reliable UWB observation data for subsequent multi-sensor fusion positioning, thereby further enhancing the long-term stability and reliability of high-precision UAV positioning in complex interference environments such as industrial plumes.

[0071] In some preferred embodiments, assuming a drone is performing an inspection task in an industrial environment, this environment may be subject to factors such as temperature fluctuations, electromagnetic interference, or structural reflections, causing the systematic bias of the UWB ranging unit to be non-constant. For example, when the drone moves from one area to another, changes in ambient temperature may cause slight changes in the propagation speed of the UWB signal, resulting in a distance deviation that drifts slowly over time. In this case, the solution of this application models this time-varying distance deviation as a deviation vector and incorporates it as part of the extended Kalman filter (EKF) state vector. The EKF uses angular velocity and linear acceleration data collected by the drone's inertial measurement unit (IMU) for state prediction, and simultaneously updates measurements by combining observation data from the UWB ranging unit, visual perception unit, and laser perception unit. During the measurement update process, the EKF not only updates the drone's position, velocity, and other states, but also estimates and updates this deviation vector in real time. For example, if the UWB ranging value consistently exceeds the distance predicted by the inertial measurement unit (IMU) by a specific value, the extended Kalman filter (EKF) gradually adjusts the estimated bias vector to reflect this persistent positive bias. Once the bias vector is accurately estimated, it is used to compensate for subsequent UWB ranging information. For instance, if the estimated bias vector indicates a distance offset of +0.1 meters, each original UWB ranging value is subtracted by 0.1 meters to obtain the corrected ranging information. In this way, even if the systematic bias of the UWB ranging unit changes dynamically, the proposed solution can track and compensate for it in real time, ensuring consistently high accuracy of UWB positioning information.

[0072] In some implementations of the above-mentioned high-precision positioning method for UAVs, the determination of the UWB positioning information of the UAV based on ranging information and the location information of preset UWB anchor points can be achieved in the following specific ways.

[0073] Specifically, the above-mentioned determination of the UWB positioning information of the UAV based on the ranging information and the position information of the preset UWB anchor points includes: using the trilateration method or the least squares method to solve the ranging information from at least three non-collinear preset UWB anchor points to obtain the preliminary UWB positioning coordinates of the UAV; and filtering and smoothing the preliminary UWB positioning coordinates to obtain the UWB positioning information of the UAV.

[0074] Trilateration is a method for determining an object's position based on distance measurements. When distance information exists between the UAV and at least three known UWB anchor points, the initial UWB positioning coordinates of the UAV can be obtained by constructing circles (spheres in three-dimensional space) with the UWB anchor points as centers and the distance information as radii. Least squares is an optimization algorithm used to estimate the UAV's position when redundant measurements exist (e.g., more than three UWB anchor points) by minimizing the sum of squared residuals between the measured values ​​and the model's predicted values. This method effectively handles measurement noise and improves positioning accuracy. Pre-determined UWB anchor points refer to UWB transmitting / receiving devices pre-deployed within the UAV's operating area whose precise locations are known. These UWB anchor points are typically set to be non-collinear to ensure the uniqueness and stability of the positioning solution. The initial UWB positioning coordinates refer to the raw position estimate obtained directly through trilateration or least squares methods, which may contain some measurement noise or instantaneous errors. Filtering and smoothing processes aim to eliminate or reduce noise and transient fluctuations in the initial UWB positioning coordinates, thereby obtaining more stable and accurate UWB positioning information. Commonly used filtering methods include Kalman filtering, extended Kalman filtering, particle filtering, or moving average filtering. Their purpose is to optimize the position estimation using continuous measurement data over a time series, thereby improving the robustness of the positioning.

[0075] The proposed solution first uses trilateration or least squares methods to perform geometric calculations or optimization estimations on ranging information from multiple UWB anchor points, thereby quickly obtaining the initial position of the UAV. Subsequently, by filtering and smoothing these initial position coordinates, random noise and instantaneous errors that may be introduced during ranging can be effectively suppressed, resulting in more stable and accurate UWB positioning information. This two-stage processing method ensures that the UWB positioning information not only reflects the current position of the UAV in a timely manner, but also significantly improves accuracy and stability, providing reliable foundational data for subsequent fusion positioning and navigation control.

[0076] The above technical solution provides a specific and efficient method for determining UWB positioning information. Using trilateration or least squares methods, the initial position of the UAV can be calculated quickly and accurately based on the ranging information acquired by the UWB ranging unit. Furthermore, by filtering and smoothing the initial UWB positioning coordinates, the impact of ranging noise and environmental interference on the positioning results can be effectively reduced, significantly improving the stability and accuracy of UWB positioning information. This provides more reliable position data support for high-precision navigation control of UAVs in complex industrial plume interference environments.

[0077] This application also discloses a high-precision positioning system for unmanned aerial vehicles (UAVs), wherein the UAV is equipped with a visual perception unit, a laser perception unit, and a UWB ranging unit; such as Figure 2 As shown, the system includes: The monitoring module 201 is used to monitor the perception data quality information of the visual perception unit and laser perception unit carried by the UAV; based on the perception data quality information, it is determined that the UAV is in an industrial plume interference state, and an interference confirmation signal is generated. The ranging module 202 is used to activate the UWB ranging unit according to the interference confirmation signal and obtain ranging information between the UAV and multiple preset UWB anchor points; The positioning module 203 is used to determine the UWB positioning information of the UAV based on the ranging information and the position information of the preset UWB anchor point; The fusion and control module 204 is used to fuse and generate the location information of the UAV based on the UWB positioning information, the perception data of the visual perception unit and the laser perception unit, and to perform navigation control on the UAV based on the location information.

[0078] This application presents an innovative solution to the pain point of traditional multi-source fusion positioning systems failing under industrial plume interference. Existing technologies suffer a sharp decline in the quality of visual and laser sensor data when exposed to dense steam plumes, leading to unreliable location information and even erroneous positioning, severely threatening UAV flight safety. This application addresses this issue by incorporating a monitoring module to evaluate sensor data quality in real time. Upon detecting industrial plume interference, the ranging module activates the UWB ranging unit. UWB technology, with its strong anti-interference capabilities and high penetration, provides stable ranging information even in harsh environments. The positioning module uses this information to determine high-precision UWB positioning information, which is then intelligently fused with the interfered visual and laser data by the fusion and control module, particularly giving higher weight to UWB positioning information under interference conditions. This mechanism effectively avoids the positioning failure problem of traditional systems under interference, ensuring that UAVs can still obtain high-precision and robust position information in complex industrial environments, thereby significantly improving the safety and mission success rate of UAV operations and demonstrating significant technological advancement.

[0079] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A high-precision positioning method for unmanned aerial vehicles (UAVs), characterized in that, The UAV is equipped with a visual sensing unit, a laser sensing unit, and a UWB ranging unit; the method includes: Monitor the perception data quality information of the visual perception unit and laser perception unit carried by the UAV; based on the perception data quality information, determine that the UAV is in an industrial plume interference state and generate an interference confirmation signal; The UWB ranging unit is activated based on the interference confirmation signal to obtain ranging information between the UAV and multiple preset UWB anchor points; Based on the ranging information and the location information of the preset UWB anchor points, the UWB positioning information of the UAV is determined; The location information of the UWB is generated by fusing the UWB positioning information, the perception data of the visual perception unit and the laser perception unit; the UWB is then used for navigation control based on the location information.

2. The high-precision positioning method for unmanned aerial vehicles according to claim 1, characterized in that, The visual perception unit includes a visual inertial odometry (VIO), and the laser perception unit includes a lidar (LiDAR); the visual inertial odometry (VIO) includes a camera and an inertial measurement unit (IMU). The quality information of the perception data from the visual perception unit and laser perception unit carried by the monitoring drone includes: Based on the visual images acquired in real time by the camera, the number of identifiable feature points in the visual images, the tracking stability of the identifiable feature points between consecutive image frames, and the local contrast of the visual images are determined; at the same time, the point cloud density of the laser sensing unit, the matching residual between the point cloud and the pre-built map, and the false echo ratio are obtained. The step of determining that the drone is in an industrial plume interference state based on the perceived data quality information includes: If the number of identifiable feature points, tracking stability, local contrast, and point cloud density are all lower than the corresponding thresholds; and the matching residual and false echo ratio are all higher than the corresponding thresholds, then the UAV is determined to be in an industrial plume interference state.

3. The high-precision positioning method for unmanned aerial vehicles according to claim 2, characterized in that, The tracking stability of identifiable feature points across consecutive image frames is determined as follows: Calculate the average pixel displacement of the tracked feature points between adjacent image frames; Calculate the standard deviation and / or variance of the average pixel displacement to quantify its jitter level; When the average pixel displacement exceeds a preset displacement threshold, and / or the standard deviation or variance exceeds a preset jitter threshold, the tracking stability of the feature point is determined to be lower than a preset stability threshold.

4. The high-precision positioning method for unmanned aerial vehicles according to claim 3, characterized in that, The fusion process generates the location information of the UAV, which is performed based on an extended Kalman filter. Based on the UWB positioning information, the perception data from the visual sensing unit and the laser sensing unit, the location information of the UAV is fused and generated, including: In the extended Kalman filter, the UWB positioning information is used as the position observation value, the output information of the visual inertial odometry (VIO) is used as the visual observation value, the matching result of the lidar (LiDAR) and the pre-built map is used as the point cloud matching observation value, and the data of the inertial measurement unit (IMU) is used for state prediction. In response to the interference confirmation signal, the measurement noise covariance between the visual observation and the point cloud matching observation in the extended Kalman filter is dynamically increased; The measurement noise covariance corresponding to the UWB positioning information is set to be lower than the increased measurement noise covariance between the visual observation and the point cloud matching observation. Based on the adjusted noise covariance of each measurement, the Kalman gain is calculated, and the UWB positioning information is used to update the state vector of the UAV, outputting the position information of the UAV.

5. The high-precision positioning method for unmanned aerial vehicles according to claim 4, characterized in that, The process involves calculating the Kalman gain based on the adjusted noise covariance of each measurement, updating the UAV's state vector using the UWB positioning information, and outputting the UAV's position information, including: A measurement matrix is ​​constructed based on the correspondence between the UWB positioning information and the position states in the state vector; The Kalman gain is calculated based on the measurement matrix, the prediction covariance matrix of the state vector, and the measurement noise covariance matrix corresponding to the UWB positioning information. The state vector is corrected based on the difference between the Kalman gain and the UWB positioning information and the state prediction value to complete the measurement update; The position state information is extracted from the updated state vector and output as the position information of the UAV.

6. The high-precision positioning method for unmanned aerial vehicles according to claim 4, characterized in that, After determining the UWB positioning information of the drone, the method further includes: Determine whether the UWB ranging unit has a systematic deviation. If so, perform a correction operation on the UWB positioning information. Determining whether the UWB ranging unit has a systematic deviation includes: calculating the estimated position of the UAV based on multiple preset UWB anchor point combinations; calculating the difference between the estimated positions; if the differences all exceed a preset difference threshold, then the UWB ranging unit is determined to have a systematic deviation.

7. The high-precision positioning method for unmanned aerial vehicles according to claim 6, characterized in that, The correction operation for the UWB positioning information includes: Based on the angular velocity and linear acceleration collected by the inertial measurement unit (IMU), the relative displacement of the UAV within a preset time window is predicted, and the inertial predicted displacement is obtained. The position change measured by the UWB ranging unit within the preset time window is obtained as the UWB measured displacement; If the difference between the inertial predicted displacement and the UWB measured displacement continues to exceed a preset error threshold within multiple consecutive time windows, it is confirmed that the UWB ranging unit has a systematic deviation. The ranging information is adaptively corrected based on the systematic bias to obtain the corrected UWB positioning information.

8. The high-precision positioning method for unmanned aerial vehicles according to claim 7, characterized in that, The adaptive correction of the ranging information based on the systematic bias includes: The systematic deviation is modeled as a deviation vector that varies with time; Based on the extended Kalman filter, the deviation vector is used as an extended component of the state variable for real-time estimation to obtain the estimated deviation vector; Based on the estimated deviation vector, the ranging information at the current moment is compensated to obtain the corrected ranging information.

9. The high-precision positioning method for unmanned aerial vehicles according to any one of claims 1-8, characterized in that, The step of determining the UWB positioning information of the UAV based on the ranging information and the location information of the preset UWB anchor point includes: The preliminary UWB positioning coordinates of the UAV are obtained by solving the ranging information from at least three non-collinear preset UWB anchor points using the trilateration method or the least squares method. The initial UWB positioning coordinates are filtered and smoothed to obtain the UWB positioning information of the UAV.

10. A high-precision positioning system for unmanned aerial vehicles (UAVs), characterized in that, The UAV is equipped with a visual sensing unit, a laser sensing unit, and a UWB ranging unit; the system includes: The monitoring module is used to monitor the quality of the perception data from the visual perception unit and the laser perception unit carried by the UAV; based on the quality of the perception data, it determines that the UAV is in an industrial plume interference state and generates an interference confirmation signal. The ranging module is used to activate the UWB ranging unit according to the interference confirmation signal and obtain ranging information between the UAV and multiple preset UWB anchor points; The positioning module is used to determine the UWB positioning information of the UAV based on the ranging information and the position information of the preset UWB anchor point; The fusion and control module is used to fuse and generate the location information of the UAV based on the UWB positioning information, the perception data of the visual perception unit and the laser perception unit, and to perform navigation control on the UAV based on the location information.

Images