Unmanned aerial vehicle uwb enhanced positioning graph optimization method for narrow roadway trolley inspection

Abstract

This invention discloses a method for optimizing the UWB-enhanced positioning map of a vehicle inspecting narrow alleyways. The method includes: hovering the UWB above the initial position of the inspection vehicle and collecting its state vector; obtaining the predicted pose of the inspection vehicle; registering the current environment of the inspection vehicle with historical laser point clouds and updating the vehicle's state vector; hovering the UWB above the mapping point and calculating the distance between the UWB and the inspection vehicle; constructing the comprehensive UWB positioning residual for the current mapping point and the corresponding optimization objective function; and based on the current position of the inspection vehicle, finding a better mapping point and switching the UWB mapping point position until the inspection vehicle completes the inspection in the alleyway, thus optimizing the distribution of mapping points in the alleyway. This method meets the accurate positioning requirements of vehicle inspection in narrow alleyways and can reduce UWB deployment and maintenance costs.

Description

Technical Field

[0001] This invention relates to the field of drone positioning, specifically to a method for optimizing UWB-enhanced positioning maps of drones used for vehicle inspection in narrow alleyways. Background Technology

[0002] In inspection tasks in narrow passages such as mine tunnels, underground utility tunnels, and industrial plants, traditional GPS positioning is unusable due to signal obstruction or loss. Ultra-wideband (UWB) positioning technology, with its high accuracy and multipath resistance, has become an alternative. As an emerging wireless positioning method, UWB has been widely used in indoor and complex environments due to its advantages such as high accuracy, low power consumption, and strong anti-interference capabilities. UWB achieves positioning by transmitting and receiving narrow pulse signals at the nanosecond or even picosecond level, utilizing parameters such as the time difference of arrival (TDOA) and angle of arrival (AOA), theoretically achieving centimeter-level positioning accuracy. However, in the special scenario of narrow passages, UWB positioning technology also encounters many challenges, among which multipath effect is one of the most prominent problems. Surfaces such as walls, floors, and ceilings in narrow passages cause multiple reflections, refractions, and scattering of UWB signals, meaning the signal received by the receiver is not a direct signal from the transmitter, but contains a large number of signal components that have traveled through different paths. These multipath signals interfere with each other, causing errors in the signal arrival time measurement, which in turn seriously affects the accuracy and reliability of UWB positioning. Complex environments such as metal structures and dense obstacles in narrow alleyways can also cause severe multipath effects in UWB signals, leading to accumulated positioning errors, drift, and even failure. Furthermore, the ground inspection vehicle is affected by factors such as wheeled odometer errors and uneven ground during its movement, making it difficult for a single UWB positioning system to meet the robustness requirements of long-term inspections.

[0003] Current solutions rely on the FAST-LIO algorithm, which fuses LiDAR (SLAM) and Inertial Measurement Unit (IMU). However, in degraded environments such as narrow alleyways, LiDAR point cloud features are sparse and highly repetitive, causing the cumulative drift error of SLAM to continuously increase. Although IMU can provide short-term, high-precision motion estimation, its error also increases over time. More importantly, existing methods lack absolute position constraints in long-distance enclosed environments, failing to meet the positioning accuracy requirements for centimeter-level stable device detection. Summary of the Invention

[0004] The purpose of this invention is to provide a method for optimizing the UWB-enhanced positioning map of a drone for vehicle inspection in narrow alleyways. This method involves pre-deploying mapping points, such as key points in narrow alleyways like corners, intersections, and the location of characteristic obstacles, allowing the drone to hover above each mapping point. Each mapping point serves as a dynamic UWB base station, and the lidar of the ground inspection vehicle is combined with the inertial measurement unit to achieve high-precision, interference-resistant positioning enhancement.

[0005] To achieve the above functions, this invention designs an optimization method for UWB enhanced positioning maps of drones used for trolley inspections in narrow alleyways. For the target alleyway, pre-mapped points within the alleyway are used as the drone's hovering points. The following steps S1-S6 are executed to optimize the UWB measurement accuracy of the drone:

[0006] Step S1: For the inspection vehicle located at the initial position, the drone flies to hover above the initial position of the inspection vehicle and collects the current state vector of the inspection vehicle; and based on the current state vector of the inspection vehicle, obtains the predicted pose of the inspection vehicle at the next moment.

[0007] Step S2: For each laser point in the laser point cloud of the environment where the inspection vehicle is located, register the current laser point cloud with the laser point cloud collected in the past, and update the state vector of the inspection vehicle.

[0008] Step S3: Hover the drone above the survey point and calculate the distance between the drone and the inspection vehicle, including measuring the distance and calculating the distance;

[0009] Step S4: Based on the predicted pose of the inspection vehicle, the laser point cloud registration process, and the distance calculation residual between the inspection vehicle and the UAV, construct the comprehensive UAV positioning residual for the current mapping point based on the graph optimization method, and construct the optimization objective function for the comprehensive UAV positioning residual.

[0010] Step S5: Based on the current position of the inspection vehicle, find a better surveying point within the preset range, switch the position of the UAV surveying point, until the inspection vehicle completes the inspection in the tunnel.

[0011] Step S6: Based on the geometric accuracy factor, the geometric accuracy factor is minimized by adjusting the position, density, and number of survey points in the roadway, thus optimizing the distribution of survey points in the roadway.

[0012] Beneficial effects: Compared with the prior art, the advantages of the present invention include:

[0013] Existing technologies require the pre-deployment of multiple fixed UWB anchor nodes within the tunnel. However, narrow tunnels are often cluttered with obstacles such as cables and pipes, leading to severe signal obstruction and limited coverage. Furthermore, if fixed points are damaged (e.g., by collision or displacement), manual recalibration is necessary, resulting in high maintenance costs. To ensure positioning accuracy and robustness, especially in non-line-of-sight (NLOS) areas, dense deployment of anchor points is often required, increasing the number of anchor points needed and thus raising costs. Additionally, re-installation and deployment are necessary to adapt to new areas or paths, further increasing long-term operating costs. This invention eliminates the need for any fixed infrastructure construction within the tunnel; only the UWB module needs to be installed on a drone. Deployment is simple: just hover the drone at the designated key point. This rapid deployment significantly reduces system deployment time and costs. Simultaneously, the centralized equipment on the drone facilitates maintenance, charging, and upgrades from the ground. While the drone itself requires maintenance, it is more centralized and convenient than maintaining fixed points scattered throughout the tunnel, significantly reducing maintenance costs. Drones can be quickly deployed to new critical points, are highly adaptable, and one system can be used in multiple different lanes, significantly improving system utilization and reducing overall cost of ownership. At the same time, this invention can effectively solve the problem of ground vehicle inspection and positioning in narrow lanes, improve inspection efficiency and safety, and provide reliable positioning support for various applications in narrow lane environments. Attached Figure Description

[0014] Figure 1 This is a flowchart of a method for optimizing the UWB enhanced positioning map of a vehicle for inspection in narrow alleyways according to an embodiment of the present invention;

[0015] Figure 2 This is an environmental schematic diagram of the UWB-enhanced positioning map optimization method for a vehicle inspecting narrow alleyways according to an embodiment of the present invention. Detailed Implementation

[0016] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and should not be used to limit the scope of protection of the present invention.

[0017] The present invention provides an optimization method for UAV UWB enhanced positioning maps during vehicle inspection in narrow alleyways. For the target alleyway, pre-mapped survey points within the alleyway are used as hovering points for the UAV, and these survey points serve as dynamic UWB base stations for the UAV. The survey points are located at corners, intersections, and the locations of characteristic obstacles. (Refer to...) Figure 1 Perform the following steps S1-S6 to optimize the UWB measurement accuracy of the UAV:

[0018] Step S1: For the inspection vehicle located at the initial position, the drone flies to hover above the initial position of the inspection vehicle and collects the current state vector of the inspection vehicle; and based on the current state vector of the inspection vehicle, obtains the predicted pose of the inspection vehicle at the next moment.

[0019] Reference Figure 2 , Figure 2 The red dots on the ground are pre-mapped survey points, and the green boxes represent the drones hovering above the inspection vehicle at the initial position.

[0020] The specific steps of step S1 are as follows:

[0021] Step S1.1: Collect the current state vector of the inspection vehicle as follows:

[0022] ;

[0023] in, This represents the state vector of the inspection vehicle. This represents the three-dimensional position coordinates of the inspection vehicle in the world coordinate system; This represents the three-dimensional velocity vector of the inspection vehicle in the world coordinate system; The attitude quaternion of the inspection vehicle in the world coordinate system is represented; the inspection vehicle is equipped with an inertial measurement unit, which includes an accelerometer and a gyroscope; This indicates accelerometer bias. Due to the manufacturing process and environmental factors of the accelerometer itself, its measurement values ​​will have a certain systematic error, i.e., bias. Accurately estimating and compensating for accelerometer bias is crucial for improving the measurement accuracy and positioning accuracy of the inertial measurement unit. This indicates the gyroscope bias, representing the deviation in the gyroscope's measurement value. Gyroscopes are used to measure the angular velocity of an object, but they are also subject to various factors that can cause bias. T indicates transpose.

[0024] Step S1.2: State prediction is the process of predicting the system state based on the measurements from the inertial measurement unit (IMU). Its core is to update the position, velocity, and attitude information in the state vector using the raw measurement data from the IMU. Physically speaking, the position of an object at a given moment is determined by its initial position and the cumulative velocity it has accumulated before that moment. In a discrete-time system, the position of the object at the next moment can be predicted by integrating the velocity. The predicted pose of the inspection vehicle at the next moment, including predicted position, predicted velocity, and predicted attitude, is calculated as follows:

[0025] ;

[0026] ;

[0027] ;

[0028] in, The rate of change of the inspection vehicle's position is used to characterize the predicted position of the inspection vehicle. The rate of change of the inspection vehicle's speed is used to characterize the predicted speed of the inspection vehicle. This represents the three-dimensional velocity vector of the inspection vehicle; This represents the raw measurement value from the accelerometer in the inertial measurement unit. It's the accelerometer bias. This is the attitude rotation matrix. Its function is to transform accelerometer measurements from the inertial measurement unit coordinate system to the world coordinate system. It is the gravity vector; by integrating the actual acceleration, the velocity of the object at the next moment can be predicted. The attitude quaternion change rate is used to characterize the predicted attitude of the inspection vehicle. For attitude quaternions, This represents the raw measurement value of the gyroscope in the inertial measurement unit (IMU), indicating the angular velocity of the IMU in the IMU coordinate system. It is the gyroscope bias, minus the gyroscope bias. More accurate angular velocity can be obtained. Through the By performing integration, the quaternion of the object's attitude at the next moment can be predicted, thereby determining the object's attitude in the world coordinate system.

[0029] Step S2: For each laser point in the laser point cloud of the environment where the inspection vehicle is located, register the current laser point cloud with the laser point cloud collected in the past, and update the state vector of the inspection vehicle.

[0030] The specific steps of step S2 are as follows:

[0031] Step S2.1: Analyze the laser points in the laser point cloud of the environment where the inspection vehicle is located. The nearest neighbor search method is used to search for the nearest neighbor in the reference point cloud, as shown in the following formula:

[0032] ;

[0033] in, It is the normal vector of the reference plane. It is a reference point on a reference plane, where the reference plane refers to a small plane in the reference point cloud (usually the previous frame point cloud or local map) that is fitted by the nearest neighbor of the current laser point and its neighboring points. It is the attitude rotation matrix. It is a translation vector. Indicates laser point Distance to the reference plane;

[0034] Step S2.2: For After weighting, the state vector is updated using an iterative extended Kalman filter (IESKF). The state vector update formula is as follows:

[0035] ;

[0036] in, This represents manifold addition, used to handle nonlinear transformations in the state space. It is the Kalman gain matrix, used to balance the weights of prediction and observation. It is the weighted average. ; This represents the state vector of the inspection vehicle. In the iterative extended Kalman filter framework, the inertial measurement unit (IMU) bias is estimated as part of the state variables. The estimated bias value is updated using the laser point cloud matching results from the observation data, and these estimates are applied for compensation in subsequent IMU data preprocessing. In environments such as long tunnels, the error accumulation effect of the IMU becomes more pronounced due to the long travel distance. Online calibration of the IMU bias can effectively suppress this error accumulation, thereby reducing the positioning drift rate.

[0037] Step S3: Hover the drone above the survey point and calculate the distance between the drone and the inspection vehicle, including measuring the distance and calculating the distance;

[0038] The specific steps of step S3 are as follows:

[0039] Step S3.1: The UAV is equipped with a UWB module, which integrates a UWB ranging model. Based on the two-way time-of-flight (TW-TOF) method, the distance between the UAV hovering above the survey point and the inspection vehicle is calculated. :

[0040] ;

[0041] Where c is the speed of light. , These are the send and receive timestamps for the drone, respectively. , These are the timestamps for sending and receiving the inspection vehicle;

[0042] Step S3.2: Install anchor points at the four wing corners of the drone, and attach corresponding tags to the inspection vehicle. Calculate the distance between the drone and the inspection vehicle. The specific formula is as follows:

[0043] ;

[0044] in, The distance between the drone and the inspection vehicle at timestamp t is calculated, and the z in the subscript represents the anchor point on the drone. This is the transformation matrix from the UAV body to world coordinates. Let be the fixed transformation matrix from the UAV body to the tag at timestamp t. The origin position in the UAV's body coordinate system. Let z be the position of anchor point z in the world coordinate system. For measuring noise.

[0045] Step S4: Based on the predicted pose of the inspection vehicle, the laser point cloud registration process, and the distance calculation residual between the inspection vehicle and the UAV, construct the comprehensive UAV positioning residual for the current mapping point based on the graph optimization method, and construct the optimization objective function for the comprehensive UAV positioning residual.

[0046] Graph optimization is a commonly used backend optimization method in SLAM (Simultaneous Localization and Mapping). It represents the robot's position and the positions of landmarks as nodes in a graph, while sensor measurements such as odometry and UWB ranging are represented as edge constraints. By minimizing the errors of all constraints, the optimal pose estimate is obtained. To overcome the cumulative drift of FAST-LIO in long-distance tunnels, a hovering UWB base station is introduced to provide absolute position constraints, using the mapped points as UWB base stations. The tightly coupled graph optimization framework with UWB and laser IMU is a multi-sensor fusion localization method that achieves high-precision and robust localization by jointly optimizing the pose and UWB base station positions.

[0047] The specific steps of step S4 are as follows:

[0048] Step S4.1: Construct the set of state vectors as follows:

[0049] ;

[0050] in, Represents a set of state vectors. Let N represent the state vector sequence of the inspection vehicle, where N is the total number of state vectors of the inspection vehicle. This represents the sequence of survey points for the UAV, where M is the total number of survey points for the UAV; these points are obtained through pre-survey in the tunnel.

[0051] Step S4.2: Construct the UAV positioning comprehensive residual, specifically as follows:

[0052] ;

[0053] in, This indicates the overall residual of the drone's positioning. This represents the pre-integrated residual of the inertial measurement unit, used to constrain motion between adjacent poses. This represents the laser point cloud matching residual, used to constrain the matching between the laser radar point cloud and the map. The value represents the UWB ranging residual, which constrains the distance between the vehicle and the UAV base station; I represents the inertial measurement unit, L represents the lidar, and U represents the UWB module.

[0054] Step S4.3: Construct the optimization objective function for the UAV positioning integrated residual as follows:

[0055] ;

[0056] in, This represents the pre-integration residual of the inertial measurement unit. This represents the laser point cloud matching residual. The value represents the UWB ranging residual; i represents time i, k represents the k-th laser point, and j represents time j.

[0057] , Let i represent the state vector of the inspection vehicle at time i and time i+1;

[0058] The inertial measurement unit pre-integrated residual Based on the current state vector of the inspection vehicle, the inspection vehicle in two consecutive frames at time i and time i+1... , Predicted pose changes;

[0059] Laser point cloud matching residual The calculation is as follows:

[0060] ;

[0061] in, This represents the pose of the inspection vehicle at time i. This indicates the pose of the inspection vehicle at time j. This represents the relative pose change of the inspection vehicle from time i to time j;

[0062] UWB ranging residual The difference between the calculated distance and the measured distance between the drone and the inspection vehicle is shown below:

[0063] ;

[0064] in, This indicates the calculated distance between the drone and the inspection vehicle. This indicates the measured distance between the drone and the inspection vehicle. Let represent the state vector of the inspection vehicle at time j. This indicates the location of the m-th mapping point of the UAV.

[0065] Step S5: Based on the current position of the inspection vehicle, find a better surveying point within the preset range, switch the position of the UAV surveying point, until the inspection vehicle completes the inspection in the tunnel.

[0066] The specific steps of step S5 are as follows:

[0067] Step S5.1: Based on the current position of the inspection vehicle, find a better survey point within the preset range. If a better survey point exists, proceed to step S5.2; otherwise, proceed to step S5.3.

[0068] The criteria for determining a better survey point is: there exists another survey point whose unobstructed distance from the survey point to the inspection vehicle is less than the unobstructed distance between the current survey point and the inspection vehicle.

[0069] Step S5.2: The UAV flies over a better mapping point and hovers to complete the UAV position switch and update the UAV positioning composite residual;

[0070] Step S5.3: The drone maintains its current position and provides the drone positioning residual for the current position.

[0071] Step S6: Based on the geometric accuracy factor, the geometric accuracy factor is minimized by adjusting the position, density, and number of survey points in the roadway, thus optimizing the distribution of survey points in the roadway.

[0072] The Geometric Dilution of Precision (GDOP) is an indicator that measures the degree to which the geometric distribution of a positioning system amplifies measurement errors. Essentially, it is the condition number of the Jacobian matrix. In UWB positioning systems, ranging errors such as multipath propagation and clock skew are inherent. The core objective of minimizing the GDOP is to optimize base station layout to reduce the amplification effect of the geometric distribution on errors, thereby achieving higher positioning accuracy with the same ranging precision.

[0073] The geometric precision factor is calculated as follows:

[0074] ;

[0075] in, Indicates geometric precision factor, The Jacobian matrix represents the relationship between the survey points and the positions of the inspection vehicle. The number of rows is the number of survey points, and the number of columns is the sum of the dimensions of the three-dimensional position error of the inspection vehicle and the dimension of the time error. Indicates trace; , , These represent the positional errors of the inspection vehicle in the x, y, and z directions, respectively. This is due to time error;

[0076] By adjusting the number and location distribution of survey points in the roadway, as well as the density of survey points at bends, the geometric accuracy factor can be adjusted. Minimize to optimize the distribution of survey points in the tunnel.

[0077] Geometric precision factor The larger the value, the greater the degree to which the positioning error is amplified by the geometric distribution; conversely, The smaller the GDOP value, the higher the positioning accuracy. GDOP=1 indicates an ideal geometric distribution of base stations, such as base stations evenly distributed around the tag; while GDOP=5 indicates a poor geometric distribution of base stations, such as base stations being collinear or coplanar. Therefore, the principle of minimizing GDOP should be followed when setting the hovering position of the drone. In narrow alleys, due to space constraints, a single drone base station may not be able to achieve an ideal spherical uniform distribution, so it is necessary to optimize as much as possible by dividing the alley into several segments, with a set of base stations arranged in each segment. At bends, the base station density should be increased, especially at outer bends, because inner bends may be blocked. Because the drone's attitude is uncertain during flight, it cannot achieve accurate UWB positioning gain during flight. The hovering position should be kept as low as possible to maximize the planar distance of the UWB positioning distance. At the same time, collinear or coplanar situations should be minimized within the inspection range of the vehicle. From the drone's perspective, the base stations should be distributed in different quadrants as much as possible. Ideally, the larger the volume of the tetrahedron formed by the four base stations and the drone, the better. Therefore, the larger the drone's wheelbase and the farther the distance between the four corners, the better the positioning accuracy.

[0078] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited to the above embodiments. Within the scope of knowledge possessed by those skilled in the art, various changes can be made without departing from the spirit of the present invention.

Claims

1. A method for optimizing the UWB-enhanced positioning map of a drone used for vehicle inspection in narrow alleyways, characterized in that, For the target tunnel, using the pre-mapped survey points in the tunnel as the hovering points of the UAV, perform the following steps S1-S6 to optimize the UAV UWB measurement accuracy: Step S1: For the inspection vehicle located at the initial position, the drone flies to hover above the initial position of the inspection vehicle and collects the current state vector of the inspection vehicle; and based on the current state vector of the inspection vehicle, obtains the predicted pose of the inspection vehicle at the next moment. Step S2: For each laser point in the laser point cloud of the environment where the inspection vehicle is located, register the current laser point cloud with the laser point cloud collected in the past, and update the state vector of the inspection vehicle. Step S3: Hover the drone above the survey point and calculate the distance between the drone and the inspection vehicle, including measuring the distance and calculating the distance; Step S4: Based on the predicted pose of the inspection vehicle, the laser point cloud registration process, and the distance calculation residual between the inspection vehicle and the UAV, construct the comprehensive UAV positioning residual for the current mapping point based on the graph optimization method, and construct the optimization objective function for the comprehensive UAV positioning residual. The specific steps of step S4 are as follows: Step S4.1: Construct the set of state vectors as follows: ； in, Represents a set of state vectors. Let N represent the state vector sequence of the inspection vehicle, where N is the total number of state vectors of the inspection vehicle. This represents the sequence of survey points for the UAV, where M is the total number of survey points for the UAV; these points are obtained through pre-survey in the tunnel. Step S4.2: Construct the UAV positioning comprehensive residual, specifically as follows: ； in, This indicates the overall residual of the drone's positioning. This represents the pre-integration residual of the inertial measurement unit. This represents the laser point cloud matching residual. The value represents the UWB ranging residual; I represents the inertial measurement unit, L represents the lidar, U represents the UWB module; i represents time i, k represents the k-th laser point, and j represents time j. Step S4.3: Construct the optimization objective function for the UAV positioning integrated residual as follows: ； in, This represents the pre-integration residual of the inertial measurement unit. This represents the laser point cloud matching residual. This represents the UWB ranging residual; , Let i represent the state vector of the inspection vehicle at time i and time i+1; Laser point cloud matching residual The calculation is as follows: ； in, This represents the pose of the inspection vehicle at time i. This indicates the pose of the inspection vehicle at time j. This represents the relative pose change of the inspection vehicle from time i to time j; UWB ranging residual The difference between the calculated distance and the measured distance between the drone and the inspection vehicle is shown below: ； in, This indicates the calculated distance between the drone and the inspection vehicle. This indicates the measured distance between the drone and the inspection vehicle. Let represent the state vector of the inspection vehicle at time j. This indicates the location of the m-th mapping point of the UAV; Step S5: Based on the current position of the inspection vehicle, find a better surveying point within the preset range, switch the position of the UAV surveying point, until the inspection vehicle completes the inspection in the tunnel. Step S6: Based on the geometric accuracy factor, the geometric accuracy factor is minimized by adjusting the position, density, and number of survey points in the roadway, thus optimizing the distribution of survey points in the roadway.

2. The method for optimizing the UWB-enhanced positioning map of a drone for trolley inspection in narrow alleyways according to claim 1, characterized in that, The specific steps of step S1 are as follows: Step S1.1: Collect the current state vector of the inspection vehicle as follows: ； in, This represents the state vector of the inspection vehicle. This represents the three-dimensional position coordinates of the inspection vehicle in the world coordinate system; This represents the three-dimensional velocity vector of the inspection vehicle in the world coordinate system; The attitude quaternion of the inspection vehicle in the world coordinate system is represented; the inspection vehicle is equipped with an inertial measurement unit, which includes an accelerometer and a gyroscope; Indicates accelerometer bias; This indicates the gyroscope bias, and T indicates transpose; Step S1.2: Calculate the predicted pose of the inspection vehicle at the next moment, including the predicted position, predicted velocity, and predicted attitude, as shown in the following formula: ； ； ； in, The rate of change of the inspection vehicle's position is used to characterize the predicted position of the inspection vehicle. The rate of change of the speed of the inspection vehicle is used to characterize the predicted speed of the inspection vehicle. This represents the three-dimensional velocity vector of the inspection vehicle; This represents the raw measurement value from the accelerometer in the inertial measurement unit. It's the accelerometer bias. It is the attitude rotation matrix. It is the gravity vector; The attitude quaternion change rate is used to characterize the predicted attitude of the inspection vehicle. For attitude quaternions, This represents the original measurement value from the gyroscope in the inertial measurement unit. It is the gyroscope bias.

3. The method for optimizing the UWB enhanced positioning map of a drone for trolley inspection in narrow alleyways according to claim 2, characterized in that, The specific steps of step S2 are as follows: Step S2.1: Analyze the laser points in the laser point cloud of the environment where the inspection vehicle is located. The nearest neighbor search method is used to search for the nearest neighbor in the reference point cloud, as shown in the following formula: ； in, It is the normal vector of the reference plane. It is a reference point on the reference plane. It is the attitude rotation matrix. It is a translation vector. Indicates laser point Distance to the reference plane; Step S2.2: For After weighting, the state vector is updated through iterative extended Kalman filtering. The state vector update formula is: ； in, To represent addition on a manifold, It is the Kalman gain matrix. It is the weighted average. ; This represents the state vector of the inspection vehicle.

4. The method for optimizing the UWB enhanced positioning map of a drone for trolley inspection in narrow alleyways according to claim 3, characterized in that, The specific steps of step S3 are as follows: Step S3.1: The UAV is equipped with a UWB module, which integrates a UWB ranging model to calculate the distance between the UAV hovering above the survey point and the inspection vehicle. : ； Where c is the speed of light. , These are the send and receive timestamps for the drone, respectively. , These are the timestamps for sending and receiving the inspection vehicle; Step S3.2: Install anchor points at the four wing corners of the drone, and attach corresponding tags to the inspection vehicle. Calculate the distance between the drone and the inspection vehicle. The specific formula is as follows: ； in, The distance between the drone and the inspection vehicle at timestamp t is calculated, and the z in the subscript represents the anchor point on the drone. This is the transformation matrix from the UAV body to world coordinates. Let be the fixed transformation matrix from the UAV body to the tag at timestamp t. The origin position in the UAV's body coordinate system. Let z be the position of anchor point z in the world coordinate system. For measuring noise.

5. The method for optimizing the UWB-enhanced positioning map of a drone for trolley inspection in narrow alleyways according to claim 4, characterized in that, The specific steps of step S5 are as follows: Step S5.1: Based on the current position of the inspection vehicle, find a better survey point within the preset range. If a better survey point exists, proceed to step S5.2; otherwise, proceed to step S5.

3. Step S5.2: The UAV flies over a better mapping point and hovers to complete the UAV position switch and update the UAV positioning composite residual; Step S5.3: The drone maintains its current position and provides the drone positioning residual for the current position.

6. The method for optimizing the UWB enhanced positioning map of a drone for trolley inspection in narrow alleyways according to claim 5, characterized in that, The criteria for determining a better survey point is: there exists another survey point whose unobstructed distance from the survey point to the inspection vehicle is less than the unobstructed distance between the current survey point and the inspection vehicle.

7. The method for optimizing the UWB enhanced positioning map of a drone for trolley inspection in narrow alleyways according to claim 6, characterized in that, The specific method for step S6 is as follows: The geometric precision factor is calculated as follows: ； in, Indicates geometric precision factor, The Jacobian matrix represents the relationship between the survey points and the positions of the inspection vehicle. The number of rows is the number of survey points, and the number of columns is the sum of the dimensions of the three-dimensional position error of the inspection vehicle and the dimension of the time error. Indicates trace; , , These represent the positional errors of the inspection vehicle in the x, y, and z directions, respectively. This is due to time error; By adjusting the number and location distribution of survey points in the roadway, as well as the density of survey points at bends, the geometric accuracy factor can be adjusted. Minimize to optimize the distribution of survey points in the tunnel.

Images