Obstacle avoidance path optimization method and system for airport unmanned inspection equipment
Patent Information
- Application Number
- CN202611247833.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-18
- Publication Date
- 2026-09-29
AI Technical Summary
机场中的环境具有结构复杂的特点，例如存在建筑物、移动车辆等静态与动态障碍，且安全要求极高，这对无人巡检设备的自主导航与避障能力提出了严峻挑战
本申请针对常规的路径规划方法在机场巡检过程中易受多类复杂障碍的综合影响，难以兼顾路径寻找效率与安全的问题，通过深入分析无人巡检设备所在局部范围内静态障碍物的分布特征，构建静态影响系数，能够对当前时刻下无人巡检设备局部范围内的静态障碍物对路径规划的影响程度进行评估；通过对点云数据分析，能够对动态障碍物的位置及距离进行识别，并基于动态障碍物的运动特征，构建动态碰撞风险系数，能够对无人巡检设备与动态障碍物之间的碰撞风险程度进行评估；进而构建综合风险系数，综合评估了静态、动态障碍物对后续路径规划的影响程度，进而对路径规划算法的目标函数中的权重系数进行实时优化，有助于提高避障路径规划效率。
Smart Images
Figure CN122835406A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of path optimization technology, specifically to a method and system for obstacle avoidance path optimization for unmanned airport inspection equipment. Background Technology
[0002] With the rapid development of unmanned and intelligent technologies in airport operations and maintenance, unmanned inspection equipment (such as unmanned vehicles) is widely used for daily inspections of areas such as runways, aprons, and perimeter fences. The airport environment is characterized by complex structures, including static and dynamic obstacles such as buildings and moving vehicles, and has extremely high safety requirements. This poses a severe challenge to the autonomous navigation and obstacle avoidance capabilities of unmanned inspection equipment.
[0003] Conventional path planning methods, such as local obstacle avoidance based on fixed rules and simple global path search, often struggle to balance path finding efficiency and safety. This is especially true in airport unmanned inspection applications, where obstacle distribution is complex, there are many types of dynamic obstacles, and their trajectories are often irregular, increasing the risk of collisions. Therefore, in the complex and ever-changing obstacle environment of airport inspections, conventional path optimization methods are easily affected by the chaotic obstacle distribution and dynamic uncertainties, resulting in low path planning efficiency. Summary of the Invention
[0004] To address the aforementioned technical problems, the purpose of this application is to provide an obstacle avoidance path optimization method and system for unmanned airport inspection equipment. The specific technical solution adopted is as follows: In a first aspect, embodiments of this application provide an obstacle avoidance path optimization method for unmanned inspection equipment at airports, the method comprising the following steps: Construct a binary raster map of the airport, where the raster is divided into obstacle raster and free raster; After the cold start phase of the unmanned inspection equipment is completed, the proportion and random distribution of obstacle grids within the local window of the unmanned inspection equipment in the binary grid map are analyzed to construct the static influence coefficient at the current moment. The system collects point cloud data in the forward area of the unmanned inspection equipment in real time, and identifies dynamic obstacles in the point cloud data at the current moment based on a pre-trained neural network model. Based on the differences in movement direction, speed, distance between the unmanned inspection equipment and each dynamic obstacle, as well as the irregularity of the movement path of each dynamic obstacle, the system constructs the dynamic collision risk coefficient at the current moment, and combines it with the static influence coefficient to construct the comprehensive risk coefficient at the current moment. Using the comprehensive risk coefficient, the weight coefficients in the objective function of the A* algorithm at the current moment are optimized, thereby planning the obstacle avoidance path for the unmanned inspection equipment.
[0005] In a preferred embodiment, the method for constructing the binary raster map of the airport is as follows: Acquire high-resolution airport images; High-resolution airport images are divided into square grid units of uniform size, with each grid unit being independent of the others. Grids containing static obstacles are designated as obstacle grids; otherwise, they are designated as free grids. Obtain a binary raster map of the airport.
[0006] In a preferred embodiment, the method for constructing the static influence coefficient at the current moment is as follows: Calculate the ratio of the number of obstacle grids in a local window of the unmanned inspection equipment to the total number of grids in the local window at the current moment; Set the grayscale value of the obstacle grid in the local window of the unmanned inspection device at the current moment to 255 and the grayscale value of the free grid to 0 to obtain the binary image of the local window; Calculate the binary image in each case. , , , The gray-level co-occurrence matrices in four directions are calculated, and the mean entropy value of all gray-level co-occurrence matrices is calculated. The static influence coefficient at the current moment is positively correlated with both the ratio and the mean.
[0007] In a preferred embodiment, the specific process of identifying dynamic obstacles in the point cloud data at the current moment is as follows: The point cloud data collected at the current moment is voxelized. The voxelized point cloud data is then used as input to a pre-trained VoxelNet network model, and the output is the position of each dynamic obstacle in the point cloud data at the current moment.
[0008] In a preferred embodiment, the method for constructing the dynamic collision risk coefficient at the current moment is as follows: Based on the differences in motion direction and speed between the unmanned inspection equipment and each dynamic obstacle at the current moment, a relative motion trend coefficient between the unmanned inspection equipment and each dynamic obstacle at the current moment is constructed. Based on the shortest distance between each dynamic obstacle and the unmanned inspection equipment measured by the lidar and its azimuth angle relative to the unmanned inspection equipment, and combined with the coordinates of the unmanned inspection equipment in the binary grid map at the current moment, the grid coordinates of each dynamic obstacle at the current moment are obtained. Find the minimum Manhattan distance between the unmanned inspection equipment and the grid coordinates of all dynamic obstacles at the current moment; Obtain the coordinates of all grids that each dynamic obstacle has passed through from the start of the unmanned inspection equipment to the current moment, and arrange the horizontal and vertical coordinates of all grids in chronological order to form the horizontal and vertical coordinate sequences of each dynamic obstacle. The mean of the sample entropy of the horizontal axis sequence and the sample entropy of the vertical axis sequence of each dynamic obstacle is denoted as the variation coefficient of each dynamic obstacle. The dynamic collision risk coefficient at the current moment is positively correlated with the maximum value of the relative motion trend coefficient between the unmanned inspection equipment and all dynamic obstacles at the current moment, and the maximum value of the variation coefficient of all dynamic obstacles, and negatively correlated with the minimum value of the Manhattan distance.
[0009] In a preferred embodiment, the formula for constructing the relative motion trend coefficient between the unmanned inspection device and each dynamic obstacle at the current moment is: In the formula, Let be the relative motion trend coefficient between the unmanned inspection equipment and the s-th dynamic obstacle at the current moment; This represents the motion vector of the unmanned inspection equipment at the current moment. Let be the motion vector of the s-th dynamic obstacle at the current moment; It is the hyperbolic tangent function; Wherein, the direction of the motion vector of the unmanned inspection device at the current moment refers to the direction from the previous grid coordinate of the unmanned inspection device to the current grid coordinate, and the magnitude refers to the speed of the unmanned inspection device at the current moment; The direction of the motion vector of the s-th dynamic obstacle at the current moment refers to the direction from the previous grid coordinate of the s-th dynamic obstacle to the current grid coordinate, and the magnitude refers to the actual moving speed of the s-th dynamic obstacle.
[0010] In a preferred embodiment, the comprehensive risk coefficient at the current moment is positively correlated with both the static influence coefficient and the dynamic collision risk coefficient.
[0011] In a preferred embodiment, the weight coefficients in the objective function of the A* algorithm at the current time are negatively correlated with the comprehensive risk coefficient.
[0012] In a preferred embodiment, the specific process of planning the obstacle avoidance path for the unmanned inspection equipment is as follows: The optimized weight coefficients at the current moment are used as the weight coefficients in the objective function of the A* algorithm to plan the obstacle avoidance path of the unmanned inspection equipment.
[0013] Secondly, embodiments of this application also provide an obstacle avoidance path optimization system for airport unmanned inspection equipment, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of any of the above-described obstacle avoidance path optimization methods for airport unmanned inspection equipment.
[0014] This application has at least the following beneficial effects: This application addresses the issue that conventional path planning methods are easily affected by multiple complex obstacles during airport inspections, making it difficult to balance path finding efficiency and safety. By deeply analyzing the distribution characteristics of static obstacles within the local area of the unmanned inspection equipment, a static influence coefficient is constructed to assess the impact of static obstacles on path planning at the current moment. Through point cloud data analysis, the location and distance of dynamic obstacles can be identified, and a dynamic collision risk coefficient is constructed based on the motion characteristics of dynamic obstacles to assess the collision risk between the unmanned inspection equipment and dynamic obstacles. Furthermore, a comprehensive risk coefficient is constructed to comprehensively evaluate the impact of static and dynamic obstacles on subsequent path planning, thereby optimizing the weight coefficients in the objective function of the path planning algorithm in real time, which helps improve the efficiency of obstacle avoidance path planning. Attached Figure Description
[0015] Figure 1 A flowchart illustrating the steps of an obstacle avoidance path optimization method for airport unmanned inspection equipment provided in one embodiment of this application; Figure 2 A flowchart illustrating the process of obtaining the comprehensive risk coefficient at the current moment, provided as an embodiment of this application. Detailed Implementation
[0016] Please see Figure 1 The diagram illustrates a flowchart of a method for optimizing obstacle avoidance paths for unmanned airport inspection equipment according to an embodiment of this application. The method includes the following steps: S1: Construct a binary raster map of the airport, where the raster is divided into obstacle raster and free raster.
[0017] This application employs the A* algorithm for path planning of unmanned airport inspection equipment. First, the airport environment needs to be modeled to ensure the algorithm runs in an accurate and structured spatial representation. For this purpose, high-resolution airport images are acquired through drone aerial photography. Then, the high-resolution airport images are rasterized. In this embodiment, the side length of the raster is set to the width of the unmanned inspection equipment, thus dividing the high-resolution airport image into square raster units of uniform size, each independent of the others. Furthermore, the raster is marked based on the presence or absence of static obstacles. Grids with static obstacles are marked as obstacle grids, which are impassable; grids without static obstacles are marked as free grids, which are passable. This yields a binary raster map of the airport. A Cartesian coordinate system is constructed with the lower left corner vertex of the raster map as the origin, allowing the position of each grid to be represented by two-dimensional coordinates.
[0018] It should be noted that the unmanned inspection equipment and moving obstacles in this application are all treated as one grid, and during the movement, they only occupy one grid.
[0019] S2: After the cold start phase of the unmanned inspection equipment is completed, analyze the proportion and random distribution of obstacle grids in the local window of the unmanned inspection equipment in the binary grid map, and construct the static influence coefficient at the current moment.
[0020] During airport inspections, high-density or haphazardly arranged static obstacles significantly reduce the passable area and increase the number of nodes required for path searching. Furthermore, the trajectories of dynamic obstacles may be difficult to predict, further reducing path planning efficiency. This application addresses these issues through the following analysis and processing.
[0021] Specifically, during the inspection process, the unmanned inspection equipment travels along the four-neighbor path of the grid. However, some areas of the airport may have densely distributed static obstacles, requiring the unmanned inspection equipment to frequently turn during inspection, which often results in the planned path not being the optimal solution. In view of this, it is necessary to obtain the distribution characteristics of static obstacles in the airport environment.
[0022] First, during the cold start phase of the unmanned inspection equipment, based on the pre-set start and end points of the unmanned inspection equipment, the default A* algorithm of the unmanned inspection equipment is used to automatically plan the path. In this embodiment, the cold start phase is set to the time period corresponding to the path of the first K (10 in this embodiment) grids starting from the start point. After the cold start phase ends, a path is set with the grid where the unmanned inspection equipment is currently located as the center and a size of [missing information]. Local window, In this embodiment, the side length is preset. Set the value to 7; then, calculate the ratio of the number of obstacle grids in the local window to the total number of grids in the local window, and use it as the local obstacle proportion coefficient at the current moment, denoted as R. The larger the R value, the higher the proportion of obstacle grids in the local area of the unmanned inspection equipment at the current moment.
[0023] Furthermore, when the proportion of obstacle grids is high but their spatial arrangement is simple, the obstacle grids have little impact on the path planning of unmanned inspection equipment. However, if the distribution of obstacles within the local area of the unmanned inspection equipment is more random and discrete, the unmanned inspection equipment may need to make more turns during the inspection process, thus reducing path planning efficiency. Since the resulting binary grid map is similar to a grayscale image, image processing methods can be used to further analyze the obstacle distribution characteristics within the local window of the unmanned inspection equipment. First, the grayscale value of the obstacle grid in the local window of the unmanned inspection equipment at the current moment is set to 255, and the grayscale value of the free grid is set to 0, to obtain a binary image of the local window; the grayscale values of the two grids are then calculated respectively. , , , The gray-level co-occurrence matrices in these four directions are used to calculate the mean of the entropy values of all gray-level co-occurrence matrices. The normalized value of this mean is recorded as the obstacle randomness factor at the current moment, which characterizes the randomness and irregularity of the distribution of static obstacles within the local window of the unmanned inspection equipment at the current moment. The larger the value, the greater the randomness and irregularity of the distribution of static obstacles within the local window of the unmanned inspection equipment at the current moment. The normalization method used is the min-max normalization method. Specifically, the local window is used to traverse the entire binary raster map sequentially from top to bottom and from left to right, with a step size of 1. The mean of the entropy values of the gray-level co-occurrence matrix corresponding to each local window is calculated, and the minimum and maximum values are calculated. This minimum-max normalization is then applied to the mean of the entropy values of the gray-level co-occurrence matrix corresponding to the local window of the unmanned inspection equipment at the current moment. The min-max normalization method is a well-known technique, and its specific process will not be elaborated further.
[0024] As a preferred implementation, a static influence coefficient is constructed based on the proportion and random distribution of obstacle grids within the local window of the unmanned inspection device in the binary grid map at the current moment. This coefficient characterizes the influence of static obstacles within the local area of the unmanned inspection device on path planning at the current moment. The method for constructing the static influence coefficient at the current moment is as follows: The ratio of the number of obstacle grids within the local window of the unmanned inspection device to the total number of grids in the local window is calculated; the grayscale value of the obstacle grids in the local window of the unmanned inspection device is set to 255, and the grayscale value of the free grids is set to 0, resulting in a binary image of the local window; the influence coefficient of the binary image in the local window is then calculated. , , , The gray-level co-occurrence matrices in four directions are calculated, and the mean entropy value of all gray-level co-occurrence matrices is statistically analyzed. The static influence coefficient at the current moment is positively correlated with both the ratio and the mean. The positive correlation means that the dependent variable increases as the independent variable increases, or decreases as the independent variable decreases.
[0025] In this embodiment, the static influence coefficient at the current moment is denoted as... Its specific expression is: In the formula, This represents the static influence coefficient at the current moment. This represents the local obstacle percentage coefficient at the current moment. This represents the obstacle random factor at the current moment. Dividing the numerator by 2 is to normalize the numerator whose value range is [0,2].
[0026] income A larger value indicates a higher proportion of static obstacles within the local window of the current grid node, or a more pronounced random distribution of static obstacles. This means the unmanned inspection device is more likely to change direction frequently at the current moment, thus affecting path planning efficiency. When there are no obstructed grids within the local window of the unmanned inspection device, the value of A is set to 0. Subsequent analysis will only consider dynamic risks.
[0027] S3: Collect point cloud data in the forward area of the unmanned inspection equipment in real time, and identify dynamic obstacles in the point cloud data at the current moment based on the pre-trained neural network model; construct the dynamic collision risk coefficient at the current moment based on the difference in movement direction, speed, distance between the unmanned inspection equipment and each dynamic obstacle, as well as the irregularity of the movement path of each dynamic obstacle, and combine it with the static influence coefficient to construct the comprehensive risk coefficient at the current moment.
[0028] Furthermore, during unmanned inspection equipment inspections, dynamic obstacles such as vehicles and pedestrians are frequently encountered, requiring timely obstacle avoidance. These dynamic obstacles are typically located within a free grid, and the unmanned inspection equipment can monitor these obstacles using its own LiDAR. Specifically, the LiDAR collects point cloud data in real-time within the unmanned inspection equipment's forward movement area, which is a 180° fan-shaped area centered on the equipment's direction of travel. The collected point cloud data is then voxelized and used as input to a pre-trained VoxelNet network model to identify various dynamic obstacles within the point cloud data. The model outputs the position of each dynamic obstacle in the current point cloud data and marks each obstacle with a bounding box. This method is well-known, and the specific process will not be elaborated further. Based on the shortest distances between each dynamic obstacle and the unmanned inspection device identified by lidar measurements, and their azimuth angles relative to the unmanned inspection device, combined with the coordinates of the unmanned inspection device in the binary grid map at the current moment, the grid coordinates of each dynamic obstacle at the current moment are obtained. For example, the grid coordinates of the unmanned inspection device at the current moment are ( , The device travels along the positive x-axis. The lidar detects a dynamic obstacle 10 meters away from the unmanned inspection equipment in the positive x-axis direction, with a grid size of 2 meters. Taking 2m as an example, the grid coordinates of the dynamic obstacle can be obtained as ( , ).
[0029] The closer the unmanned inspection equipment gets to an obstacle, the greater the risk of collision, and the more timely the obstacle avoidance should be. For example, if the unmanned inspection equipment and a dynamic obstacle are moving towards each other, the closer the distance between them, or the more irregular the trajectory of the dynamic obstacle, the higher the risk of collision.
[0030] Therefore, firstly, the previous grid cell in the current unmanned inspection device's travel path is obtained. The direction from the previous grid cell coordinates to the current grid cell coordinates is taken as the current motion direction of the unmanned inspection device. The magnitude of the unmanned inspection device's current speed is used as the vector magnitude (the speed is obtained by the unmanned inspection device's own speed sensor, unit: m / s). The motion vector of the unmanned inspection device at the current moment is constructed and denoted as . Then, taking the s-th dynamic obstacle at the current moment as an example, we obtain the actual moving speed of the s-th dynamic obstacle. Similarly, we take the direction from the previous grid coordinate of the s-th dynamic obstacle to the current grid coordinate as the moving direction of the s-th dynamic obstacle at the current moment, and use the current moving speed of the s-th dynamic obstacle as the magnitude of the vector to construct the moving vector of the s-th dynamic obstacle at the current moment, denoted as [vector]. In this embodiment, the actual moving speed of each dynamic obstacle refers to the ratio of the actual displacement of each dynamic obstacle in the previous L (L is 5 in this embodiment) moments to the total duration. Specifically, if it is detected that the number of captured moments recorded for each dynamic obstacle at the current moment is less than L, then the total actual displacement within the saved period is divided by its corresponding duration, and this value is used as the actual moving speed of each dynamic obstacle.
[0031] It should be noted that if the previous grid coordinate of the same dynamic obstacle is the same as the current grid coordinate, then the motion vector direction of the dynamic obstacle at the current moment will be assigned a zero value and used in subsequent calculations.
[0032] The formula for calculating the relative motion trend coefficient between the unmanned inspection equipment and the s-th dynamic obstacle at the current moment is as follows: In the formula, Let be the relative motion trend coefficient between the unmanned inspection equipment and the s-th dynamic obstacle at the current moment; This represents the motion vector of the unmanned inspection equipment at the current moment. Let be the motion vector of the s-th dynamic obstacle at the current moment; The function is a hyperbolic tangent function, which maps the dot product between the motion vectors of the unmanned inspection equipment and the dynamic obstacle to the range of (-1,1), thereby stably representing the relative motion trend characteristics of the two.
[0033] If in the formula A positive value indicates a higher likelihood that the unmanned inspection device and the s-th dynamic obstacle are moving towards each other, and the greater their speeds, the greater the risk of collision. If... The value of is negative; the smaller the value, the lower the risk of collision between the unmanned inspection equipment and the dynamic obstacle, indicating that they are moving in the same direction. Therefore, the obtained... It reflects the collision risk level between the unmanned inspection equipment and the s-th dynamic obstacle at the current moment; the larger the value, the higher the collision risk between the unmanned inspection equipment and the s-th dynamic obstacle at the current moment.
[0034] Furthermore, the closer the unmanned inspection equipment is to dynamic obstacles, the higher the risk of collision. Therefore, the Manhattan distance between the unmanned inspection equipment and the grid coordinates of all dynamic obstacles at the current moment is calculated. The minimum value among all Manhattan distances is recorded as the distance risk coefficient at the current moment; the smaller the value, the closer the unmanned inspection equipment is to the dynamic obstacles at the current moment.
[0035] However, the movement trajectories of dynamic obstacles may be irregular, such as the frequent speed changes of ground personnel, further increasing the risk of collision. Therefore, the coordinates of all grids traversed by the s-th dynamic obstacle from the start of the unmanned inspection equipment's inspection to the current moment are obtained. The x and y coordinates of all grids are arranged in chronological order to form the x-coordinate sequence and y-coordinate sequence of the s-th dynamic obstacle, respectively. The mean of the sample entropy of the x-coordinate sequence and the sample entropy of the y-coordinate sequence of the s-th dynamic obstacle is recorded as the variation coefficient of the s-th dynamic obstacle; the larger the value, the more irregular the movement path of the s-th dynamic obstacle at the current moment. In the above sample entropy calculation process, the embedding dimension m and similarity tolerance r are pre-set for trajectory sequence feature comparison; in this embodiment, the embedding dimension m is set to 2, and the similarity tolerance r is set to within 0.1 to 0.2 times the standard deviation of the corresponding dynamic obstacle's x-coordinate sequence.
[0036] As a preferred implementation, a dynamic collision risk coefficient is constructed based on the differences in movement direction, speed, and distance between the unmanned inspection device and each dynamic obstacle at the current moment, as well as the irregularity of the movement paths of each dynamic obstacle. This coefficient characterizes the degree of collision risk between the unmanned inspection device and the dynamic obstacles at the current moment. The method for constructing the dynamic collision risk coefficient at the current moment is as follows: constructing the relative motion trend coefficient between the unmanned inspection device and each dynamic obstacle at the current moment; calculating the minimum Manhattan distance between the grid coordinates of the unmanned inspection device and all dynamic obstacles at the current moment; obtaining the variation coefficient of each dynamic obstacle; calculating the maximum value of the variation coefficient of all dynamic obstacles at the current moment. The dynamic collision risk coefficient at the current moment is positively correlated with the relative motion trend coefficient and the maximum value of the variation coefficient, and negatively correlated with the minimum Manhattan distance. The negative correlation means that the dependent variable decreases as the independent variable increases, or the dependent variable increases as the independent variable decreases.
[0037] In this embodiment, the dynamic collision risk coefficient at the current moment is denoted as... Its specific expression is: In the formula, This represents the dynamic collision risk coefficient at the current moment. This represents the maximum value among the relative motion trend coefficients between the unmanned inspection equipment and all dynamic obstacles at the current moment. It represents the maximum value among all dynamic obstacle variation coefficients at the current moment; The distance risk coefficient at the current moment; This is a preset constant used to prevent the denominator from being 0, and its value range is [0.01, 0.1]. In this embodiment, it is set to 0.01. For the normalization function, this embodiment uses the minimum-maximum normalization method for normalization. The minimum and maximum values are determined based on the corresponding parameter sets of all moments in the historical inspection process of the unmanned inspection equipment. If the minimum and maximum values are equal, then let... , The normalized values of all are equal to 0, let The normalized value is equal to 1, indicating that the unmanned inspection equipment is relatively stationary relative to dynamic obstacles during the historical inspection process.
[0038] Specifically, if no dynamic obstacle is detected at the current moment, the dynamic collision risk coefficient at the current moment is set to 0, indicating that there is no risk of collision with a dynamic obstacle.
[0039] The larger the value of E, the higher the risk of collision between the unmanned inspection equipment and dynamic obstacles at the current moment.
[0040] Furthermore, based on the above analysis, parameter A characterizes the density and spatial dispersion of static obstacles in the local environment of the unmanned inspection equipment at the current moment; parameter E quantifies the collision risk between the unmanned inspection equipment and dynamic obstacles at the current moment. Therefore, as a preferred implementation, a comprehensive risk coefficient is constructed based on the static influence coefficient and dynamic collision risk coefficient at the current moment to characterize the combined influence of static and dynamic obstacles on path planning at the current moment. The comprehensive risk coefficient at the current moment is positively correlated with both the static influence coefficient and the dynamic collision risk coefficient. The flowchart for obtaining the comprehensive risk coefficient at the current moment is as follows. Figure 2 As shown.
[0041] In this embodiment, the comprehensive risk coefficient at the current moment is denoted as... Its formula is: In the formula, This represents the overall risk coefficient at the current moment. This represents the static influence coefficient at the current moment. This represents the dynamic collision risk coefficient at the current moment. As a normalization function, this embodiment uses the minimum-maximum normalization method for normalization, and its minimum and maximum values are determined based on the corresponding parameter set of all moments in the historical inspection process of the unmanned inspection equipment; , These are the preset first weight and the preset second weight, respectively. Its specific value can be determined according to actual needs. For example, in environments where dynamic obstacles are frequently encountered, a larger value needs to be set. To increase the impact of dynamic collision risks, a larger value needs to be set in inspection environments dominated by static obstacles. The value is set to increase the degree of influence on static obstacles. In this embodiment, it is set to... , .
[0042] The resulting comprehensive risk coefficient represents the combined impact of static and dynamic obstacles on path planning at the current moment, providing a basis for subsequent path planning. A higher value indicates a greater combined impact of static and dynamic obstacles on path planning at the current moment.
[0043] S4: Using the comprehensive risk coefficient, optimize the weight coefficients in the objective function of the A* algorithm at the current moment, and then plan the obstacle avoidance path of the unmanned inspection equipment.
[0044] Furthermore, based on the comprehensive risk coefficient at the current moment, the weight coefficients in the objective function of the path planning algorithm are optimized to obtain a more efficient and smooth inspection path.
[0045] Specifically, the objective function of the A* algorithm is set as follows: In the formula, The objective function is... The cost function; For heuristic functions; The weighting coefficient is used. When the influence of obstacles is greater, the weight of the heuristic function should be actively reduced so that the objective function focuses more on the calculation of the actual cost function, thereby prioritizing the avoidance of collision risks in path planning; conversely, a larger heuristic function weight can be set to obtain a more convenient inspection path.
[0046] Therefore, based on the comprehensive risk coefficient at the current moment, the weight coefficients in the objective function of the A* algorithm at the current moment are optimized. The weight coefficients in the objective function of the A* algorithm at the current moment are negatively correlated with the comprehensive risk coefficient.
[0047] In this embodiment, the specific optimization formula for optimizing the weight coefficients in the objective function at the current moment is as follows: In the formula, The optimized weight coefficients at the current moment; To preset the initial weight coefficients, and In this embodiment, 1.5 is used; Let be the comprehensive risk coefficient at the current moment, and its value range is [0,1]. Therefore, the optimized weight coefficient has a value range of [0.5,1.5].
[0048] Then, the optimized weight coefficients at the current moment are used as weight coefficients in the objective function of the A* algorithm to plan the obstacle avoidance path of the unmanned inspection equipment. The A* algorithm is a well-known technique and will not be described in detail here.
[0049] Similarly, the weight coefficients in the objective function at all subsequent time points are optimized, thereby planning the path at all time points.
[0050] Optimizing the obstacle avoidance path of unmanned inspection equipment at airports using the above methods helps improve the efficiency of path replanning.
[0051] Based on the same inventive concept as the above method, this application also provides an obstacle avoidance path optimization system for airport unmanned inspection equipment, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of the above-described obstacle avoidance path optimization method for airport unmanned inspection equipment.
Claims
1. A method for optimizing obstacle avoidance paths for unmanned inspection equipment at airports, characterized in that, The method includes the following steps: Construct a binary raster map of the airport, where the raster is divided into obstacle raster and free raster; After the cold start phase of the unmanned inspection equipment is completed, the proportion and random distribution of obstacle grids within the local window of the unmanned inspection equipment in the binary grid map are analyzed to construct the static influence coefficient at the current moment. The system collects point cloud data in the forward area of the unmanned inspection equipment in real time, and identifies dynamic obstacles in the point cloud data at the current moment based on a pre-trained neural network model. Based on the differences in movement direction, speed, distance between the unmanned inspection equipment and each dynamic obstacle, as well as the irregularity of the movement path of each dynamic obstacle, the system constructs the dynamic collision risk coefficient at the current moment, and combines it with the static influence coefficient to construct the comprehensive risk coefficient at the current moment. Using the comprehensive risk coefficient, the weight coefficients in the objective function of the A* algorithm at the current moment are optimized, thereby planning the obstacle avoidance path for the unmanned inspection equipment.
2. The obstacle avoidance path optimization method for unmanned airport inspection equipment as described in claim 1, characterized in that, The method for constructing the binary raster map of the airport is as follows: Acquire high-resolution airport images; High-resolution airport images are divided into square grid units of uniform size, with each grid unit being independent of the others. Grids containing static obstacles are designated as obstacle grids; otherwise, they are designated as free grids. Obtain a binary raster map of the airport.
3. The obstacle avoidance path optimization method for unmanned airport inspection equipment as described in claim 1, characterized in that, The method for constructing the static influence coefficient at the current moment is as follows: Calculate the ratio of the number of obstacle grids in a local window of the unmanned inspection equipment to the total number of grids in the local window at the current moment; Set the grayscale value of the obstacle grid in the local window of the unmanned inspection device at the current moment to 255 and the grayscale value of the free grid to 0 to obtain the binary image of the local window; Calculate the binary image in each case. , , , The gray-level co-occurrence matrices in four directions are calculated, and the mean entropy value of all gray-level co-occurrence matrices is calculated. The static influence coefficient at the current moment is positively correlated with both the ratio and the mean.
4. The obstacle avoidance path optimization method for unmanned airport inspection equipment as described in claim 1, characterized in that, The specific process for identifying dynamic obstacles in the point cloud data at the current moment is as follows: The point cloud data collected at the current moment is voxelized. The voxelized point cloud data is then used as input to a pre-trained VoxelNet network model, and the output is the position of each dynamic obstacle in the point cloud data at the current moment.
5. The obstacle avoidance path optimization method for unmanned airport inspection equipment as described in claim 1, characterized in that, The method for constructing the dynamic collision risk coefficient at the current moment is as follows: Based on the differences in motion direction and speed between the unmanned inspection equipment and each dynamic obstacle at the current moment, a relative motion trend coefficient between the unmanned inspection equipment and each dynamic obstacle at the current moment is constructed. Based on the shortest distance between each dynamic obstacle and the unmanned inspection equipment measured by the lidar and its azimuth angle relative to the unmanned inspection equipment, and combined with the coordinates of the unmanned inspection equipment in the binary grid map at the current moment, the grid coordinates of each dynamic obstacle at the current moment are obtained. Find the minimum Manhattan distance between the unmanned inspection equipment and the grid coordinates of all dynamic obstacles at the current moment; Obtain the coordinates of all grids that each dynamic obstacle has passed through from the start of the unmanned inspection equipment to the current moment, and arrange the horizontal and vertical coordinates of all grids in chronological order to form the horizontal and vertical coordinate sequences of each dynamic obstacle. The mean of the sample entropy of the horizontal axis sequence and the sample entropy of the vertical axis sequence of each dynamic obstacle is denoted as the variation coefficient of each dynamic obstacle. The dynamic collision risk coefficient at the current moment is positively correlated with the maximum value of the relative motion trend coefficient between the unmanned inspection equipment and all dynamic obstacles at the current moment, and the maximum value of the variation coefficient of all dynamic obstacles, and negatively correlated with the minimum value of the Manhattan distance.
6. The obstacle avoidance path optimization method for unmanned airport inspection equipment as described in claim 5, characterized in that, The formula for constructing the relative motion trend coefficient between the unmanned inspection equipment and each dynamic obstacle at the current moment is: In the formula, Let be the relative motion trend coefficient between the unmanned inspection equipment and the s-th dynamic obstacle at the current moment; This represents the motion vector of the unmanned inspection equipment at the current moment. Let be the motion vector of the s-th dynamic obstacle at the current moment; It is the hyperbolic tangent function; Wherein, the direction of the motion vector of the unmanned inspection device at the current moment refers to the direction from the previous grid coordinate of the unmanned inspection device to the current grid coordinate, and the magnitude refers to the speed of the unmanned inspection device at the current moment; The direction of the motion vector of the s-th dynamic obstacle at the current moment refers to the direction from the previous grid coordinate of the s-th dynamic obstacle to the current grid coordinate, and the magnitude refers to the actual moving speed of the s-th dynamic obstacle.
7. The obstacle avoidance path optimization method for unmanned airport inspection equipment as described in claim 1, characterized in that, The overall risk coefficient at the current moment is positively correlated with both the static influence coefficient and the dynamic collision risk coefficient.
8. The obstacle avoidance path optimization method for unmanned airport inspection equipment as described in claim 1, characterized in that, The weight coefficients in the objective function of the A* algorithm at the current time are negatively correlated with the comprehensive risk coefficient.
9. The obstacle avoidance path optimization method for unmanned airport inspection equipment as described in claim 1, characterized in that, The specific process for planning the obstacle avoidance path of the unmanned inspection equipment is as follows: The optimized weight coefficients at the current moment are used as the weight coefficients in the objective function of the A* algorithm to plan the obstacle avoidance path of the unmanned inspection equipment.
10. An obstacle avoidance path optimization system for airport unmanned inspection equipment, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the obstacle avoidance path optimization method for unmanned airport inspection equipment as described in any one of claims 1-9.