Multi-unmanned aerial vehicle dynamic path planning method and device based on escape optimization support vector machine
By using a path planning method based on escape optimization support vector machine, the real-time performance and safety issues of path planning for multiple UAVs in tunnels and underground spaces were solved, enabling efficient and safe operation in complex dynamic environments.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-31
- Publication Date
- 2026-04-14
AI Technical Summary
Existing multi-UAV path planning algorithms suffer from poor real-time performance and are prone to getting trapped in local optima in complex dynamic environments such as tunnels and underground spaces, making it difficult to meet the requirements for efficient and safe operations. In particular, they are difficult to achieve accuracy and reliability in environments with signal interference and dynamic obstacles.
An escape optimization support vector machine-based approach is adopted. By constructing an environmental information database, using an SVM model with radial basis function kernel function for environmental classification, combining escape optimization algorithm to plan the UAV path, and adjusting in real time to cope with environmental changes, a spatiotemporal conflict matrix is constructed for path optimization.
It enables efficient and safe path planning for multiple UAVs in tunnels and underground spaces, ensuring real-time performance and global optimization in complex dynamic environments, and improving the efficiency and safety of multi-UAV collaborative operations.
Smart Images
Figure CN121857782A_ABST
Abstract
Description
[0001] This application relates to the field of multi-UAV dynamic path planning technology, and in particular to a method and apparatus for multi-UAV dynamic path planning based on escape optimization support vector machine. Background Technology
[0002] With the continuous advancement of infrastructure construction, the scale and complexity of tunnel and underground space engineering projects are constantly expanding. In engineering scenarios such as railway and highway tunnel construction, underground utility tunnel construction, and mining, higher demands are being placed on the efficiency and accuracy of tasks such as project progress monitoring, environmental safety assessment, and disaster emergency response. Multi-UAV collaborative operations, with their advantages of flexibility, high efficiency, convenience, and rapid deployment, are gradually becoming a key technological means for data acquisition, surveying and detection, and emergency rescue in the field of tunnel and underground space engineering. For example, in long-distance tunnel construction, multiple UAVs can quickly complete tasks such as tunnel interior topographic mapping and crack detection; when underground spaces experience collapses or water seepage, multiple UAVs can quickly penetrate into the disaster site to conduct disaster reconnaissance and personnel search and rescue, greatly improving the efficiency of engineering operations and emergency response.
[0003] However, the unique operating environment of tunnels and underground spaces presents numerous severe challenges to multi-UAV path planning. On one hand, this environment is highly enclosed and features complex terrain, with narrow, winding interior spaces and numerous irregular fixed obstacles such as tunnel walls, support structures, and construction equipment. Simultaneously, dynamic obstacles arise during operations, such as moving engineering vehicles, construction workers, and debris from collapses. The location and trajectory of these dynamic obstacles are difficult to predict, significantly increasing the difficulty of path planning. On the other hand, tunnels and underground spaces suffer from severe signal interference, making it difficult to effectively cover the area with Global Navigation Satellite System (GNSS) signals. This leads to decreased UAV positioning accuracy and unstable communication links, rendering traditional satellite-based path planning methods ineffective. Furthermore, the tunnel environment also presents challenges such as poor lighting conditions, high dust concentrations, and large temperature and humidity variations, further exacerbating the complexity of the environment and placing extremely high demands on the accuracy, real-time performance, and reliability of multi-UAV path planning. Currently, multi-UAV path planning algorithms all have significant limitations: 1) Traditional path planning algorithms, such as A* algorithm and Dijkstra's algorithm, are based on graph search principles and plan UAV routes by calculating the shortest path between nodes. These algorithms can find the theoretically optimal path in simple static environments, but in complex dynamic environments such as tunnels and underground spaces, they require traversing and calculating a large number of nodes, resulting in extremely high computational complexity and poor real-time performance, which cannot meet the needs of multiple UAVs for rapid response to dynamic obstacles. 2) Existing intelligent algorithms lack the ability to adapt to changes in the environment. Once the environment changes, the path needs to be recalculated, making it difficult to ensure the continuity and safety of drone operations. 3) Some intelligent optimization algorithms, such as genetic algorithms and particle swarm optimization, improve the efficiency and quality of path planning to a certain extent by simulating biological evolution or swarm intelligence behavior, and can handle some complex environmental problems. However, when dealing with high-dimensional and nonlinear environmental data in tunnels and underground spaces, these algorithms are prone to getting trapped in local optima, resulting in a non-globally optimal planned path, which may lead the drone into dangerous areas. 4) The parameter settings of the algorithm have a significant impact on the path planning results. In complex and ever-changing engineering environments, it is difficult to determine the optimal parameter combination, which increases the difficulty and uncertainty of algorithm application. 5) Support Vector Machine (SVM), as a machine learning algorithm based on statistical learning theory, performs excellently in pattern recognition, data classification, and other fields, and can effectively classify and extract features from complex environmental data. However, in multi-UAV path planning, SVM is mainly used for preliminary identification of environmental features and cannot directly generate optimized paths in dynamic environments, lacking global search and dynamic adjustment capabilities for path planning.
[0004] Therefore, existing path planning technologies are insufficient to meet the actual needs of efficient and safe operation of multiple drones in tunnel and underground space engineering. Summary of the Invention
[0005] In view of the above problems, this application provides a dynamic path planning method for multiple UAVs based on escape optimization support vector machine to meet the needs of multiple UAVs operating in complex dynamic environments (such as signal interference, dynamic obstacles, narrow spaces, etc.). It aims to solve the problems of poor real-time performance and easy getting trapped in local optima in traditional path planning algorithms in complex environments, and improve the operating efficiency and safety of multiple UAVs in tunnel and underground space engineering.
[0006] In a first aspect, embodiments of this application provide a dynamic path planning method for multiple unmanned aerial vehicles (UAVs) based on escape optimization support vector machines, the method comprising: Using drones to acquire environmental information about tunnels and underground spaces, the tunnel space is divided into three-dimensional grid cells, and an environmental information database is established based on the attributes of the grid cells; Feature vectors of grid cells are extracted from the environmental information database, and a training set is constructed based on the feature vectors and environmental category labels. We select a radial basis kernel function to construct an SVM model, and optimize the penalty parameter of the SVM model through cross-validation on the training set to obtain a trained environment category prediction SVM model. The real-time collected environmental information is input into the environmental category prediction SVM model, and the environmental category prediction results are output. Hazardous areas are identified based on environmental category prediction results, and the optimal path for each drone is obtained using an escape optimization algorithm. For different UAVs, a spatiotemporal conflict matrix is constructed to detect whether there are conflicting paths and to adjust the conflicting paths. The environmental change index is calculated in real time. If the environmental change index exceeds the change threshold, the above steps are repeated to achieve dynamic path adjustment.
[0007] Optionally, the step of using drones to acquire environmental information about the tunnel and underground space, dividing the tunnel space into three-dimensional grid cells, and establishing an environmental information database based on the attributes of the grid cells includes: By utilizing multiple sensors mounted on the drone, environmental information about tunnels and underground spaces is acquired, and the data collected by each type of sensor is compiled into a dataset. D ; The collection area of the tunnel and underground space is divided into Each grid cell is associated with multiple attribute information, and each attribute information is a set. D One of the elements; An environmental information database is constructed based on grid cells and their attribute information. ,in, Indicates coordinates as Grid cells, express Related attribute information.
[0008] Optionally, the step of determining the danger zone based on the environmental category prediction results and using an escape optimization algorithm to obtain the optimal path for each UAV includes: Based on the environmental category prediction results, grid cells classified as dynamic obstacles or high-risk environments are designated as hazardous areas. Based on the dangerous area, an escape optimization algorithm is used to update the drone path and plan the optimal path for each drone.
[0009] Optionally, the step of updating the drone path based on the danger zone using an escape optimization algorithm and planning the optimal path for each drone includes: Initialize the ESC algorithm parameters and generate an initial population, which contains the following: n Each of the five drones has P randomly generated initial paths, and each initial path represents an individual. During the exploration phase, the fitness value of each initial path is calculated based on the fitness function; Individuals in the population are sorted in ascending order based on their fitness values, and the population is divided into calm group, follower group, and panic group. The positions of individuals in the calm group, follower group, and panic group are updated respectively, and the individuals with the best performance in the search phase are added to the elite pool. During the utilization phase, the position of each individual in the elite pool is iteratively updated using the Levy flight strategy until the maximum number of iterations is reached. Or the rate of change of fitness value is less than the threshold. When the optimal path is reached, output the optimal path. .
[0010] Optionally, calculating the fitness value of each initial path based on the fitness function includes: Obtain the path length of the P initial paths for each drone. ; Determine the path safety of each initial path. ; Determining the multi-UAV cooperation of each initial path ; Will , and We perform weighted combinations to obtain the fitness values of each initial path.
[0011] Optionally, the step of constructing a spatiotemporal conflict matrix for the optimal path of each UAV, using the spatiotemporal conflict matrix to detect whether conflicting paths exist, and adjusting the conflicting paths includes: Building drones m The optimal path with another drone n The spatiotemporal conflict matrix between the optimal paths; Determine if the spatiotemporal conflict matrix is 1. If it is, it means that there is a conflict between the two paths, and proceed to the next step. Otherwise, output the optimal path. Drones at the point of conflict n Insert a waiting node, where the waiting node represents the waiting time.
[0012] Secondly, embodiments of this application provide a multi-UAV dynamic path planning device based on escape optimization support vector machine, implemented based on the above method, the device comprising: The acquisition module is used to acquire environmental information of tunnels and underground spaces using drones, divide the tunnel space into three-dimensional grid cells, and establish an environmental information database based on the attributes of the grid cells. The extraction module is used to extract feature vectors of grid cells from the environmental information database and construct a training set based on the feature vectors and environmental category labels. The model building module is used to select the radial basis kernel function to construct the SVM model, and optimize the penalty parameter of the SVM model through cross-validation on the training set to obtain a trained environment category prediction SVM model. The prediction module is used to input the real-time collected environmental information into the environmental category prediction SVM model and output the environmental category prediction results. The path output module determines the danger zone based on the environmental category prediction results and uses an escape optimization algorithm to obtain the optimal path for each UAV. The conflict handling module is used to construct a spatiotemporal conflict matrix for the optimal path of different UAVs, use the spatiotemporal conflict matrix to detect whether there are conflicting paths, and adjust the conflicting paths. The dynamic adjustment module is used to calculate the environmental change index in real time. If the environmental change index exceeds the change threshold, the route is replanned to achieve dynamic route adjustment.
[0013] Compared with the prior art, the specific beneficial effects of the present invention are as follows: This invention addresses the complexity and real-time requirements of dynamic path planning for multiple unmanned aerial vehicles (UAVs) in tunnel and underground space engineering. It constructs a dynamic path planning system for multiple UAVs by deeply integrating escape optimization algorithms with support vector machines (SVMs). Utilizing the accurate classification capabilities of SVMs for tunnel and underground space environmental features, combined with the global-local search mechanism of ESCs simulating biological escape behavior, it achieves efficient path planning and real-time adjustment for multiple UAVs in complex dynamic environments, ensuring the safety and efficiency of collaborative operations among multiple UAVs. Attached Figure Description
[0014] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments of this application will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0015] Figure 1 This is a flowchart of the method proposed in this invention; Figure 2 This is a block diagram illustrating the principle of the Escape Optimization (ESC) algorithm. Detailed Implementation
[0016] Exemplary embodiments of this application will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of this application are shown in the drawings, it should be understood that this application may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of this application and to fully convey the scope of this application to those skilled in the art.
[0017] Example 1: As Figure 1 As shown in the figure, this embodiment provides a multi-UAV dynamic path planning method based on escape optimization support vector machine. The method includes the following steps: Step 1: Use drones to acquire environmental information about the tunnel and underground space, divide the tunnel space into three-dimensional grid cells, and establish an environmental information database based on the attributes of the grid cells; Using a drone equipped with multiple sensors, including a lidar (such as the Velodyne VLP-16, with an accuracy of ±2cm) and a camera, the drone scans the tunnel and underground space at a speed of v, collecting 3D coordinates, obstacle positions, dynamic targets, reflection intensity, and image data every u seconds. Let the data collected by each sensor be a set. ,in Indicates the current sensor's number The data from each sampling point includes the tunnel's three-dimensional coordinates (length, width, and height), reflection intensity (i.e., the energy intensity parameter of the lidar echo signal), which is used to characterize the scanned object's ability to reflect the laser beam, image pixels, and other information.
[0018] The collection area of the tunnel and underground space is divided into There are 3D voxel grids, each with a size of m × m × m. Each grid cell is associated with multiple attribute information, and its coordinates are... Grid cells The attribute information is ,in, Represents grid cells The The attribute information is data collected by the sensor and can include obstacle type, presence of obstacle, safety level (level 1-5, with higher values indicating higher risk), reflection intensity, image grayscale and texture features, tunnel spatial morphology (including three-dimensional coordinates, point cloud density, curvature, clearance height, etc.), and dynamic target status (including target speed, direction, and optical flow changes, etc.). like Represents grid cells The first attribute information is "whether there is an obstacle (0 for no, 1 for yes)". Represents grid cells The second attribute information is "obstacle type (s for safe, g for fixed obstacle, d for dynamic obstacle)".
[0019] Based on grid cells and their attribute information, an environmental information database covering tunnels and underground spaces is constructed. .
[0020] Step 2: Extract feature vectors of grid cells from the environmental information database, and construct a training set based on the feature vectors and environmental category labels; Specifically, since the environmental information database contains multi-dimensional attribute data for path planning, based on point cloud statistical analysis, image feature extraction, and multi-frame motion analysis, geometric features (including three-dimensional coordinates, point cloud density, curvature, and clearance height) for characterizing the spatial morphology of the tunnel, entity features (including obstacle type, size, reflection intensity, image grayscale and texture features, and safety level) for identifying obstacle types and safety levels, and motion features (including target speed, direction, and optical flow changes) for reflecting the dynamic state of targets can be extracted from the environmental information database. These extracted features are then combined into a total feature vector X, which is used for environmental category prediction in a support vector machine model.
[0021] Furthermore, environmental information database The extracted feature vector set is denoted as ,in For including the first i Each grid cell attribute and the feature vector of the surrounding grid, wherein the surrounding grid refers to the adjacent grid cells within a certain range around the current grid cell; Each feature vector The corresponding tag is The corresponding environment category (i.e., the "obstacle type" in the attribute information) is determined by the set of feature vectors. The training set consists of the labels and the labels. (Training set = {feature vectors, corresponding environment categories}).
[0022] Step 3: Select the radial basis kernel function to construct the SVM model, and optimize the penalty parameter of the SVM model on the training set through cross-validation to obtain the trained environment category prediction SVM model; The radial basis function (RBF kernel) is a key kernel function in support vector machines (SVM). It is mainly used to deal with nonlinear classification problems. It maps data to a high-dimensional space, making data that was originally linearly inseparable linearly separable in the high-dimensional space.
[0023] This application uses the radial basis function (RBF) to construct the SVM model, and its kernel function formula is as follows: ; in, For kernel function parameters, For Euclidean distance For the first 1 eigenvector For the first 1 eigenvector; Penalty parameters of the SVM model through cross-validation and Optimize the selection based on the penalty parameter. and The SVM model is trained on the training set by minimizing the following objective function, and the SVM model with the last model parameter adjustment is used as the trained model and used as the environment category prediction SVM model. The objective function is: st ; in, For the weight vector, For bias terms, As slack variables, This is a kernel function mapping.
[0024] Step 4: Input the real-time collected environmental information into the environmental category prediction SVM model and output the environmental category prediction results; Once the model is trained, the trained environment category prediction SVM model can be used to classify real-time environment data and output the environment category prediction results. .
[0025] Step 5: Determine the danger zone based on the environmental category prediction results, and use the escape optimization algorithm to obtain the optimal path for each UAV; Optionally, step 5 may include the following sub-steps: Step 5.1: Based on the environmental category prediction results, grid cells with environmental categories of dynamic obstacles or high risk are designated as hazardous areas; Specifically, based on the classification results of the SVM model output according to environmental categories, the set of grid cells containing dynamic obstacle categories or high-risk categories can be marked as danger zones. ,in, Indicates the first A dangerous area.
[0026] Step 5.2: Based on the danger zone, update the drone path using the escape optimization algorithm and plan the optimal path for each drone; Escape Optimization (ESC) is a metaheuristic algorithm inspired by crowd evacuation behavior. This algorithm, drawing inspiration from crowd evacuation behavior, is used to solve real-world cases and benchmark problems. Figure 2 Step 5.2 includes the following sub-steps: Step 5.2.1: Initialize the ESC algorithm parameters and generate an initial population, the initial population containing for nEach of the five drones has P randomly generated initial paths, and each initial path represents an individual. Specifically, let the number of drones be... Randomly generated for each drone Initial path , .in Indicates the first The first path Each initial path contains nodes, and each path contains... There are 10 nodes, each corresponding to a grid cell coordinate, and... Each node is distributed based on the tunnel's central axis. Each individual in the population represents a potential path. Step 5.2.2: Exploration phase, calculate the fitness value of each initial path according to the fitness function; Specifically, the fitness function is: ; in, For path length, For path security, For multi-drone collaboration, The weighting coefficients and ; The path safety can be the reciprocal of the number of times a dangerous area is traversed. For example, if the Pth path traverses a dangerous area 4 times, then the path safety is 1 / 4. The multi-UAV cooperation can be the... i The reciprocal of the degree of path conflict between the first drone and all other drones, where, when the first... i The drone and the first j At the same time, drones t When the distance between path nodes is less than the safety threshold R, it is considered a conflict. "Conflict degree" refers to the ratio of the number of conflicts to the total number of path nodes. Step 5.2.3: Sort the population individuals in ascending order according to their fitness values, and divide the population into calm group, follower group and panic group, and update the position of individuals in calm group, follower group and panic group respectively; Specifically, the calm group consists of individuals with the highest fitness, responsible for guiding the search direction. The follower group consists of individuals with moderate fitness, mimicking the behavior of the calm group. The panicked group consists of individuals with low fitness, conducting random exploration. These three groups can be divided into three groups at a ratio of 15%, 35%, and 50%, respectively. The follower group moves with the calm group towards a safe area, storing the best-performing individuals from the exploration phase in an elite pool as potential safe exits.
[0027] The formula for updating an individual's location is: ; in, Indicates the first t The position of the i-th drone in the next iteration. Indicates the first t At the +1st iteration, the... i The location of the drone. Indicates the first t During the nth iteration The speed of the drone For inertial weights, As a learning factor, for Random numbers, For individuals Best historical position This is the globally optimal position.
[0028] Meanwhile, an adaptive weighting mechanism is introduced during path updates: ; in, These are the maximum and minimum values of the inertia weight. This represents the maximum number of iterations. Step 5.2.4: During the utilization phase, iteratively update the position of each individual in the elite pool until the maximum number of iterations is reached. Or the rate of change of fitness value is less than the threshold. When the optimal path is reached, output the optimal path. ; Optionally, the Levy flight strategy is used to update the position of each individual in the elite pool, and the individual position update formula is: ; Among them, inertia weight , For individuals Best historical position The globally optimal position; By updating path nodes through the Levy flight strategy, the global search capability of ESC path planning can be improved, enabling path planning to avoid getting stuck in local optima in areas with many dynamic obstacles, narrow tunnels, and local risk concentrations.
[0029] Step 6: For the optimal path of different UAVs, construct a spatiotemporal conflict matrix, use the spatiotemporal conflict matrix to detect whether there are conflicting paths, and adjust the conflicting paths accordingly; Optionally, the optimal path for each drone. ,in, nGiven the number of drones, conflicting paths are detected based on existing conflict detection and resolution algorithms, and conflicting paths can be adjusted by inserting waiting nodes or adjusting the path order.
[0030] Let the paths of the two drones be respectively and ,and The drones are assigned their respective serial numbers; a spatiotemporal conflict matrix is established between the optimal paths of the two drones. : ; in, Indicates drone i In its path u The three-dimensional coordinates of each node Indicates drone j In its path v The three-dimensional coordinates of each node. The spatial conflict distance threshold, For drones i Reaching its path at time points u To indicate drones j The path to reach it v The time of each node The time difference threshold, To indicate drones i The first in the path sequence u 1 path node To indicate drones j The first in the path sequence v One path node; like This indicates that there is a conflict between the two paths, and the drone is located before the conflict node. Insert a waiting node, with a waiting time of [time value missing]. Second, In the formula, Indicates time interval, thus through Adjust conflicting paths; The "conflict node" refers to a path location point in the same time series where the path nodes of two or more UAVs are located in the same grid cell or the distance between them is less than a preset safety threshold R. In other words, the spatial-temporal point meets the potential collision conditions and requires avoidance or adjustment operations.
[0031] Step 7: Calculate the environmental change index in real time. If the environmental change index exceeds the change threshold, repeat steps 1-6 above to achieve dynamic path adjustment. Specifically, the environmental change index is calculated in real time using the following formula. : ; in, To add more dynamic obstacles, This represents the total number of obstacles. When new dynamic obstacles or environmental information updates are detected, steps 1-6 above are re-executed to achieve dynamic path adjustment.
[0032] Example 2: This example provides a multi-UAV dynamic path planning device based on escape optimization support vector machine, implemented according to a multi-UAV dynamic path planning method based on escape optimization support vector machine. The device includes: The acquisition module is used to acquire environmental information of tunnels and underground spaces using drones, divide the tunnel space into three-dimensional grid cells, and establish an environmental information database based on the attributes of the grid cells. The extraction module is used to extract feature vectors of grid cells from the environmental information database and construct a training set based on the feature vectors and environmental category labels. The model building module is used to select the radial basis kernel function to construct the SVM model, and optimize the penalty parameter of the SVM model through cross-validation on the training set to obtain a trained environment category prediction SVM model. The prediction module is used to input the real-time collected environmental information into the environmental category prediction SVM model and output the environmental category prediction results. The path output module determines the danger zone based on the environmental category prediction results and uses an escape optimization algorithm to obtain the optimal path for each UAV. The conflict handling module is used to construct a spatiotemporal conflict matrix for the optimal path of different UAVs, use the spatiotemporal conflict matrix to detect whether there are conflicting paths, and adjust the conflicting paths. The dynamic adjustment module is used to calculate the environmental change index in real time. If the environmental change index exceeds the change threshold, the route is replanned to achieve dynamic route adjustment.
[0033] The device in this application embodiment can be an electronic device or a component within an electronic device, such as an integrated circuit or a chip. The electronic device can be a terminal or other devices besides a terminal. For example, the electronic device can be a GPU BOX, mobile phone, tablet computer, laptop computer, PDA, in-vehicle electronic device, mobile internet device (MID), augmented reality (AR) / virtual reality (VR) device, robot, wearable device, ultra-mobile personal computer (UMPC), netbook, or personal digital assistant (PDA), etc. It can also be a server, network attached storage (NAS), personal computer (PC), television (TV), ATM, or self-service machine, etc. This application embodiment does not specifically limit the device.
[0034] The device in this application embodiment can be a device with an operating system. The operating system can be Android, Linux, Windows, or other possible operating systems; this application embodiment does not specifically limit it.
[0035] This application provides an electronic device, which includes a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the multi-UAV dynamic path planning method based on escape optimization support vector machine as described in Embodiment 1 above.
[0036] This application also provides a computer-readable storage medium storing a computer program / instructions thereon, which, when executed by a processor, implements the steps in the multi-UAV dynamic path planning method based on escape optimization support vector machine disclosed in Embodiment 1 of this application.
[0037] This application also provides a computer program product that, when run on an electronic device, enables the processor to execute the steps of the multi-UAV dynamic path planning method based on escape optimization support vector machine disclosed in Embodiment 1 of this application.
[0038] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.
[0039] This application describes embodiments with reference to flowchart illustrations and / or block diagrams of methods, apparatuses, electronic devices, and computer program products according to embodiments of this application. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing terminal device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal device, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.
[0040] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing terminal device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.
[0041] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal equipment, causing a series of operational steps to be performed on the computer or other programmable terminal equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable terminal equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.
[0042] Although preferred embodiments of the present application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the embodiments of the present application.
[0043] Finally, it should be noted that in this text, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes said element.
[0044] The above provides a detailed description of the multi-UAV dynamic path planning method and apparatus based on escape optimization support vector machine provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.
Claims
1. A dynamic path planning method for multiple unmanned aerial vehicles (UAVs) based on escape optimization support vector machines, characterized in that, The method includes: Using drones to acquire environmental information about tunnels and underground spaces, the tunnel space is divided into three-dimensional grid cells, and an environmental information database is established based on the attributes of the grid cells; Feature vectors of grid cells are extracted from the environmental information database, and a training set is constructed based on the feature vectors and environmental category labels. We select a radial basis kernel function to construct an SVM model, and optimize the penalty parameter of the SVM model through cross-validation on the training set to obtain a trained environment category prediction SVM model. The real-time collected environmental information is input into the environmental category prediction SVM model, and the environmental category prediction results are output. Hazardous areas are identified based on environmental category prediction results, and the optimal path for each drone is obtained using an escape optimization algorithm. For different UAVs, a spatiotemporal conflict matrix is constructed to detect whether there are conflicting paths and to adjust the conflicting paths. The environmental change index is calculated in real time. If the environmental change index exceeds the change threshold, the above steps are repeated to achieve dynamic path adjustment.
2. The method according to claim 1, characterized in that, The method of using drones to acquire environmental information about tunnels and underground spaces, dividing the tunnel space into three-dimensional grid cells, and establishing an environmental information database based on the attributes of the grid cells includes: By utilizing multiple sensors mounted on the drone, environmental information about the tunnel and underground space is acquired, and the data collected by each type of sensor is used to construct a set D; The collection area of the tunnel and underground space is divided into Each grid cell is associated with multiple attribute information, and each attribute information is an element in set D; An environmental information database is constructed based on grid cells and their attribute information. ,in, Indicates coordinates as Grid cells, express Related attribute information.
3. The method according to claim 2, characterized in that, The process of identifying hazardous areas based on environmental category prediction results and using an escape optimization algorithm to obtain the optimal path for each UAV includes: Based on the environmental category prediction results, grid cells classified as dynamic obstacles or high-risk environments are designated as hazardous areas. Based on the dangerous area, an escape optimization algorithm is used to update the drone path and plan the optimal path for each drone.
4. The method according to claim 3, characterized in that, The process of updating drone paths based on dangerous areas using an escape optimization algorithm to plan the optimal path for each drone includes: Initialize the ESC algorithm parameters and generate an initial population, which contains P initial paths randomly generated for each of the n UAVs, with each initial path representing one individual; During the exploration phase, the fitness value of each initial path is calculated based on the fitness function; Individuals in the population are sorted in ascending order based on their fitness values, and the population is divided into calm group, follower group, and panic group. The positions of individuals in the calm group, follower group, and panic group are updated respectively, and the individuals with the best performance in the search phase are added to the elite pool. During the utilization phase, the position of each individual in the elite pool is iteratively updated using the Levy flight strategy until the maximum number of iterations is reached. Or the rate of change of fitness value is less than the threshold. When the optimal path is reached, output the optimal path. .
5. The method according to claim 4, characterized in that, The calculation of the fitness value of each initial path based on the fitness function includes: Obtain the path length of the P initial paths for each drone. ; Determine the path safety of each initial path. ; Determining the multi-UAV cooperation of each initial path ; Will , and We perform weighted combinations to obtain the fitness values of each initial path.
6. The method according to claim 5, characterized in that, The optimal path for each UAV is determined by constructing a spatiotemporal conflict matrix. This matrix is then used to detect conflicting paths and adjust them accordingly. Construct the spatiotemporal conflict matrix between the optimal path of drone m and the optimal path of another drone n; Determine if the spatiotemporal conflict matrix is 1. If it is, it means that there is a conflict between the two paths, and proceed to the next step. Otherwise, output the optimal path. A waiting node is inserted before the conflict node for drone n, where the waiting node represents the waiting time.
7. A multi-UAV dynamic path planning device based on escape optimization support vector machine, characterized in that, The apparatus, implemented based on any one of claims 1-6, comprises: The acquisition module is used to acquire environmental information of tunnels and underground spaces using drones, divide the tunnel space into three-dimensional grid cells, and establish an environmental information database based on the attributes of the grid cells. The extraction module is used to extract feature vectors of grid cells from the environmental information database and construct a training set based on the feature vectors and environmental category labels. The model building module is used to select the radial basis kernel function to construct the SVM model, and optimize the penalty parameter of the SVM model through cross-validation on the training set to obtain a trained environment category prediction SVM model. The prediction module is used to input the real-time collected environmental information into the environmental category prediction SVM model and output the environmental category prediction results. The path output module determines the danger zone based on the environmental category prediction results and uses an escape optimization algorithm to obtain the optimal path for each UAV. The conflict handling module is used to construct a spatiotemporal conflict matrix for the optimal path of different UAVs, use the spatiotemporal conflict matrix to detect whether there are conflicting paths, and adjust the conflicting paths. The dynamic adjustment module is used to calculate the environmental change index in real time. If the environmental change index exceeds the change threshold, the route is replanned to achieve dynamic route adjustment.