Community inspection vehicle path planning method, equipment and medium

Abstract

The embodiment of the invention discloses a community inspection vehicle path planning method and device and a medium, and the method comprises the steps: inputting a community map picture into a pre-trained generative adversarial network model when an inspection vehicle carries out community inspection, and obtaining a path picture with a preliminary plan; executing a specified path planning algorithm from a starting point to a target point and from the target point to the starting point based on the path map picture with the preliminary planning, the starting point coordinate and the target point coordinate to obtain a first spanning tree and a second spanning tree; and obtaining a shortest path from the starting point to the target point based on the first spanning tree and the second spanning tree. According to the embodiment of the invention, preliminary planning is carried out by using the pre-trained generative adversarial network model, and the picture with the potential path can be quickly generated. The time and computing resource consumption for searching a large number of invalid paths in a traditional path planning method are reduced, and the path planning efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a method, device and medium for planning a path for a community patrol vehicle. Background Art

[0002] With the continuous development of science and technology, the construction of smart communities has gradually become the focus of people's attention. As an important mobile robot, the smart community inspection vehicle plays an important role in the safety inspection and management work in the community. In a smart community, the inspection vehicle needs to have efficient and accurate path planning capabilities to ensure that it can reach the destination quickly and safely and complete the inspection task.

[0003] Traditional path planning methods may generate a large number of invalid paths during the search process, which need to be eliminated. However, a lot of time and computing resources may be wasted when processing these invalid paths, which may affect the efficiency and accuracy of community inspections. Summary of the invention

[0004] One or more embodiments of this specification provide a community patrol vehicle path planning method, device and medium for solving the technical problems raised by the background technology.

[0005] One or more embodiments of this specification adopt the following technical solutions:

[0006] One or more embodiments of this specification provide a community inspection vehicle path planning method, the method comprising:

[0007] When the inspection vehicle conducts community inspections, the community map image is input into the pre-trained generative adversarial network model to obtain a path image with preliminary planning;

[0008] Based on the path map image with preliminary planning, the coordinates of the starting point and the coordinates of the target point, respectively executing a specified path planning algorithm from the starting point to the target point and from the target point to the starting point to obtain a first spanning tree and a second spanning tree;

[0009] Based on the first spanning tree and the second spanning tree, a shortest path from the starting point to the target point is obtained.

[0010] It should be noted that the embodiments of this specification have the following beneficial effects through the above contents:

[0011] Improve path planning efficiency: By using a pre-trained generative adversarial network model for preliminary planning, images with potential paths can be quickly generated. This reduces the time and computing resource consumption of searching for a large number of invalid paths in traditional path planning methods, and improves the efficiency of path planning.

[0012] Enhanced path accuracy: Combined with the specified path planning algorithm, a bidirectional path search is performed from the starting point and the target point to obtain the first spanning tree and the second spanning tree. This bidirectional search can more comprehensively consider the path possibilities, thereby improving the accuracy of finding the shortest path.

[0013] Adapting to complex community environments: This method can cope with various changes and uncertainties in community environments. The generative adversarial network model can learn the characteristics and patterns of the community, thereby providing reasonable preliminary path planning in different situations. The specified path planning algorithm can be further optimized according to actual conditions to ensure that the inspection vehicle can adapt to complex community layouts and dynamic environments.

[0014] Improve community inspection results: Efficient and accurate route planning helps inspection vehicles reach their destinations quickly and safely and complete inspection tasks. This can improve community safety and management efficiency, and detect and handle potential problems in a timely manner.

[0015] Save time and resources: Reduce the search and processing of invalid paths, saving time and computing resources. This is especially important for community inspection systems with limited resources, which can improve the overall performance and sustainability of the system.

[0016] Furthermore, before inputting the community map image into the pre-trained generative adversarial network model to obtain a path image with preliminary planning, the method also includes:

[0017] By randomly generating multiple map images as data samples;

[0018] Execute the A* algorithm from the starting point to the target point on each map image to obtain a map image with the shortest path marked;

[0019] The map image with the shortest path marked is input into a generative adversarial network model for training to obtain a trained generative adversarial network model.

[0020] It should be noted that the embodiments of this specification have the following beneficial effects through the above contents:

[0021] Improve the accuracy of path planning: By executing the A* algorithm on a large number of randomly generated map images, marking the shortest path, and inputting these map images with the shortest path marked into the generative adversarial network model for training, the model can learn the shortest path patterns and features under different map layouts. This helps to improve the accuracy of the generative adversarial network model in preliminary path planning and reduce the generation of invalid paths.

[0022] Enhance the generalization ability of the model: Using randomly generated map images for training can increase the diversity and complexity of the data, allowing the model to adapt to a variety of different community map layouts. This helps improve the generalization ability of the model, enabling it to more accurately generate preliminary planned paths when faced with actual community maps.

[0023] Improve the efficiency of path planning: The trained generative adversarial network model can quickly generate a path image with a preliminary plan, reducing the time and computing resource consumption of searching for a large number of invalid paths in traditional path planning methods. This helps to improve the efficiency of path planning, allowing inspection vehicles to reach their destinations faster and complete inspection tasks.

[0024] Promoting the development of smart communities: Efficient and accurate path planning is an important part of building smart communities. By improving the accuracy and efficiency of path planning, the safety and management efficiency of communities can be improved, thus promoting the development of smart communities.

[0025] Furthermore, after obtaining the map image with the shortest path marked, the method further includes:

[0026] The pixel points where the shortest path in the map image marking the shortest path is located are expanded and bolded.

[0027] It should be noted that the embodiments of this specification have the following beneficial effects through the above contents:

[0028] Improve the visualization of the path: By expanding and bolding the pixels where the shortest path is located, the shortest path can be made more obvious and prominent on the map image. This helps inspectors or related personnel to see the planned path more intuitively, improves the visualization of the path, and reduces the possibility of misunderstanding or misjudgment.

[0029] Enhanced path recognizability: The bolded shortest path is easier to identify and distinguish, especially in complex map environments. This helps improve the recognizability of the path, allowing inspection vehicles or other equipment to more accurately follow the planned path and reduce the risk of deviation from the path.

[0030] Emphasize the importance of the shortest path: Expanding and bolding can emphasize the importance of the shortest path and make it more prominent in the map image. This helps guide inspectors or related personnel to give priority to the shortest path, improving inspection efficiency and work results.

[0031] Facilitates path tracking and monitoring: The bolded shortest path is more visible on the map, making it easier to track and monitor the path. This is important for real-time monitoring of the inspection vehicle's driving trajectory, evaluating the effectiveness of path planning, and performing subsequent optimization and adjustments.

[0032] Furthermore, the specified path planning algorithm is the RRT* algorithm.

[0033] It should be noted that the embodiments of this specification have the following beneficial effects through the above contents:

[0034] Efficiency: The RRT* algorithm is a path planning algorithm based on random sampling, which can quickly search for feasible paths in complex environments. Compared with other traditional algorithms, the RRT* algorithm has advantages in computational efficiency and can find a better path in a shorter time.

[0035] Strong adaptability: The RRT* algorithm is applicable to a variety of different types of environments, including static and dynamic environments. It can handle the presence and changes of obstacles, as well as uncertain factors in the environment. This makes its application in complex environments such as smart communities more reliable.

[0036] Optimality: The RRT* algorithm can find a path close to the optimal one by continuously optimizing the path. It can avoid falling into the local optimal solution and improve the quality of the path by introducing random sampling and local optimization in the search process.

[0037] Scalability: The RRT* algorithm has good scalability and can be applied to large-scale environments and multi-robot systems. It can be combined with other algorithms or techniques to further improve the performance of path planning.

[0038] Furthermore, obtaining the shortest path from the starting point to the target point based on the first spanning tree and the second spanning tree includes:

[0039] When the first spanning tree meets the second spanning tree, or the distance between the first spanning tree and the second spanning tree is less than a set threshold, the first iteration is completed to obtain a first shortest path;

[0040] Based on a preset number of iterations, repeatedly executing the specified path planning algorithm from the starting point to the target point and from the target point to the starting point based on the path map image with the preliminary plan, the starting point coordinates and the target point coordinates, respectively, to obtain a first spanning tree and a second spanning tree; when the first spanning tree meets the second spanning tree, or the distance between the first spanning tree and the second spanning tree is less than a set threshold, obtaining multiple second shortest paths;

[0041] Based on the first shortest path and the plurality of second shortest paths, a shortest path from the starting point to the target point is obtained.

[0042] It should be noted that the embodiments of this specification have the following beneficial effects through the above contents:

[0043] Improve the accuracy of path planning: By executing the specified path planning algorithm multiple times, multiple shortest paths can be obtained. Based on these paths, the optimal path can be further analyzed and selected, thereby improving the accuracy of path planning.

[0044] Enhance the reliability of path planning: Through multiple iterations, the local optimal solution that may appear in a single planning can be avoided. The existence of multiple shortest paths increases the possibility of finding the global optimal solution, thereby enhancing the reliability of path planning.

[0045] Consider different starting points and target points: This method simultaneously performs path planning from the starting point and the target point, and obtains two spanning trees. This can take into account the path possibilities in different directions, provide more options, and help find a better shortest path.

[0046] Adapt to complex environments: In complex environments, the path may be affected by many factors, such as obstacles, terrain, etc. Through multiple iterations and generating multiple shortest paths, it can better adapt to these complex situations and find a more suitable path.

[0047] Improve efficiency: Although multiple iterations may increase some computing time, by finding multiple shortest paths and comparing them, time can be saved when finally selecting a path. This avoids the situation where multiple attempts at different paths may be required in a single planning, improving overall efficiency.

[0048] Flexibility: The method can be adjusted according to specific needs and environments. For example, parameters such as the number of iterations and thresholds can be set according to actual conditions to adapt to different scenarios and requirements.

[0049] Optimize path selection: Based on multiple shortest paths, further analysis and optimization can be performed. Other factors such as path smoothness and safety can be considered to comprehensively select the most appropriate path.

[0050] Scalability: This method has certain scalability and can be applied to more complex path planning problems, such as multiple target points, dynamic environments, etc. Through appropriate adjustments and extensions, it can meet the needs of different application scenarios.

[0051] Further, based on the path map image with preliminary planning, the coordinates of the starting point and the coordinates of the target point, respectively, a specified path planning algorithm is executed from the starting point to the target point, and from the target point to the starting point to obtain a first spanning tree and a second spanning tree, including:

[0052] Determine the preset point coordinates based on the starting point coordinates and the target point coordinates;

[0053] Based on the path map image with the preliminary plan, determining whether the preset point coordinates are within the obstacle;

[0054] If the preset point coordinates are not within the obstacle, a specified path planning algorithm is executed from the starting point coordinates to the preset point coordinates and from the target point coordinates to the preset point coordinates to obtain the first spanning tree and the second spanning tree.

[0055] It should be noted that the embodiments of this specification have the following beneficial effects through the above contents:

[0056] Improved path planning accuracy: By determining the coordinates of the preset points and checking whether they are within obstacles, the path planning algorithm can be prevented from making invalid calculations within obstacles. This helps improve the accuracy of path planning and ensures that the generated path does not pass through obstacles.

[0057] Reduce the amount of calculation: If the coordinates of the preset point are within the obstacle, there is no need to execute the path planning algorithm from the starting point or the target point to that point. This can reduce unnecessary calculations and improve the efficiency of the algorithm.

[0058] Increase the feasibility of the path: By excluding the coordinates of the preset points that are within obstacles, it can be ensured that the generated path is feasible and will not be blocked by obstacles. This increases the feasibility and practicality of the path.

[0059] Improve system response speed: Reducing unnecessary calculations can speed up path planning and improve system response speed. This is very important for application scenarios with high real-time requirements, such as autonomous driving or robot navigation.

[0060] Enhance system stability: Avoiding path planning within obstacles can reduce the possibility of algorithm errors or anomalies and enhance system stability and reliability.

[0061] Optimize path selection: By properly selecting preset point coordinates, you can guide the path planning algorithm to generate a better path. For example, you can choose to set preset points in open areas or key locations to obtain a shorter, smoother, or more specific path.

[0062] Improve user experience: Accurate and feasible path planning can provide a better user experience. Users can get satisfactory path planning results more quickly, improve work efficiency or enjoy better services.

[0063] Furthermore, if the preset point coordinates are within an obstacle, the method further includes:

[0064] A specified path planning algorithm is executed from the starting point coordinates to the target point coordinates to obtain a corresponding path planning.

[0065] It should be noted that the embodiments of this specification have the following beneficial effects through the above contents:

[0066] Ensure the continuity of path planning: Even if the preset point coordinates are within obstacles, the corresponding path planning can still be obtained by executing the specified path planning algorithm directly from the starting point coordinates to the target point coordinates. This ensures that the path planning will not be interrupted when encountering obstacles, but can continue, providing a continuous solution.

[0067] Increase the flexibility of path planning: This method increases the flexibility of path planning. When the preset point is not available, it can be planned directly from the starting point to the target point, avoiding the situation where a feasible path cannot be found due to the limitation of the preset point. This can better adapt to different environments and situations and improve the success rate of path planning.

[0068] Improve the efficiency of path planning: Directly planning the path from the starting point to the target point can avoid additional calculations and judgments at the preset points. This reduces the amount of calculation and improves the efficiency of path planning, especially in complex environments or when path planning needs to be generated quickly.

[0069] Furthermore, the preset point coordinates include middle point coordinates, one-third point coordinates and two-thirds point coordinates.

[0070] One or more embodiments of this specification provide a community inspection vehicle path planning device, including:

[0071] at least one processor; and,

[0072] a memory communicatively connected to the at least one processor; wherein,

[0073] The memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to:

[0074] When the inspection vehicle conducts community inspections, the community map image is input into the pre-trained generative adversarial network model to obtain a path image with preliminary planning;

[0075] Based on the path map image with preliminary planning, the coordinates of the starting point and the coordinates of the target point, respectively executing a specified path planning algorithm from the starting point to the target point and from the target point to the starting point to obtain a first spanning tree and a second spanning tree;

[0076] Based on the first spanning tree and the second spanning tree, a shortest path from the starting point to the target point is obtained.

[0077] One or more embodiments of this specification provide a non-volatile computer storage medium storing computer executable instructions, which can achieve the following when executed by a computer:

[0078] When the inspection vehicle conducts community inspections, the community map image is input into the pre-trained generative adversarial network model to obtain a path image with preliminary planning;

[0079] Based on the path map image with preliminary planning, the coordinates of the starting point and the coordinates of the target point, respectively executing a specified path planning algorithm from the starting point to the target point and from the target point to the starting point to obtain a first spanning tree and a second spanning tree;

[0080] Based on the first spanning tree and the second spanning tree, a shortest path from the starting point to the target point is obtained.

[0081] At least one of the above technical solutions adopted in the embodiments of this specification can achieve the following beneficial effects:

[0082] Improve path planning efficiency: By using a pre-trained generative adversarial network model for preliminary planning, images with potential paths can be quickly generated. This reduces the time and computing resource consumption of searching for a large number of invalid paths in traditional path planning methods, and improves the efficiency of path planning.

[0083] Enhanced path accuracy: Combined with the specified path planning algorithm, a bidirectional path search is performed from the starting point and the target point to obtain the first spanning tree and the second spanning tree. This bidirectional search can more comprehensively consider the path possibilities, thereby improving the accuracy of finding the shortest path.

[0084] Adapting to complex community environments: This method can cope with various changes and uncertainties in community environments. The generative adversarial network model can learn the characteristics and patterns of the community, thereby providing reasonable preliminary path planning in different situations. The specified path planning algorithm can be further optimized according to actual conditions to ensure that the inspection vehicle can adapt to complex community layouts and dynamic environments.

[0085] Improve community inspection results: Efficient and accurate route planning helps inspection vehicles reach their destinations quickly and safely and complete inspection tasks. This can improve community safety and management efficiency, and detect and handle potential problems in a timely manner.

[0086] Save time and resources: Reduce the search and processing of invalid paths, saving time and computing resources. This is especially important for community inspection systems with limited resources, which can improve the overall performance and sustainability of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0087] In order to more clearly illustrate the technical solutions in the embodiments of this specification or the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art description. Obviously, the drawings described below are only some embodiments recorded in this specification. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative labor. In the drawings:

[0088] Figure 1 A flowchart of a community inspection vehicle path planning method provided for one or more embodiments of this specification;

[0089] Figure 2 A schematic diagram of an initial shortest path provided for one or more embodiments of this specification;

[0090] Figure 3 A schematic diagram of the shortest path initially planned for one or more embodiments of this specification;

[0091] Figure 4 A schematic diagram of the structure of a community patrol vehicle path planning device provided for one or more embodiments of this specification. DETAILED DESCRIPTION

[0092] The embodiments of this specification provide a community patrol vehicle path planning method, device and medium.

[0093] In order to enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the drawings in the embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this specification, not all of the embodiments. Based on the embodiments of this specification, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of this specification.

[0094] Figure 1 A schematic diagram of a process flow of a community inspection vehicle path planning method provided in one or more embodiments of this specification, which can be executed by a community inspection vehicle path planning system. Certain input parameters or intermediate results in the process allow manual intervention and adjustment to help improve accuracy.

[0095] The method steps of the embodiment of this specification are as follows:

[0096] S102, when the inspection vehicle conducts community inspection, the community map image is input into the pre-trained generative adversarial network model to obtain a path image with preliminary planning.

[0097] In the embodiment of this specification, before inputting the community map image into the pre-trained generative adversarial network model to obtain the path image with preliminary planning, it is necessary to train the generative adversarial network model, as follows:

[0098] A plurality of map images are randomly generated as data samples; an A* algorithm from a starting point to a target point is executed on each map image to obtain a map image with a shortest path marked; the map image with the shortest path marked is input into a generative adversarial network model for training to obtain a trained generative adversarial network model.

[0099] It should be noted that the above training to obtain a generative adversarial network model can be implemented through the following specific implementation plan:

[0100] Data preparation: Generate multiple map images randomly. You can use a random map generation algorithm or an existing map dataset. For each map image, determine the coordinates of the starting point and the target point. Execute the A* algorithm from the starting point to the target point on each map image to obtain the shortest path. Mark the shortest path on the map image, for example, using different colors or lines.

[0101] GAN model training: Build a GAN model, including a generator and a discriminator. Input the map images of the labeled shortest paths into the GAN model as training data. Train the generator to learn to generate map images similar to the real labeled shortest paths. Train the discriminator to distinguish between the real labeled shortest paths map images and the images generated by the generator. Through continuous iterative training, the performance of the generator and the discriminator is gradually improved.

[0102] Model evaluation: Use test data to evaluate the trained generative adversarial network model. You can calculate the similarity or error between the generated map image and the real labeled shortest path map image. Adjust and optimize the model based on the evaluation results.

[0103] It should be noted that the embodiments of this specification have the following beneficial effects through the above contents:

[0104] Improve the accuracy of path planning: By executing the A* algorithm on a large number of randomly generated map images, marking the shortest path, and inputting these map images with the shortest path marked into the generative adversarial network model for training, the model can learn the shortest path patterns and features under different map layouts. This helps to improve the accuracy of the generative adversarial network model in preliminary path planning and reduce the generation of invalid paths.

[0105] Enhance the generalization ability of the model: Using randomly generated map images for training can increase the diversity and complexity of the data, allowing the model to adapt to a variety of different community map layouts. This helps improve the generalization ability of the model, enabling it to more accurately generate preliminary planned paths when faced with actual community maps.

[0106] Improve the efficiency of path planning: The trained generative adversarial network model can quickly generate a path image with a preliminary plan, reducing the time and computing resource consumption of searching for a large number of invalid paths in traditional path planning methods. This helps to improve the efficiency of path planning, allowing inspection vehicles to reach their destinations faster and complete inspection tasks.

[0107] Promoting the development of smart communities: Efficient and accurate path planning is an important part of building smart communities. By improving the accuracy and efficiency of path planning, the safety and management efficiency of communities can be improved, thus promoting the development of smart communities.

[0108] Furthermore, after obtaining the map image with the shortest path marked thereon, the pixel points where the shortest path is located in the map image with the shortest path marked thereon may be expanded and thickened.

[0109] It should be noted that the embodiments of this specification have the following beneficial effects through the above contents:

[0110] Improve the visualization of the path: By expanding and bolding the pixels where the shortest path is located, the shortest path can be made more obvious and prominent on the map image. This helps inspectors or related personnel to see the planned path more intuitively, improves the visualization of the path, and reduces the possibility of misunderstanding or misjudgment.

[0111] Enhanced path recognizability: The bolded shortest path is easier to identify and distinguish, especially in complex map environments. This helps improve the recognizability of the path, allowing inspection vehicles or other equipment to more accurately follow the planned path and reduce the risk of deviation from the path.

[0112] Emphasize the importance of the shortest path: Expanding and bolding can emphasize the importance of the shortest path and make it more prominent in the map image. This helps guide inspectors or related personnel to give priority to the shortest path, improving inspection efficiency and work results.

[0113] Facilitates path tracking and monitoring: The bolded shortest path is more visible on the map, making it easier to track and monitor the path. This is important for real-time monitoring of the inspection vehicle's driving trajectory, evaluating the effectiveness of path planning, and performing subsequent optimization and adjustments.

[0114] S104, based on the path map image with preliminary planning, the starting point coordinates and the target point coordinates, respectively, from the starting point to the target point, and from the target point to the starting point, execute a specified path planning algorithm to obtain a first spanning tree and a second spanning tree.

[0115] In the embodiment of the present specification, the specified path planning algorithm is the RRT* algorithm. The first spanning tree and the second spanning tree may be planned paths obtained based on the RRT* algorithm.

[0116] It should be noted that the embodiments of this specification have the following beneficial effects through the above contents:

[0117] Efficiency: The RRT* algorithm is a path planning algorithm based on random sampling, which can quickly search for feasible paths in complex environments. Compared with other traditional algorithms, the RRT* algorithm has advantages in computational efficiency and can find a better path in a shorter time.

[0118] Strong adaptability: The RRT* algorithm is applicable to a variety of different types of environments, including static and dynamic environments. It can handle the presence and changes of obstacles, as well as uncertain factors in the environment. This makes its application in complex environments such as smart communities more reliable.

[0119] Optimality: The RRT* algorithm can find a path close to the optimal one by continuously optimizing the path. It can avoid falling into the local optimal solution and improve the quality of the path by introducing random sampling and local optimization in the search process.

[0120] Scalability: The RRT* algorithm has good scalability and can be applied to large-scale environments and multi-robot systems. It can be combined with other algorithms or techniques to further improve the performance of path planning.

[0121] S106: Obtain the shortest path from the starting point to the target point based on the first spanning tree and the second spanning tree.

[0122] In an embodiment of the present specification, the first iteration can be completed to obtain the first shortest path when the first spanning tree meets the second spanning tree, or the distance between the first spanning tree and the second spanning tree is less than a set threshold; based on a preset number of iterations, the specified path planning algorithm is repeatedly executed from the starting point to the target point, and from the target point to the starting point, respectively, based on the path map image with preliminary planning, the starting point coordinates and the target point coordinates, to obtain the first spanning tree and the second spanning tree; when the first spanning tree meets the second spanning tree, or the distance between the first spanning tree and the second spanning tree is less than a set threshold, multiple second shortest paths are obtained; based on the first shortest path and the multiple second shortest paths, the shortest path from the starting point to the target point is obtained.

[0123] It should be noted that the shortest path from the starting point to the target point can be obtained by the following specific implementation scheme:

[0124] Step 1: Get a map image with a preliminary plan of the path, the coordinates of the starting point, and the coordinates of the destination point. Set the number of iterations.

[0125] Step 2: Based on the coordinates of the starting point and the target point, execute the specified path planning algorithm from the starting point to the target point and from the target point to the starting point to obtain the first spanning tree and the second spanning tree. Determine whether the first spanning tree and the second spanning tree meet, or whether the distance between them is less than the set threshold. If the conditions are met, complete the first iteration and obtain the first shortest path.

[0126] Step 3: Repeat step 2 based on the preset number of iterations. In each iteration, a new first spanning tree and a new second spanning tree are obtained, and it is determined whether they meet or whether the distance is less than a threshold. If the conditions are met, multiple second shortest paths are obtained.

[0127] Step 4: Based on the first shortest path and the plurality of second shortest paths, determine the shortest path from the starting point to the target point. A comparison algorithm or other methods may be used to select the shortest path.

[0128] Step 5: Output the shortest path from the starting point to the target point.

[0129] It should be noted that the embodiments of this specification have the following beneficial effects through the above contents:

[0130] Improve the accuracy of path planning: By executing the specified path planning algorithm multiple times, multiple shortest paths can be obtained. Based on these paths, the optimal path can be further analyzed and selected, thereby improving the accuracy of path planning.

[0131] Enhance the reliability of path planning: Through multiple iterations, the local optimal solution that may appear in a single planning can be avoided. The existence of multiple shortest paths increases the possibility of finding the global optimal solution, thereby enhancing the reliability of path planning.

[0132] Consider different starting points and target points: This method simultaneously performs path planning from the starting point and the target point, and obtains two spanning trees. This can take into account the path possibilities in different directions, provide more options, and help find a better shortest path.

[0133] Adapt to complex environments: In complex environments, the path may be affected by many factors, such as obstacles, terrain, etc. Through multiple iterations and generating multiple shortest paths, it can better adapt to these complex situations and find a more suitable path.

[0134] Improve efficiency: Although multiple iterations may increase some computing time, by finding multiple shortest paths and comparing them, time can be saved when finally selecting a path. This avoids the situation where multiple attempts at different paths may be required in a single planning, improving overall efficiency.

[0135] Flexibility: The method can be adjusted according to specific needs and environments. For example, parameters such as the number of iterations and thresholds can be set according to actual conditions to adapt to different scenarios and requirements.

[0136] Optimize path selection: Based on multiple shortest paths, further analysis and optimization can be performed. Other factors such as path smoothness and safety can be considered to comprehensively select the most appropriate path.

[0137] Scalability: This method has certain scalability and can be applied to more complex path planning problems, such as multiple target points, dynamic environments, etc. Through appropriate adjustments and extensions, it can meet the needs of different application scenarios.

[0138] Further, based on the path map image with preliminary planning, the starting point coordinates and the target point coordinates, respectively, from the starting point to the target point, and from the target point to the starting point, a designated path planning algorithm is executed to obtain the first spanning tree and the second spanning tree, and the preset point coordinates can be determined based on the starting point coordinates and the target point coordinates; based on the path map image with preliminary planning, it is determined whether the preset point coordinates are within an obstacle; if the preset point coordinates are not within an obstacle, the first spanning tree and the second spanning tree are obtained by executing the designated path planning algorithm from the starting point coordinates to the preset point coordinates, and from the target point coordinates to the preset point coordinates. The preset point coordinates of the embodiments of this specification may include middle point coordinates, one-third point coordinates and two-thirds point coordinates.

[0139] It should be noted that the first spanning tree and the second spanning tree may be obtained by the following specific implementation scheme:

[0140] Determine the preset point coordinates: Calculate the coordinates of the middle point, one-third point, and two-thirds point based on the coordinates of the starting point and the target point. The middle point coordinates can be determined by the average value of the coordinates of the starting point and the target point.

[0141] Determine whether the preset point coordinates are within obstacles: For each preset point coordinate, check it on the path map image with the preliminary plan. Image recognition or other methods can be used to determine whether the preset point coordinates overlap with obstacles.

[0142] Execute the specified path planning algorithm: If the preset point coordinates are not within the obstacle, execute the specified path planning algorithm from the starting point coordinates to the preset point coordinates, and from the target point coordinates to the preset point coordinates. The path planning algorithm can be selected according to specific needs, such as A* algorithm, Dijkstra algorithm, etc. After executing the path planning algorithm, the first spanning tree and the second spanning tree are obtained.

[0143] Subsequent processing: According to the first spanning tree and the second spanning tree, further path optimization, path selection or other related processing can be performed.

[0144] It should be noted that the embodiments of this specification have the following beneficial effects through the above contents:

[0145] Improved path planning accuracy: By determining the coordinates of the preset points and checking whether they are within obstacles, the path planning algorithm can be prevented from making invalid calculations within obstacles. This helps improve the accuracy of path planning and ensures that the generated path does not pass through obstacles.

[0146] Reduce the amount of calculation: If the coordinates of the preset point are within the obstacle, there is no need to execute the path planning algorithm from the starting point or the target point to that point. This can reduce unnecessary calculations and improve the efficiency of the algorithm.

[0147] Increase the feasibility of the path: By excluding the coordinates of the preset points that are within obstacles, it can be ensured that the generated path is feasible and will not be blocked by obstacles. This increases the feasibility and practicality of the path.

[0148] Improve system response speed: Reducing unnecessary calculations can speed up path planning and improve system response speed. This is very important for application scenarios with high real-time requirements, such as autonomous driving or robot navigation.

[0149] Enhance system stability: Avoiding path planning within obstacles can reduce the possibility of algorithm errors or anomalies and enhance system stability and reliability.

[0150] Optimize path selection: By properly selecting preset point coordinates, you can guide the path planning algorithm to generate a better path. For example, you can choose to set preset points in open areas or key locations to obtain a shorter, smoother, or more specific path.

[0151] Improve user experience: Accurate and feasible path planning can provide a better user experience. Users can get satisfactory path planning results more quickly, improve work efficiency or enjoy better services.

[0152] Furthermore, if the preset point coordinates are within an obstacle, a specified path planning algorithm may be executed from the starting point coordinates to the target point coordinates to obtain a corresponding path planning.

[0153] It should be noted that the embodiments of this specification have the following beneficial effects through the above contents:

[0154] Ensure the continuity of path planning: Even if the preset point coordinates are within obstacles, the corresponding path planning can still be obtained by executing the specified path planning algorithm directly from the starting point coordinates to the target point coordinates. This ensures that the path planning will not be interrupted when encountering obstacles, but can continue, providing a continuous solution.

[0155] Increase the flexibility of path planning: This method increases the flexibility of path planning. When the preset point is not available, it can be planned directly from the starting point to the target point, avoiding the situation where a feasible path cannot be found due to the limitation of the preset point. This can better adapt to different environments and situations and improve the success rate of path planning.

[0156] Improve the efficiency of path planning: Directly planning the path from the starting point to the target point can avoid additional calculations and judgments at the preset points. This reduces the amount of calculation and improves the efficiency of path planning, especially in complex environments or when path planning needs to be generated quickly.

[0157] It should be noted that the smart community inspection vehicle is a mobile robot that integrates a variety of advanced technologies and is mainly used for security inspections and management work within the community. It has functions such as license plate recognition and positioning, data transmission and processing, AI recognition and analysis, path navigation and planning. Path planning is one of the core technologies for community inspection vehicles to achieve navigation and complete inspection tasks. Path planning starts from a set starting point and, taking into account various factors in the community, calculates and plans an effective path to the destination, while ensuring that the path meets certain performance indicators.

[0158] Generative Adversarial Networks (GAN) is a deep learning model that is trained by adversarial networks. Its main goal is to generate new data that is highly similar to real data through the adversarial learning process between the two networks.

[0159] The main task of the generative network is to generate new data samples, while the discriminative network is responsible for judging whether these data samples are real, that is, identifying whether the results generated by the generative network are real data or generated fakes. During the training process, the generative network will constantly try to generate more realistic data that can deceive the discriminative network, while the discriminative network will strive to improve its own discrimination ability and identify the fakes generated by the generative network. This adversarial process allows the two networks to make progress together, thereby generating more realistic and diverse data.

[0160] The RRT* algorithm is an asymptotically optimal path planning based on Rapidly-exploring Random Trees (RRT). The algorithm completes path planning through steps such as random sampling, tree growth, and path search. As the number of RRT* iterations increases, the path will gradually shorten. However, when the path planning map environment scene is large or the map environment is complex, the random sampling in the early stage of the RRT* algorithm will lead to reduced search efficiency.

[0161] The embodiments of this specification provide a community patrol vehicle path planning method based on a generative adversarial network and an improved RRT*, so as to solve the problem of low efficiency when using the RRT* algorithm for path planning.

[0162] The embodiments of this specification are implemented through the following technical solutions:

[0163] A community inspection vehicle path planning method based on a generative adversarial network and improved RRT*, the implementation method comprising the following steps:

[0164] Step 1: Feed the map image into the generative adversarial network model and output a path image with preliminary planning;

[0165] Step 2: Input the path map image with preliminary planning, the coordinates of the starting point and the target point, and execute the RRT* algorithm from the starting point to the target point and from the target point to the starting point respectively;

[0166] Step 3: When two spanning trees meet or the distance is less than the set threshold, the first iteration ends, and a path is obtained and recorded as the shortest path;

[0167] Step 4: Continue to execute steps 2 and 3. If the generated path is smaller than the shortest path, update the shortest path.

[0168] Step 5: After all iterations are completed, the planned path is output.

[0169] A community inspection vehicle path planning method based on generative adversarial networks and improved RRT*, wherein step 1 specifically includes the following steps:

[0170] Step 1.1: Generate a large number of map images randomly as data samples;

[0171] Step 1.2: Execute the A* algorithm from the starting point to the target point on each generated map image to obtain the shortest path, such as Figure 2 The schematic diagram of the initial shortest path shown in FIG. 1 is shown in FIG. 1 , and the pixels on the shortest path are expanded and bolded;

[0172] Step 1.3: The randomly generated map images are sent to the generative adversarial network model, and the images with the shortest path of the A* algorithm are sent to the discriminant network for training to obtain the trained generative adversarial network;

[0173] Step 1.4: Send the planning map image to the trained adversarial network and output an image with a path, such as Figure 3 A schematic diagram of the shortest path initially planned is shown.

[0174] A community inspection vehicle path planning method based on generative adversarial networks and improved RRT*, wherein step 2 specifically includes the following steps:

[0175] Step 2.1: Input map M, RRT*iteration number K, starting point coordinates (x s ,y s ) and the target point coordinates (x g ,y g );

[0176] Step 2.2: Send the map M into the generative adversarial network trained in step 1 to obtain the map M′.

[0177] Step 2.3: Calculate the coordinates of the middle point Perform collision detection to determine whether the point (x1, y1) is within the obstacle. If it is within the obstacle, calculate Continue collision detection and calculate if it is within the obstacle

[0178]

[0179] Step 2.4: Get the middle point in step 2.3, and execute the operation from the starting point coordinate (x s ,y s ) to the middle point and from the target point target point coordinates (x g ,y g) to the middle point. If the middle point is not obtained in step 2.3, execute the two RRTs from the starting point coordinates (x s ,y s ) to the target point coordinate (x g ,y g ) is a path planning.

[0180] Step 2.5 Iterate the RRT* algorithm K times and output the shortest path.

[0181] The beneficial effects of the embodiments of this specification are:

[0182] The embodiments of this specification perform map preprocessing based on a generative adversarial network, and use a large amount of map data to execute the A* algorithm of the shortest path as a training set of the generative adversarial network to train the generative adversarial network, so that it can learn the A* algorithm to generate the shortest path.

[0183] The embodiments of this specification aim at the problem of low efficiency of RRT*. Generative adversarial network is used to initially form the path sampling range, accelerate the execution speed of RRT* algorithm, and bidirectionally execute RRT* algorithm based on the principle that the straight line between two points is the shortest, further accelerating the efficiency of path planning.

[0184] The embodiment of this specification is based on a large number of map images generated by the A* algorithm to train a generative adversarial network to learn the path generated by the A* algorithm, and at the same time selects the sampling points of the RRT* algorithm in the generated path to reduce the problem of low efficiency of random sampling in the initial stage of the RRT* algorithm. The embodiment of this specification also selects a bidirectional RRT* algorithm in combination with the map environment to improve the efficiency of path planning.

[0185] Figure 4 A schematic diagram of a community inspection vehicle path planning device provided for one or more embodiments of this specification includes:

[0186] at least one processor; and,

[0187] a memory communicatively connected to the at least one processor; wherein,

[0188] The memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to:

[0189] When the inspection vehicle conducts community inspections, the community map image is input into the pre-trained generative adversarial network model to obtain a path image with preliminary planning;

[0190] Based on the path map image with preliminary planning, the coordinates of the starting point and the coordinates of the target point, respectively executing a specified path planning algorithm from the starting point to the target point and from the target point to the starting point to obtain a first spanning tree and a second spanning tree;

[0191] Based on the first spanning tree and the second spanning tree, a shortest path from the starting point to the target point is obtained.

[0192] One or more embodiments of this specification provide a non-volatile computer storage medium storing computer executable instructions, which can achieve the following when executed by a computer:

[0193] When the inspection vehicle conducts community inspections, the community map image is input into the pre-trained generative adversarial network model to obtain a path image with preliminary planning;

[0194] Based on the path map image with preliminary planning, the coordinates of the starting point and the coordinates of the target point, respectively executing a specified path planning algorithm from the starting point to the target point and from the target point to the starting point to obtain a first spanning tree and a second spanning tree;

[0195] Based on the first spanning tree and the second spanning tree, a shortest path from the starting point to the target point is obtained.

[0196] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device and non-volatile computer storage medium embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiments.

Claims

1. A community inspection vehicle path planning method, characterized in that: The method comprises: When the inspection vehicle conducts community inspections, the community map image is input into the pre-trained generative adversarial network model to obtain a path image with preliminary planning; Based on the path map image with preliminary planning, the coordinates of the starting point and the coordinates of the target point, respectively executing a specified path planning algorithm from the starting point to the target point and from the target point to the starting point to obtain a first spanning tree and a second spanning tree; Based on the first spanning tree and the second spanning tree, a shortest path from the starting point to the target point is obtained.

2. The method according to claim 1, characterized in that Before inputting the community map image into the pre-trained generative adversarial network model to obtain a path image with preliminary planning, the method further includes: By randomly generating multiple map images as data samples; Execute the A* algorithm from the starting point to the target point on each map image to obtain a map image with the shortest path marked; The map image with the shortest path marked is input into a generative adversarial network model for training to obtain a trained generative adversarial network model.

3. The method according to claim 2, characterized in that After obtaining the map image with the shortest path marked, the method further includes: The pixel points where the shortest path in the map image marking the shortest path is located are expanded and bolded.

4. The method according to claim 1, characterized in that: The specified path planning algorithm is the RRT* algorithm.

5. The method according to claim 4, characterized in that The obtaining the shortest path from the starting point to the target point based on the first spanning tree and the second spanning tree includes: When the first spanning tree meets the second spanning tree, or the distance between the first spanning tree and the second spanning tree is less than a set threshold, the first iteration is completed to obtain a first shortest path; Based on a preset number of iterations, repeatedly executing the specified path planning algorithm from the starting point to the target point and from the target point to the starting point based on the path map image with the preliminary plan, the starting point coordinates and the target point coordinates, respectively, to obtain a first spanning tree and a second spanning tree; when the first spanning tree meets the second spanning tree, or the distance between the first spanning tree and the second spanning tree is less than a set threshold, obtaining multiple second shortest paths; Based on the first shortest path and the plurality of second shortest paths, a shortest path from the starting point to the target point is obtained.

6. The method according to claim 4, characterized in that Based on the path map image with preliminary planning, the coordinates of the starting point and the coordinates of the target point, respectively executing a specified path planning algorithm from the starting point to the target point and from the target point to the starting point to obtain a first spanning tree and a second spanning tree, including: Determine the preset point coordinates based on the starting point coordinates and the target point coordinates; Based on the path map image with the preliminary plan, determining whether the preset point coordinates are within the obstacle; If the preset point coordinates are not within the obstacle, a specified path planning algorithm is executed from the starting point coordinates to the preset point coordinates and from the target point coordinates to the preset point coordinates to obtain the first spanning tree and the second spanning tree.

7. The method according to claim 6, characterized in that If the preset point coordinates are within the obstacle, the method further includes: A specified path planning algorithm is executed from the starting point coordinates to the target point coordinates to obtain a corresponding path planning.

8. The method according to claim 6, characterized in that The preset point coordinates include middle point coordinates, one-third point coordinates and two-thirds point coordinates.

9. A community patrol vehicle path planning device, characterized in that: include: at least one processor; as well as, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to: When the inspection vehicle conducts community inspections, the community map image is input into the pre-trained generative adversarial network model to obtain a path image with preliminary planning; Based on the path map image with preliminary planning, the coordinates of the starting point and the coordinates of the target point, respectively executing a specified path planning algorithm from the starting point to the target point and from the target point to the starting point to obtain a first spanning tree and a second spanning tree; Based on the first spanning tree and the second spanning tree, a shortest path from the starting point to the target point is obtained.

10. A non-volatile computer storage medium, characterized in that: Computer executable instructions are stored, and when the computer executable instructions are executed by a computer, the following can be achieved: When the inspection vehicle conducts community inspections, the community map image is input into the pre-trained generative adversarial network model to obtain a path image with preliminary planning; Based on the path map image with preliminary planning, the coordinates of the starting point and the coordinates of the target point, respectively executing a specified path planning algorithm from the starting point to the target point and from the target point to the starting point to obtain a first spanning tree and a second spanning tree; Based on the first spanning tree and the second spanning tree, a shortest path from the starting point to the target point is obtained.

Images