Unmanned aerial vehicle cluster flight path planning method
Through the combination of deep learning model and B l ender system, the drone cluster flight path is automatically generated, which solves the problem of time-consuming and labor-consuming manual operation in the existing technology, and realizes an efficient and low-cost drone lighting performance path. planning.
Patent Information
- Application Number
- CN202411374707.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-29
- Publication Date
- 2024-12-10
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The path planning of existing drone lighting performances relies on manual operation, which is low in automation, time-consuming and labor-intensive, making it difficult to achieve simultaneous flight and dynamic styling display of hundreds or even thousands of drones.
The deep learning model is used to automatically generate the flight trajectory of the drone cluster, generate 3D models through the Stab l e Dreamfus i on system, and initialize the drone objects in B l end, and match the drone using the two-two approximate minimum total distance method. With the vertex of the 3D model, a linear flight path is generated, collision detection and path correction are performed, and the flight path file is output.
It realizes automated planning of drone cluster flight paths, improves efficiency and reduces difficulty, and can quickly generate flight paths that meet customer needs, meeting high timeliness and low cost requirements.
Smart Images
Figure CN119105553A_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the technical field of unmanned aerial vehicles, and in particular relates to a flight path planning method for a cluster of unmanned aerial vehicles. Background Art
[0002] With the advancement of drone technology, organizing large numbers of drone swarms to fly and carry out various military and civilian activities has also developed rapidly, especially civilian projects such as light shows have been widely used. The current mainstream path planning method for drone light shows is very primitive. The main process is completed manually by designers using 3D design software, with low automation and time-consuming and labor-intensive. This technology mainly describes a method that automatically generates a 3D model of spatial modeling according to customer needs and further automatically generates the flight path, light color, brightness and movement of all drones in the cluster based on these 3D models. The technology automatically generates the flight trajectory of drone clusters based on a deep learning large model and develops a good user interface. Users only need to enter a simple description through voice or text to automatically generate the desired flight path.
[0003] The spatial flight position of a drone is usually calculated by combining the longitude and latitude data obtained by the GNSS satellite navigation module with altitude data such as a barometer, and is generally described by a navigation coordinate system (WGS-84). The NED coordinate system is also a relative coordinate system that is used a lot in flight path planning. The N-north axis points to the north of the earth; the E-east axis points to the east of the earth; and the D-earth axis is perpendicular to the earth's surface and points downward. These three numbers determine its relative spatial position. When applied to drone performances or other flights with a smaller spatial range, the navigation coordinates can be approximately converted from a known navigation coordinate origin (whose longitude, latitude, and altitude are all known) based on the relative spatial coordinates NED (x, y, z). Therefore, the coordinates used in this article are all NED relative coordinate systems, and their positions are determined by the spatial origin. The program controls the automatic flight of a drone, generally based on the waypoint coordinate path planned in advance, and this waypoint can also be converted into a NED spatial coordinate point. Therefore, the flight path planning of drones can essentially be regarded as the planning of the (time + coordinate point) set in the NED space coordinate system, that is, the set of (time, x, y, z). The flight path of a drone is a file that stores the set data. If it is a single drone, the user can directly mark the spatial position of the waypoints through the ground station software to form a spatial flight path. However, when we are faced with a scene such as a light show that requires hundreds or even thousands of drones to fly at the same time, it is impossible to manually mark the coordinate waypoints and form various shapes. In the application of drone light shows, it is usually necessary for a large number of drones to form one or more 3D stereoscopic patterns in space, and appropriately match the LED light display on the drone to create a performance animation effect. The existing practice is to use 3D design software for modeling, construct the 3D model that the customer needs to perform, and parse these 3D models into the same number of vertex sets according to the number of drones. Each vertex represents a drone. When these drones light up in the air, they can perform the designed shape. The entire flight process is actually an animation playback process. Each shape is actually similar to the key frame of animation design. The switching of different shapes is actually similar to the automatic filling of frames between key frames in animation. The flight path of the drone between two vertices is a straight line by default. If a collision occurs during detection, it needs to bend. This method is currently widely used in this industry. Since the design of 3D models requires professional designers, and the process of generating the entire flight path is not automated, it requires a lot of manpower. Summary of the invention
[0004] The purpose of the present invention is to provide a method for planning the flight path of a drone cluster to solve the problems raised in the above-mentioned background technology.
[0005] To achieve the above object, the present invention provides the following technical solution: a method for planning a flight path of a drone cluster, comprising the following steps:
[0006] S1. Perform input preprocessing;
[0007] S2. Generate 3D model through Stab le Dreamfus i on system;
[0008] S3. Initialize the drone object in the system Blackender, arrange and combine according to the preset number of columns, import each 3D model and parse the vertex set, match the drone with the 3D model vertices using the method of pairwise approximate minimum total distance, generate a straight line flight path table, and perform collision detection and path correction;
[0009] S4. Output the flight path file.
[0010] Preferably, in S3, the method of pairwise approximate minimum total distance specifically includes:
[0011] Mark all vertices of the next 3D model as uncombined. For each vertex of the current 3D model, search for uncombined vertices in the next 3D model, find the vertex with the closest spatial distance, combine it, and mark it as a combined vertex.
[0012] Preferably, in S3, the collision detection specifically includes:
[0013] Based on the preset minimum distance parameter minD i st, the distance between drones is calculated in real time. When the distance between any two drones is found to be less than minD i st, the collision handling program is called to correct the path.
[0014] Preferably, in S3, the preset number of columns co l is dynamically adjusted according to the number of drones n.
[0015] Preferably, the S1 specifically includes:
[0016] Receive natural language text input by the user, segment the text, and identify the number of drones n and the corresponding modeling command sentences.
[0017] Preferably, the S2 specifically includes:
[0018] Using the Stab le Dreamfusion system, a corresponding 3D model is generated according to each modeling command sentence, and the generated 3D model is saved to a specified directory.
[0019] Preferably, the S4 specifically includes:
[0020] Save the flight path of each drone as a csv file to form a flight path file set.
[0021] Preferably, in S4, the format of the csv file is (time, x, y, z), where time is the current time, and x, y, z represent relative spatial coordinate points in the coordinate system.
[0022] Preferably, the step of outputting the flight path file further includes:
[0023] S5. Conduct cluster flight.
[0024] Preferably, the S5 specifically comprises the steps of:
[0025] First, the drones are arranged in a rectangular shape on the ground. Then the entire cluster executes the take off command, and the cluster flies to the location of the vertex set of the first 3D model and stays there. Then it flies to the location of the vertex set of the second 3D model and stays there. In this process, more 3D models are added as needed. Finally, the return command is executed from the current aerial position to land at the original take-off position.
[0026] Compared with the prior art, the present invention has the following beneficial effects:
[0027] The present invention applies text to 3D large model technology to UAV cluster flight path planning, and automatically plans the flight path of UAVs. In the past, this cluster path planning process was completed manually with the help of Blender design tools. Therefore, compared with traditional methods, this method is more efficient and less difficult. The flight path planning is quickly generated through simple human-computer interaction, meeting the requirements of high timeliness and low cost. BRIEF DESCRIPTION OF THE DRAWINGS
[0028] Figure 1 is a flow chart of the present invention;
[0029] Figure 2 It is an operation process diagram of the present invention. DETAILED DESCRIPTION
[0030] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0031] See also Figure 1-Figure 2 As shown, the present invention provides the following technical solutions:
[0032] Example 1
[0033] Start the system: First, start the main server of the UAV cluster flight performance system and load the necessary software and libraries, including the interface library of Stab le Dreamfusion and B l ender.
[0034] Input configuration: Receive the performance description text (such as "500 drones, a whatPeace Dove, a go l den cup") and basic parameters of drone flight (such as flight speed v, time interval, number of columns col, minimum distance minD i st, etc.) entered by the user through the user interface.
[0035] Preprocessing input: Preprocess the input text, including sentence segmentation, identifying the number of drones, and modeling command sentences.
[0036] Loop call Stable Dreamfusion: For each modeling command phrase, loop call Stable Dreamfusion interface to generate the corresponding 3D model and save it to the specified directory.
[0037] Verify the model: Check whether the generated 3D model meets expectations, such as model quantity, shape, etc., to ensure the smooth progress of subsequent steps.
[0038] Initialize Blackender: Initialize the Blackender environment through Blackender's development interface bpy, preparing for subsequent 3D model processing and path planning.
[0039] Create drone objects: Create drone objects in BlueDraw equal to the number of drones, and arrange them on the ground plane in the specified number of columns. For each 3D model, use BlueDraw's API to import it and parse the vertices. Each 3D model needs to parse out at least as many vertices as the number of drones, find the vertex combination that matches the number of drones, and calculate the flight path of the drone from the current position to the target vertex.
[0040] For consecutive 3D models, repeat the above process to ensure that the drone can fly to the vertices of each model in sequence.
[0041] Consider the risk of collision between drones when calculating flight paths.
[0042] The preset minimum distance parameter minD i st is used for collision detection. If a potential collision is found, the collision handler is called to correct the path.
[0043] According to the UAV flight efficiency and safety requirements, the generated flight path is optimized, such as reducing unnecessary flight distance, optimizing flight speed, etc.
[0044] Path data collation: The flight path data of each UAV is collated into (time, x, y, z) format, and the time interval is considered to generate continuous flight frames.
[0045] Save the flight path data of each drone into a separate CSV file with the drone number in the file name for easy identification and loading later.
[0046] Drone preparation: deploy the drone cluster to the take-off point and check the status of the drones (such as battery power, communication connection, etc.).
[0047] Load Flight Paths: Load the generated CSV file into the drones’ control system, ensuring that each drone knows its flight path.
[0048] The drone swarm executes the take-off command and begins flying along the predetermined flight path.
[0049] During the flight, the drone control system monitors the status of the drone in real time and makes adjustments as needed.
[0050] When the drone reaches the designated location, it displays lighting, color and other effects according to the performance requirements.
[0051] After completing all the performance actions, the drone returns to the take-off point according to the predetermined path and executes the landing command.
[0052] End of performance and feedback: After the drone swarm flight show is over, data from the performance (such as flight trajectory, performance parameters, etc.) is collected and a report is generated for subsequent analysis and improvement.
[0053] Example 2
[0054] first step:
[0055] Input preprocessing is to segment the natural language text input according to the command requirements first, and then identify the number of drones m and the command sentences (1 to n sentences) that need to be modeled.
[0056] Step 2:
[0057] for(i=1~n);
[0058] Call the Stab le Dreamfusion interface, input the ith modeling sentence, create the ith 3D model according to the ith modeling sentence, and save the ith 3D model to the specified directory.
[0059] Step 3:
[0060] 1) Initialize the blender development interface bpy.
[0061] 2) Create m drone objects on the ground plane, and automatically arrange and combine them according to col*(m / col). If the number cannot be divided, an error will be reported. (The col here can be adjusted as a parameter, for example, col=15 when n<=400, col=20 when n<=900, col=30 when n<=1600, col=40 when n>1600).
[0062] 3)for(i=1~n)
[0063] Call bpy to import the i-th 3D model into Blender to open it, and automatically parse out the m vertices of this 3D model according to the number of drones m.
[0064] When i=1, calculate the distances between the m vertices and the ground drones, and find the pairwise combination with the minimum total distance (refer to additional step 1), so that the pair of combinations has the same number, that is, the same drone. Then calculate the straight line flight path table between the two points according to the preset flight speed parameter v.
[0065] When i>1, calculate the distance between the m vertices and the m vertices of the previous 3D model, and find the pairwise combination with the minimum total distance (refer to additional step 1), so that the number of this pair of combinations is the same, that is, the same drone. Then calculate the straight line flight path table between the two points according to the preset flight speed parameter v.
[0066] When i=n, according to the preset flight speed parameter v, a straight-line flight path table is calculated for the numbered UAV at the current vertex to fly back to the take-off point.
[0067] 4) Perform collision detection. According to the preset minimum distance parameter minD i st, calculate the distance between two drones at the same time. If the distance is less than the minimum distance, it means a collision has occurred, and call the collision processing program to make corrections.
[0068] Step 4:
[0069] Summarize the flight path tables of the m drones generated above and save them in csv format to the specified directory.
[0070] Additional Step 1:
[0071] Method to find the minimum total distance between two pairs: mark all vertices of the next 3D model as uncombined. For (i = 1 ~ m), the i-th vertex of the current 3D model is used as the starting point, search and find all uncombined vertices of the next 3D model, and find the uncombined vertex k with the closest spatial distance. Combine the two closest vertices above. Mark vertex k as a combined vertex.
[0072] Embodiment 3
[0073] When generating a flight path, the flight speed is dynamically adjusted according to the relative position and distance between the drones. For example, when the drone cluster needs to quickly change formation to reach the next 3D model vertex, the flight speed can be appropriately increased; while when approaching the target position or requiring precise control, the flight speed is reduced.
[0074] Steps: When calculating the flight path, introduce a speed adjustment factor based on the current distance between the drones and the target distance.
[0075] When the drone approaches the target apex, it automatically reduces its speed to ensure accurate arrival and reduce the risk of collision.
[0076] When the cluster changes its formation, the flight speed is dynamically adjusted according to the complexity and time requirements of the formation change.
[0077] Embodiment 4:
[0078] Optimize the vertex matching algorithm to reduce computational complexity and improve matching accuracy. Use more efficient data structures and algorithms to speed up the nearest neighbor search between vertices.
[0079] step:
[0080] Use a space partitioning data structure such as a KD-tree or an octree to index and store the vertices of a 3D model to speed up nearest neighbor searches.
[0081] In the matching process, the space partitioning structure is first used to quickly locate the possible candidate vertex set, and then accurate matching is performed in the candidate set.
[0082] Introduce heuristic search strategies, such as A* algorithm or Dijkstra algorithm, to find the global optimal or approximately optimal vertex matching solution.
[0083] Embodiment five:
[0084] During the flight of the drone, the flight path is dynamically adjusted according to real-time environmental information and flight status to respond to emergencies or improve performance effects.
[0085] Steps: Integrate sensors (such as GPS, visual sensors, etc.) on the drone to collect environmental information and flight status data in real time.
[0086] Design real-time path planning algorithms to dynamically adjust the flight path based on the collected data. For example, when an obstacle is detected, automatically plan a detour path.
[0087] Introducing communication mechanisms into drone swarms enables information sharing and collaborative planning among drones to improve overall flight efficiency and safety.
[0088] Embodiment six:
[0089] By optimizing lighting and color configurations and introducing special effects (such as smoke, sparks, etc.), the visual effects of drone cluster flight performances can be enhanced.
[0090] Steps: Design the lighting and color configuration of the drone according to the performance content and scene requirements. For example, when forming a specific pattern or text, use specific lights and colors to highlight it.
[0091] Introduce special effects devices (such as smoke machines, spark launchers, etc.) to release special effects during the flight of drone clusters to increase the shock and viewing experience of the performance.
[0092] Simulate and preview lighting, color and special effects in Blueender to ensure that the expected effects can be achieved during the actual performance.
[0093] Implementation method seven:
[0094] Design the system to support a variety of different performance scenarios and environmental conditions, such as indoors, outdoors, during the day, at night, etc.
[0095] Steps: Analyze the impact of different scenarios and environmental conditions on UAV flight performances and develop corresponding adaptive strategies.
[0096] Introducing scene configuration parameters into the system design allows users to adjust flight path, lighting, color and other parameters according to actual needs.
[0097] Carry out sufficient testing and verification before the actual performance to ensure that the system can operate stably in different scenarios and environmental conditions and achieve the expected performance effect.
[0098] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.
Claims
1. A method for planning flight paths of drone swarms, characterized by: The steps include: S1. Perform input preprocessing; S2. 3D model generation through the Stable Dreamfusion system; S3. Initialize the drone object in the system Blender, arrange and combine according to the preset number of columns col, import each 3D model and parse the vertex set, calculate the approximate minimum distance between the vertex sets of the two 3D models in front and behind by using the method of pairwise approximate minimum total distance and match them one by one, generate a straight line flight path table, and perform collision detection and path correction; S4. Output the flight path file.
2. A method for planning a flight path for a drone swarm according to claim 1, characterized in that: In S3, the method of pairwise approximation of the minimum total distance specifically includes: Mark all vertices of the next 3D model as uncombined. For each vertex of the current 3D model, search for uncombined vertices in the next 3D model, find the vertex with the closest spatial distance, combine it, and mark it as a combined vertex.
3. The method for planning a flight path of a drone cluster according to claim 1, characterized in that: In S3, the collision detection specifically includes: Based on the preset minimum distance parameter minDist, the distance between UAVs is calculated in real time. When the distance between any two UAVs is found to be less than minDist, the collision handling procedure is called to correct the path.
4. The method for planning a flight path of a drone cluster according to claim 1, characterized in that: In S3, the preset number of columns col is dynamically adjusted according to the number of drones n, specifically: col=15 when n≤400, col=20 when n≤900, col=30 when n≤1600, and col=40 when n>1600.
5. The method for planning a flight path of a drone cluster according to claim 1, characterized in that: The S1 specifically includes: Receive natural language text input by the user, segment the text, and identify the number of drones n and the corresponding modeling command sentences.
6. The method for planning a flight path of a drone cluster according to claim 1, characterized in that: The S2 specifically includes: Using the Stable Dreamfusion system, a corresponding 3D model is generated according to each modeling command sentence, and the generated 3D model is saved in a specified directory.
7. The method for planning a flight path for a drone swarm according to claim 1, characterized in that: The S4 specifically includes: Save the flight path of each drone as a csv file to form a flight path file set.
8. The method for planning a flight path for a drone swarm according to claim 7, characterized in that: In S4, the format of the csv file is (time, x, y, z), where time is the current time, and x, y, z represent relative spatial coordinate points in the coordinate system.
9. The method for planning a flight path of a drone cluster according to claim 1, characterized in that: The step of outputting the flight path file further includes: S5. Conduct cluster flight.
10. The method for planning a flight path for a drone swarm according to claim 9, characterized in that: The S5 specifically includes the steps of: First, the drones are arranged in a rectangular shape on the ground, and then the entire cluster executes a take off command, and the cluster flies to the location of the vertex set of the first 3D model and stays there; Then fly to the location of the vertex set of the second 3D model and stay there; In this process, more 3D models are added as needed; finally, a return command is executed from the current aerial position to land at the original take-off position.