An optimal path generation method and system based on low-altitude airspace three-dimensional scene

Abstract

The present invention discloses an optimal path generation method and system based on a low-altitude airspace three-dimensional scene, which relates to the field of three-dimensional geographic information, including obtaining three-dimensional terrain data and meteorological parameters, and building an intelligent edge computing network, generating a preliminary path according to the three-dimensional terrain data and meteorological parameters; the aircraft takes off through the generated preliminary path, the edge computing node collects three-dimensional data from the sensor in real time, updates the preliminary path through the three-dimensional data, uses a federated learning framework to centrally train the global path planning model, and distributes the latest global path planning model parameters back to each sensor; extracts three-dimensional features from the global path planning model parameters, and fuses the three-dimensional features using a multimodal fusion strategy; uses a deep learning model to analyze the fused three-dimensional features, and designs an RRT algorithm to generate a path proposal based on the recognition results. The present invention achieves comprehensive perception of the low-altitude airspace environment by acquiring three-dimensional terrain data and building an edge computing network.

Description

Technical Field

[0001] The present invention relates to the field of three-dimensional geographic information, and in particular to an optimal path generation method and system based on a low-altitude airspace three-dimensional scene. Background Art

[0002] In recent years, with the rapid development of drone technology and its widespread application in logistics distribution, environmental monitoring and emergency response, path planning in low-altitude airspace has become a research hotspot. Traditional path planning methods mainly rely on static map data and preset flight plans, which are difficult to adapt to complex and changing three-dimensional terrain and meteorological conditions in real time. Although previous studies have shown that the accuracy of path planning can be improved by integrating geographic information systems (GIS) and weather forecast models, these methods usually lack effective response mechanisms to dynamic environmental changes.

[0003] Existing technologies mostly use centralized computing architecture, which not only increases communication delays, but also limits the flexibility and scalability of the system. Especially for long-duration or large-scale missions, real-time data processing capabilities and collaborative work between edge nodes are particularly important. Summary of the invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides an optimal path generation method based on low-altitude airspace three-dimensional scenes, which solves the problems of insufficient real-time and dynamic adaptability, the limitations of centralized computing architecture and the limitations of data processing capabilities and collaborative work, and solves the insufficient accuracy of obstacle detection and path planning, and also solves the lack of emergency response mechanism.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, the present invention provides an optimal path generation method and system based on a low-altitude airspace three-dimensional scene, which includes acquiring three-dimensional terrain data and meteorological parameters, and building an intelligent edge computing network to generate a preliminary path based on the three-dimensional terrain data and meteorological parameters.

[0008] The aircraft takes off through the generated preliminary path. During the flight, the edge computing node collects three-dimensional data from the sensor in real time, updates the preliminary path through the three-dimensional data, and adjusts the flight trajectory. The global path planning model is centrally trained using the federated learning framework, and the latest global path planning model parameters are distributed back to each sensor.

[0009] The global path planning model parameters collected by the sensor are preprocessed, three-dimensional features are extracted from the preprocessed global path planning model parameters, and the three-dimensional features are fused using a multimodal fusion strategy.

[0010] The deep learning model is used to analyze the fused 3D features, identify potential obstacles, and design the RRT algorithm to generate path proposals based on the identification results.

[0011] The path proposal is sent to the aircraft for execution. After receiving the path proposal, the aircraft adjusts the flight operation according to the planned path.

[0012] Flight status data is collected through sensors and monitoring equipment, and when deviations and abnormal situations occur, re-planning and emergency measures are immediately triggered.

[0013] As a preferred solution of the optimal path generation method and system based on the low-altitude airspace three-dimensional scene of the present invention, wherein: the acquisition of three-dimensional terrain data and meteorological parameters, and the construction of an intelligent edge computing network, and the generation of a preliminary path according to the three-dimensional terrain data and meteorological parameters include the following steps:

[0014] Obtain accurate and up-to-date 3D terrain data from the Laozi Cloud Platform, and obtain real-time and historical meteorological parameters through satellite remote sensing technology;

[0015] Install lightweight edge computing devices on aircraft, deploy intelligent edge computing nodes, and set up wireless communications;

[0016] Using the federated learning framework, the optimal preliminary path is generated based on the three-dimensional terrain data and meteorological parameters before takeoff, expressed as,

[0017]

[0018] Where x is the position vector, t is the time vector, P(x,t) is the optimal preliminary path at position x and time t, T is the time vector of the flight mission, λ is a decay factor, τ is the integral variable, z(τ) is the sensor 3D data, D(z(τ)) is the information filtering function, C is the normalization function, N is the number of edge nodes involved in the calculation, i is the index vector of the edge nodes involved in the calculation, M is the environmental assessment function, and W is the weight function.

[0019] As a preferred solution of the optimal path generation method and system based on the low-altitude airspace three-dimensional scene of the present invention, the aircraft takes off through the generated preliminary path. During the flight, the edge computing node collects three-dimensional data from the sensor in real time, updates the preliminary path through the three-dimensional data, and adjusts the flight trajectory. The global path planning model is centrally trained using the federated learning framework, and the latest global path planning model parameters are distributed back to each sensor, including the following steps:

[0020] The preliminary path planning results are uploaded to the flight control panel of the aircraft, and the aircraft starts the engine and propeller according to the preset takeoff procedure;

[0021] Various sensors on the aircraft collect position, attitude, obstacle information and weather conditions in real time, and the edge computing node pre-processes the received position, attitude, obstacle information and weather conditions;

[0022] Update the preliminary path based on the position, attitude, obstacle information and weather conditions collected in real time and adjust the flight trajectory accordingly;

[0023] The global path planning model is trained centrally using a federated learning framework in the cloud. The parameters of the global path planning model are continuously optimized based on the newly collected position, posture, obstacle information, and weather conditions, and the latest global path planning model parameters are distributed back to each sensor.

[0024] As a preferred solution of the optimal path generation method and system based on the low-altitude airspace three-dimensional scene of the present invention, the global path planning model parameters collected by the sensor are preprocessed, three-dimensional features are extracted from the preprocessed global path planning model parameters, and the three-dimensional features are fused using a multimodal fusion strategy, including the following steps:

[0025] The latest global path planning model parameters received by the sensor are converted into a unified data format to generate a global path planning model parameter file;

[0026] Parse the global path planning model parameter file and extract the weight matrix, bias term and activation function;

[0027] Normalize the extracted weight matrix, bias term and activation function to obtain normalized global path planning model parameters;

[0028] The feature extraction algorithm is applied to extract three-dimensional features from the normalized global path planning model parameters, and the extracted three-dimensional features are subjected to multi-modal fusion to obtain a fused three-dimensional feature vector.

[0029] As a preferred solution of the optimal path generation method and system based on the low-altitude airspace three-dimensional scene of the present invention, wherein: the use of the deep learning model to analyze the fused three-dimensional features, identify potential obstacles, and design the RRT algorithm to generate a path proposal based on the identification results includes the following steps:

[0030] Using the pre-trained deep learning model, the fused 3D features are analyzed to identify potential obstacle locations, expressed as,

[0031]

[0032] Among them, O(x,t) means that an obstacle is detected at position x and time t. is the obstacle existence probability output by the deep learning model, J is the number of sensor types, j is the index vector of sensor types, is the local characteristic intensity of the j-th sensor at position x and time t, is the noise level of the j-th sensor at position x and time t;

[0033] According to the identified obstacle positions, the RRT algorithm is used to randomly expand the tree structure in the search space to find a collision-free path from the start point to the end point and generate a path proposal.

[0034] As a preferred solution of the optimal path generation method and system based on the low-altitude airspace three-dimensional scene of the present invention, wherein: the path proposal is sent to the aircraft for execution, and after the aircraft receives the path proposal, the flight operation is adjusted according to the planned path, including the following steps:

[0035] The generated path proposal is sent to the flight control panel of the aircraft. After receiving the path proposal, the flight control panel begins to verify the validity and safety of the path proposal. After the verification, the path proposal is applied to the flight operation. The flight control panel monitors the position and attitude of the aircraft in real time to ensure that the aircraft always flies along the planned path.

[0036] As a preferred solution of the optimal path generation method and system based on low-altitude airspace three-dimensional scene of the present invention, wherein: the flight status data is collected by sensors and monitoring equipment, and when deviations and abnormalities occur, re-planning and emergency measures are immediately triggered, including the following steps:

[0037] The flight status data is monitored in real time through sensors, monitoring equipment and the flight control panel. When the flight control panel detects that the aircraft deviates from the planned path by more than the set deviation threshold, it is marked as a deviation event. The aircraft immediately starts the re-planning procedure and uses the latest environmental data and flight status information to regenerate a safe flight path. After the new path proposal is generated, it is verified and resent to the flight control panel.

[0038] When the flight control panel detects that the aircraft has not deviated from the planned path by more than the set deviation threshold, it will make fine adjustments according to the normal flight path to keep the aircraft on the correct track;

[0039] The aircraft keeps the emergency response mechanism activated to deal with emergencies. When the situation is critical, the aircraft immediately initiates emergency measures.

[0040] In a second aspect, the present invention provides an optimal path generation system based on a low-altitude airspace three-dimensional scene, including a data acquisition module, a trajectory adjustment module, a model optimization module, a deep learning module, a path proposal module and an anomaly detection module;

[0041] The data acquisition module acquires three-dimensional terrain data and meteorological parameters, and builds an intelligent edge computing network to generate a preliminary path based on the three-dimensional terrain data and meteorological parameters;

[0042] In the trajectory adjustment module, the aircraft takes off through the generated preliminary path. During the flight, the edge computing node collects three-dimensional data from the sensor in real time, updates the preliminary path through the three-dimensional data, and adjusts the flight trajectory. The global path planning model is centrally trained using the federated learning framework, and the latest global path planning model parameters are distributed back to each sensor.

[0043] The model optimization module preprocesses the global path planning model parameters collected by the sensor, extracts three-dimensional features from the preprocessed global path planning model parameters, and fuses the three-dimensional features using a multimodal fusion strategy;

[0044] The deep learning module uses a deep learning model to analyze the fused three-dimensional features, identify potential obstacles, and designs an RRT algorithm to generate a path proposal based on the identification results;

[0045] The path proposal module sends the path proposal to the aircraft for execution. After receiving the path proposal, the aircraft adjusts the flight operation according to the planned path.

[0046] The anomaly detection module collects flight status data through sensors and monitoring equipment, and immediately triggers re-planning and emergency measures when deviations and abnormal situations occur.

[0047] In a third aspect, the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, any step of the optimal path generation method based on a low-altitude airspace three-dimensional scene as described in the first aspect of the present invention is implemented.

[0048] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the optimal path generation method based on a low-altitude airspace three-dimensional scene as described in the first aspect of the present invention.

[0049] The beneficial effects of the present invention are as follows: by acquiring three-dimensional terrain data and meteorological parameters and building an intelligent edge computing network, a comprehensive perception of the low-altitude airspace environment is achieved. Building an intelligent edge computing network improves the real-time processing capability and response speed of the system and avoids the communication delay caused by the centralized architecture. The aircraft takes off according to the preliminary path and collects three-dimensional data in real time during the flight, thereby realizing dynamic path updates and flight trajectory adjustments. The global path planning model is centrally trained using a federated learning framework and the latest parameters are distributed back to each node, thereby further improving the adaptability and accuracy of the global path planning model. By preprocessing sensor data and extracting three-dimensional features, effective analysis of complex environmental information is achieved. The use of a multimodal fusion strategy to integrate data from different sources enhances the perception of environmental changes, immediately triggers re-planning and emergency measures, and ensures the safety and controllability of the flight mission. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0051] Figure 1 This is a flow chart of the optimal path generation method based on the low-altitude airspace three-dimensional scene in Example 1.

[0052] Figure 2 This is a module diagram of the optimal path generation system based on the low-altitude airspace three-dimensional scene in Example 1. DETAILED DESCRIPTION

[0053] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0054] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0055] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0056] Example 1, reference Figure 1 and Figure 2 , which is the first embodiment of the present invention, provides an optimal path generation method and system based on a low-altitude airspace three-dimensional scene, comprising the following steps:

[0057] S1. Obtain three-dimensional terrain data and meteorological parameters, build an intelligent edge computing network, and generate a preliminary path based on the three-dimensional terrain data and meteorological parameters, as follows:

[0058] Register with Laozi Cloud Platform, download the required 3D geographic data via FTP, specify the required geographic range and resolution, including elevation model and land cover information, convert the format of the downloaded data, apply Gaussian filtering to remove noise, apply projection transformation to convert coordinates into a coordinate system, then calibrate the coordinate system using feature matching, select a satellite data provider, download the required meteorological parameters via FTP, decode, project and convert the raw data, and perform quality checks on the processed data to ensure its accuracy and completeness;

[0059] Select lightweight edge computing devices suitable for aircraft, configure operating systems, driver and middleware installation and configuration, deep learning framework, sensor data processing library, deploy intelligent edge computing nodes, ensure that each node can independently process local data and work with other nodes, set up wireless communication, and configure security protocols to protect the security of data transmission;

[0060] The 3D terrain data obtained from the Laozi cloud platform and the meteorological parameters obtained through satellite remote sensing technology are integrated, and the preliminary path is generated using the pre-trained deep learning model. The path is optimized by combining the 3D terrain data and meteorological parameters, which is expressed as,

[0061]

[0062] Where x is the position vector, indicating the current position; t is the time vector, indicating the current time point; P(x,t) is the optimal preliminary path at position x and time t; T is the time vector of the flight mission, indicating the time period of the flight mission; λ is an attenuation factor used to control the influence of past data; τ is the integral variable, indicating the time from the start time to the current time; z(τ) is the multimodal sensor data collected at time τ; D(z(τ)) is the information filtering function, which adjusts the path weight according to the multimodal sensor data. α is a tuning parameter used to adjust the sensitivity of information filtering, C is a normalization function used to sum all paths to 1, is the terrain safety factor, which is used to ensure terrain safety, Ω is the location set, which represents the set of all locations that the aircraft may reach during the entire mission, N is the number of edge nodes involved in the calculation, i is the index vector of the edge nodes involved in the calculation, and M is the environmental assessment function, M = exp(-γ·∥xx i ∥ 2 ), γ is the distance sensitivity parameter, which is used to adjust the response speed of the environmental assessment function M to the distance change, x i is the position of the ith edge node, W is the weight function, is the wind speed value at position x, which is used to calculate the weight function W, δ i is the bias parameter of the i-th edge node, which is used to adjust the basic value of the weight function W.

[0063] Further meteorological parameters are temperature, humidity, wind speed, wind direction, atmospheric pressure, precipitation, cloud cover, visibility, sunshine duration and solar radiation intensity.

[0064] S2. The aircraft takes off along the generated preliminary path. During the flight, the edge computing node collects 3D data from the sensor in real time, updates the preliminary path through the 3D data, and adjusts the flight trajectory. The global path planning model is trained centrally using the federated learning framework, and the latest global path planning model parameters are distributed back to each sensor. The details are as follows:

[0065] The generated preliminary path is finally verified on the ground station, and the path data is converted into a format that can be recognized by the flight control system. The path data is uploaded to the flight control panel of the aircraft through a wireless communication link. The flight control panel receives and parses the path data, stores it in the local memory, and prepares to execute the takeoff procedure. According to the preset takeoff procedure, the flight control panel starts the engine and propeller, and performs self-test and initialization operations;

[0066] The GPS on the aircraft obtains precise location coordinates, the IMU obtains the attitude parameters of the aircraft, the lidar and visual camera detect obstacles in the surrounding environment, and the meteorological sensor obtains temperature, humidity, wind speed, and wind direction. The collected meteorological data is transmitted to the edge computing node in real time, and Gaussian filtering is used to remove isolated noise points in the meteorological data, and the data timestamps of the sensors are ensured to be consistent.

[0067] The preliminary path is updated using the RRT algorithm based on the preprocessed position coordinates, attitude parameters, obstacles in the surrounding environment, and meteorological data. The flight control panel adjusts the path, attitude, and speed of the aircraft in real time based on the updated path information.

[0068] Using the federated learning framework, the global path planning model is trained centrally in the cloud. During the training process, the historical data and newly collected data are combined to continuously optimize the global path planning model parameters, and the latest global path planning model parameters are distributed back to each edge computing node to ensure that each node uses the latest global path planning model for path planning. Incremental update technology is used to reduce parameter transmission and improve update efficiency.

[0069] S3. Preprocess the global path planning model parameters collected by the sensor, extract three-dimensional features from the preprocessed global path planning model parameters, and fuse the three-dimensional features using a multimodal fusion strategy, as follows:

[0070] Ensure that the global path planning model parameters uploaded by all edge computing nodes have the same structure and format. Collect the latest global path planning model parameters from each edge node on the cloud server and merge them into a unified data set. Convert the merged global path planning model parameters into a standard data format to ensure that it is easy to parse and store. Save the converted data as a global path planning model parameter file. The naming convention should include a timestamp or other identifier.

[0071] Use the corresponding library in the programming language to read the generated parameter file, parse the parameters of each layer of the global path planning model according to the file format, normalize the extracted weight matrix, bias term and activation function, and obtain the normalized global path planning model parameters;

[0072] Use the principal component analysis feature extraction algorithm to apply the selected algorithm to extract features from the normalized weight matrix, bias term, and activation function to obtain a set of low-dimensional feature vectors that can characterize the characteristics of the global path planning model. Prepare three-dimensional features from different sources, select multimodal fusion to generate the final fused three-dimensional feature vector, and verify the fused three-dimensional feature vector to ensure its effectiveness and robustness.

[0073] S4. Use the deep learning model to analyze the fused 3D features, identify potential obstacles, and design the RRT algorithm to generate path proposals based on the identification results, as follows:

[0074] Collect images and point cloud data related to the aircraft environment, annotate the collected data, mark the location, category and other information of obstacles, provide supervision signals for fine-tuning the deep learning model, apply data enhancement technology to expand the training set, improve the generalization ability of the deep learning model, adjust hyperparameters according to the task, ensure that the deep learning model can converge quickly on new data, select the mean square error loss function, ensure that the deep learning model is optimized in the right direction, use the validation set to evaluate the performance of the deep learning model, monitor the accuracy, recall rate, F1 score, adjust the deep learning model structure or training strategy, apply quantization and pruning technology to reduce the size and computational complexity of the deep learning model, and ensure that it can run efficiently on edge devices to form a deep learning model;

[0075] The pre-trained deep learning model is loaded on the edge computing node, and the fused three-dimensional feature vector is used as the input of the deep learning model. The deep learning model outputs the coordinate position of each obstacle, which is expressed as,

[0076]

[0077] Among them, O(x,t) means that an obstacle is detected at position x and time t. is the obstacle existence probability output by the deep learning model, reflecting the probability of an obstacle at the current position and time point. J is the number of sensor types, the total number of different sensor types used for obstacle detection. j is the index vector of the sensor type, used to identify the specific sensor type. is the local feature intensity of the j-th sensor at position x and time t, which is used to reflect the significant feature intensity detected by the sensor at the current position and time point. is the noise level of the j-th sensor at position x and time t, which is used to reflect the noise level of the sensor at the current position and time point;

[0078] According to the identified obstacle positions, the RRT algorithm is used to generate the optimal path. First, the RRT algorithm is initialized, the starting point and the end point are set, the search space boundary and the obstacle area are defined, a sample point is randomly selected in the search space, and a new edge is extended from the nearest node in the existing tree structure to the sample point. If the new edge does not intersect with any known obstacle, it is added to the tree structure; otherwise, the extension is abandoned. When a node in the tree structure approaches the end point, the node is directly connected to the end point to form a complete collision-free path. The generated path is smoothed and other optimization operations are performed to ensure that it meets the dynamic constraints and safety requirements of the aircraft.

[0079] S5. Send the path proposal to the aircraft for execution. After receiving the path proposal, the aircraft adjusts the flight operation according to the planned path, as follows:

[0080] The path proposal is sent to the flight control panel of the aircraft through the ground station and the cloud server. After receiving the path proposal, the flight control panel returns a confirmation message to the sender, checks whether the path proposal conforms to the predefined data format and grammatical rules, and confirms whether each waypoint in the path proposal is within the physical capability of the aircraft. It uses geometric methods to check whether the path proposal intersects with known obstacles and confirms whether the path proposal complies with local aviation regulations and airspace restrictions. After the above verification, if the path proposal is judged to be valid and safe, proceed to the next step; otherwise, a new path proposal is generated and the verification process is repeated. When the path proposal is judged to be valid and safe, the flight control panel loads the path proposal into the internal navigation and gradually starts the flight program according to the preset flight plan. During the flight, data from GPS, IMU, lidar, visual cameras and sensors are collected in real time to ensure a comprehensive understanding of the aircraft status. The flight control panel adjusts the attitude and speed of the aircraft according to the waypoint information in the path proposal to ensure that it always flies along the planned path.

[0081] S6. Collect flight status data through sensors and monitoring equipment. When deviations and abnormalities occur, re-planning and emergency measures are immediately triggered, as follows:

[0082] According to the mission requirements and flight environment, a deviation threshold is set to determine whether the aircraft has deviated from the planned path. The deviation threshold can be based on position deviation, attitude deviation and comprehensive consideration of multiple factors. Data from GPS, IMU, LiDAR, visual cameras and sensors are collected in real time to ensure a comprehensive understanding of the aircraft status. When the flight control panel detects that the aircraft deviates from the planned path by more than the set deviation threshold, it is immediately marked as a deviation event and triggers the re-planning procedure to send a new path proposal to the flight control panel to prepare for the execution of a new flight plan.

[0083] When the flight control panel detects that the aircraft has not deviated from the planned path by more than the set deviation threshold, it will make fine adjustments according to the normal flight path to keep the aircraft on the correct track;

[0084] The aircraft always keeps the emergency response mechanism activated to deal with emergencies. When the situation is critical, the aircraft will perform emergency obstacle avoidance to quickly adjust the flight path, avoid new obstacles, automatically return to the home position, navigate to the nearest safe landing point or return to the take-off point and emergency landing, and choose a suitable location for emergency landing to ensure the safety of the aircraft and payload.

[0085] A critical situation refers to any situation that may cause the aircraft to be unable to continue to fly safely or pose a threat to people and property on the ground, including but not limited to sensor failure, communication interruption, power system failure, sudden appearance of obstacles, and extreme weather conditions.

[0086] This embodiment also provides an optimal path generation system based on a low-altitude airspace three-dimensional scene, including: a data acquisition module, a trajectory adjustment module, a model optimization module, a deep learning module, a path proposal module and an anomaly detection module;

[0087] The data acquisition module acquires three-dimensional terrain data and meteorological parameters, builds an intelligent edge computing network, and generates a preliminary path based on the three-dimensional terrain data and meteorological parameters. In the trajectory adjustment module, the aircraft takes off through the generated preliminary path. During the flight, the edge computing node collects three-dimensional data from the sensor in real time, updates the preliminary path through the three-dimensional data, and adjusts the flight trajectory. The global path planning model is centrally trained using the federated learning framework, and the latest global path planning model parameters are distributed back to each sensor. The model optimization module preprocesses the global path planning model parameters collected by the sensor, extracts three-dimensional features from the preprocessed global path planning model parameters, and fuses the three-dimensional features using a multimodal fusion strategy. The deep learning module uses a deep learning model to analyze the fused three-dimensional features, identify potential obstacles, and design an RRT algorithm based on the identification results to generate a path proposal. The path proposal module sends the path proposal to the aircraft for execution. After receiving the path proposal, the aircraft adjusts the flight operation according to the planned path. The anomaly detection module collects flight status data through sensors and monitoring equipment, and immediately triggers re-planning and emergency measures when deviations and abnormalities occur.

[0088] This embodiment also provides a computer device, which is suitable for the optimal path generation method and system based on low-altitude airspace three-dimensional scenes, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute computer-executable instructions to implement the optimal path generation method and system based on low-altitude airspace three-dimensional scenes as proposed in the above embodiments.

[0089] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0090] The present embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the method and system for generating an optimal path based on a low-altitude airspace three-dimensional scene as proposed in the above embodiment are implemented; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as a static random access memory (SRAM), an electrically erasable programmable read-only memory (EEPROM), an erasable programmable read-only memory (EPROM), a programmable read-only memory (PROM), a read-only memory (ROM), a magnetic memory, a flash memory, a disk or an optical disk.

[0091] In summary, the present invention achieves comprehensive perception of the low-altitude airspace environment by: acquiring three-dimensional terrain data and meteorological parameters and building an intelligent edge computing network; building an intelligent edge computing network improves the real-time processing capability and response speed of the system and avoids the communication delay caused by the centralized architecture; the aircraft takes off according to the preliminary path and collects three-dimensional data in real time during the flight, thereby realizing dynamic path updating and flight trajectory adjustment; the global path planning model is centrally trained using a federated learning framework and the latest parameters are distributed back to each node, thereby further improving the adaptability and accuracy of the global path planning model; the sensor data is preprocessed and three-dimensional features are extracted to achieve effective analysis of complex environmental information; the multimodal fusion strategy is used to integrate data from different sources, thereby enhancing the system's perception of environmental changes and immediately triggering re-planning and emergency measures to ensure the safety and controllability of the flight mission.

[0092] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. An optimal path generation method based on a low-altitude airspace three-dimensional scene, characterized by: include, Obtain 3D terrain data and meteorological parameters, and build an intelligent edge computing network to generate a preliminary path based on the 3D terrain data and meteorological parameters; The aircraft takes off along the generated preliminary path. During the flight, the edge computing node collects 3D data from sensors in real time, updates the preliminary path through the 3D data, and adjusts the flight trajectory. The global path planning model is trained centrally using the federated learning framework, and the latest global path planning model parameters are distributed back to each sensor. Preprocessing the global path planning model parameters collected by the sensor, extracting three-dimensional features from the preprocessed global path planning model parameters, and fusing the three-dimensional features using a multimodal fusion strategy; Use the deep learning model to analyze the fused 3D features, identify potential obstacles, and design the RRT algorithm to generate path proposals based on the identification results; The path proposal is sent to the aircraft for execution. After receiving the path proposal, the aircraft adjusts the flight operation according to the planned path. Flight status data is collected through sensors and monitoring equipment, and when deviations and abnormal situations occur, re-planning and emergency measures are immediately triggered.

2. The optimal path generation method based on the low-altitude airspace three-dimensional scene according to claim 1, characterized in that: The method of obtaining three-dimensional terrain data and meteorological parameters, and constructing an intelligent edge computing network, and generating a preliminary path according to the three-dimensional terrain data and meteorological parameters includes the following steps: Obtain three-dimensional terrain data from the Laozi cloud platform and obtain real-time and historical meteorological parameters through satellite remote sensing technology; Install lightweight edge computing devices on aircraft, deploy intelligent edge computing nodes, and set up wireless communications; Using the federated learning framework, the optimal preliminary path is generated based on the three-dimensional terrain data and meteorological parameters before takeoff, expressed as, Where x is the position vector, t is the time vector, P(x,t) is the optimal preliminary path at position x and time t, T is the time vector of the flight mission, λ is a decay factor, τ is the integral variable, z(τ) is the sensor 3D data, D(z(τ)) is the information filtering function, C is the normalization function, N is the number of edge nodes involved in the calculation, i is the index vector of the edge nodes involved in the calculation, M is the environmental assessment function, and W is the weight function.

3. The optimal path generation method based on the low-altitude airspace three-dimensional scene as claimed in claim 2, characterized in that: The aircraft takes off along the generated preliminary path. During the flight, the edge computing node collects three-dimensional data from the sensor in real time, updates the preliminary path through the three-dimensional data, and adjusts the flight trajectory. The global path planning model is trained centrally using the federated learning framework, and the latest global path planning model parameters are distributed back to each sensor. The following steps are included: The preliminary path planning results are uploaded to the flight control panel of the aircraft, and the aircraft starts the engine and propeller according to the preset takeoff procedure; Various sensors on the aircraft collect position, attitude, obstacle information and weather conditions in real time, and the edge computing node pre-processes the received position, attitude, obstacle information and weather conditions; Update the preliminary path based on the position, attitude, obstacle information and weather conditions collected in real time and adjust the flight trajectory accordingly; The global path planning model is centrally trained using a federated learning framework in the cloud, and the parameters of the global path planning model are optimized by combining the newly collected position, posture, obstacle information, and weather conditions. The latest global path planning model parameters are distributed back to each sensor.

4. The optimal path generation method based on the low-altitude airspace three-dimensional scene as claimed in claim 3 is characterized in that: Preprocessing the global path planning model parameters collected by the sensor, extracting three-dimensional features from the preprocessed global path planning model parameters, and fusing the three-dimensional features using a multimodal fusion strategy includes the following steps: The latest global path planning model parameters received by the sensor are converted into a unified data format to generate a global path planning model parameter file; Parse the global path planning model parameter file and extract the weight matrix, bias term and activation function; Normalize the extracted weight matrix, bias term and activation function to obtain normalized global path planning model parameters; The feature extraction algorithm is applied to extract three-dimensional features from the normalized global path planning model parameters, and the extracted three-dimensional features are subjected to multi-modal fusion to obtain a fused three-dimensional feature vector.

5. The optimal path generation method based on the low-altitude airspace three-dimensional scene as claimed in claim 4 is characterized in that: The method of using a deep learning model to analyze the fused three-dimensional features, identifying obstacles, and designing an RRT algorithm to generate a path proposal based on the identification results includes the following steps: Using the pre-trained deep learning model, the fused 3D feature vector is analyzed to identify the obstacle location, which is expressed as, Among them, O(x,t) means that an obstacle is detected at position x and time t. is the obstacle existence probability output by the deep learning model, J is the number of sensor types, j is the index vector of sensor types, is the local characteristic intensity of the j-th sensor at position x and time t, is the noise level of the j-th sensor at position x and time t; According to the identified obstacle positions, the RRT algorithm is used to randomly expand the tree structure in the search space to find a collision-free path from the start point to the end point and generate a path proposal.

6. The optimal path generation method based on the low-altitude airspace three-dimensional scene as claimed in claim 5, characterized in that: The sending of the path proposal to the aircraft for execution, and the aircraft adjusting the flight operation according to the planned path after receiving the path proposal, includes the following steps: The generated path proposal is sent to the flight control panel of the aircraft. After receiving the path proposal, the flight control panel begins to verify the validity and safety of the path proposal. After the verification, the path proposal is applied to the flight operation. The flight control panel monitors the position and attitude of the aircraft in real time.

7. The optimal path generation method based on the low-altitude airspace three-dimensional scene according to claim 6 is characterized in that: The flight status data is collected by sensors and monitoring equipment, and when deviations and abnormalities occur, re-planning and emergency measures are immediately triggered, including the following steps: Set a deviation threshold and monitor flight status data in real time through sensors, monitoring equipment and the flight control panel. When the flight control panel detects that the aircraft deviates from the planned path by more than the deviation threshold, it is marked as a deviation event. The aircraft immediately starts the re-planning procedure and uses the latest environmental data and flight status information to regenerate a safe flight path. After the new path proposal is generated, it is verified and resent to the flight control panel. When the flight control panel detects that the aircraft has not deviated from the planned path by more than the deviation threshold, it will make fine adjustments according to the normal flight path to keep the aircraft on the correct track; The aircraft keeps the emergency response mechanism activated to deal with emergencies. When the situation is critical, the aircraft immediately initiates emergency measures.

8. An optimal path generation system based on a low-altitude airspace three-dimensional scene, based on the optimal path generation method and system based on a low-altitude airspace three-dimensional scene according to any one of claims 1 to 7, characterized in that: Including data acquisition module, trajectory adjustment module, model optimization module, deep learning module, path proposal module and anomaly detection module; The data acquisition module acquires three-dimensional terrain data and meteorological parameters, and builds an intelligent edge computing network to generate a preliminary path based on the three-dimensional terrain data and meteorological parameters; In the trajectory adjustment module, the aircraft takes off through the generated preliminary path. During the flight, the edge computing node collects three-dimensional data from the sensor in real time, updates the preliminary path through the three-dimensional data, and adjusts the flight trajectory. The global path planning model is centrally trained using the federated learning framework, and the latest global path planning model parameters are distributed back to each sensor. The model optimization module preprocesses the global path planning model parameters collected by the sensor, extracts three-dimensional features from the preprocessed global path planning model parameters, and fuses the three-dimensional features using a multimodal fusion strategy; The deep learning module uses a deep learning model to analyze the fused three-dimensional features, identify potential obstacles, and designs an RRT algorithm to generate a path proposal based on the identification results; The path proposal module sends the path proposal to the aircraft for execution. After receiving the path proposal, the aircraft adjusts the flight operation according to the planned path. The anomaly detection module collects flight status data through sensors and monitoring equipment, and immediately triggers re-planning and emergency measures when deviations and abnormal situations occur.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the optimal path generation method and system based on the low-altitude airspace three-dimensional scene as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the optimal path generation method and system based on a low-altitude airspace three-dimensional scene as described in any one of claims 1 to 7 are implemented.

Images