Unmanned aerial vehicle low-altitude route intelligent planning and cooperation system and method based on multi-source perception and edge calculation

Abstract

The invention discloses an unmanned aerial vehicle low-altitude route intelligent planning and cooperation system and method based on multi-source perception and edge calculation, and relates to the technical field of unmanned aerial vehicle route planning. Comprising a multi-source data sensing module, a data processing module, a planning execution module and a feedback monitoring module, and the multi-source data sensing module is used for acquiring meteorological data and airborne data and generating an initial data set; the data processing module is used for generating an environment model and inputting the initial data set into the environment model to obtain physical prediction data and obstacle information data; the planning execution module is used for generating a planning path according to the physical prediction data and the obstacle information data and controlling the unmanned aerial vehicle group to fly; and the feedback monitoring module is used for monitoring the flight state of the unmanned aerial vehicle group in real time and feeding data back to the ground control unit. According to the invention, the flight safety, efficiency and cooperative capability of the unmanned aerial vehicle group are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of UAV route planning, and particularly to an intelligent UAV low-altitude route planning and coordination system and method based on multi-source perception and edge computing. Background Art

[0002] In the prior art, the research on UAV low-altitude route planning mainly focuses on static path optimization and obstacle avoidance in local areas. However, with the diversification of UAV application scenarios, especially in complex and dynamically changing environments, the existing technical methods often struggle to cope with the changing flight conditions. Traditional path planning methods usually rely on static environmental data, such as calculating paths based on pre-set geographical information and obstacle data. This is ineffective in adapting to rapidly changing environmental conditions, especially the dynamic changes in the wind field.

[0003] Currently, some studies attempt to optimize routes by combining meteorological information, especially the application of wind field prediction. Such methods usually rely on traditional meteorological models and simple wind field data prediction models. However, traditional wind field prediction methods have significant deficiencies in accuracy and timeliness. Especially when predicting wind field changes in the far future, it is difficult to provide real-time and accurate data. This results in the inability to synchronously update wind field prediction and route planning in practical applications, making it difficult to adapt to complex changes during flight and affecting flight safety and efficiency.

[0004] Meanwhile, most existing UAV route planning methods focus on the path optimization of a single UAV and rarely involve the scheduling problem of multi-UAV cooperative flight. In a complex mission environment, the cooperation and resource scheduling among multiple UAVs become the key to improving mission efficiency and flight safety. However, the existing technology has a relatively simple consideration of multi-UAV cooperative operations and lacks research on how to achieve effective multi-UAV resource scheduling and route coordination in a dynamically changing environment, especially under low-altitude wind field conditions.

[0005] Another key issue is that the existing systems have weak capabilities for real-time monitoring and adjustment during flight, especially in the low-altitude environment where wind fields and other dynamic factors change significantly. Most existing flight control systems rely on fixed preset rules and models for control and lack the ability to adaptively adjust based on real-time data. Especially when factors such as wind speed and wind direction change significantly, the existing systems are unable to effectively respond, resulting in the flight path deviating from the optimal trajectory and even affecting the success rate of the flight mission.

[0006] Generally speaking, there are many deficiencies in the prior art for low-altitude route planning of unmanned aerial vehicles (UAVs): the timeliness and accuracy of dynamic wind field prediction are insufficient, and it is impossible to adjust the route in real time according to the wind field changes; there is a lack of effective resource scheduling and path coordination for multi-UAV cooperative operations; the real-time perception and adaptive adjustment capabilities during flight are insufficient.

[0007] Therefore, providing an intelligent planning and coordination system and method for UAV low-altitude routes based on multi-source perception and edge computing to solve the difficulties existing in the prior art is an urgent problem for those skilled in the art. Summary of the Invention

[0008] In view of this, the present invention provides an intelligent planning and coordination system and method for UAV low-altitude routes based on multi-source perception and edge computing, which improves the flight safety, efficiency and coordination ability of UAV swarms.

[0009] To achieve the above object, the present invention adopts the following technical solutions:

[0010] An intelligent planning and coordination system for UAV low-altitude routes based on multi-source perception and edge computing includes a multi-source data perception module, a data processing module, a planning and execution module, and a feedback and monitoring module. The multi-source data perception module, the first input end of the data processing module, the first output end of the planning and execution module, and the input end of the feedback and monitoring module are connected in sequence. The second output end of the planning and execution module is connected to the input end of the multi-source data perception module, and the output end of the feedback and monitoring module is connected to the second input end of the data processing module;

[0011] The multi-source data perception module is used to obtain meteorological data and airborne data and generate an initial data set;

[0012] The data processing module is used to generate an environmental model, and input the initial data set into the environmental model to obtain physical prediction data and obstacle information data;

[0013] The planning and execution module is used to generate a planned path according to the physical prediction data and obstacle information data and control the flight of the UAV swarm;

[0014] The feedback and monitoring module is used to monitor the flight state of the UAV swarm in real time and feed back the data to the ground control unit.

[0015] Optionally, the multi-source data perception module obtains corresponding key data through integrating ground meteorological stations, satellite remote sensing, and airborne sensors, and generates an initial data set.

[0016] Optionally, the environmental model includes a preprocessing unit that applies PredRNN to an encoder-decoder framework, constructs a loss function using a physical information neural network constraint mechanism, and fuses the initial data set.

[0017] Optionally, the encoder-decoder framework includes adopting spatio-temporal memory unit parameters M and establishing a zigzag memory propagation path, enabling the bottom layer of the next time step to directly obtain the information of the deepest layer of the previous time step;

[0018] The loss function is set as follows:

[0019]

[0020] where u and v are the x-axis and y-axis components of the wind field wind speed from time step t1 to t10 respectively, t is the corresponding time step, fu is the Coriolis component force in the u direction, fv is the Coriolis component force in the v direction, F u is the residual term of the control equation for the u component, F v is the residual term of the control equation for the v component, MSE is the total mean square error, MSE u is the mean square error corresponding to the wind speed of u, MSE v is the mean square error corresponding to the wind speed of v, is the mean square error of the residual term Fu, is the mean square error of the residual term Fv, N is the total number of data points, u i is the u-component velocity value of the i-th data point, is the u-component velocity value of the i-th data point predicted by the neural network, v i is the v-component velocity value of the i-th data point, is the v-component velocity value of the i-th data point predicted by the neural network, is the F of the i-th data point in the numerical solution u residual term, is the F of the i-th data point predicted by the neural network u value (i.e., the PDE residual), is the F of the i-th data point in the numerical solution v residual term, is the F of the i-th data point predicted by the neural network v residual term.

[0021] Optionally, the planned path fuses the flight distance, turbulence intensity, and battery energy consumption coefficient and combines the improved A* algorithm for path optimization. The core cost function expression of the improved A* algorithm is:

[0022]

[0023] where H(n k ) is the estimated cost from the current node n k to the target node n m considering the influence of wind speed, k is the current node index, representing the k-th node in the search path, m is the index of the end node, di For node n in spherical coordinates i-1 and n i The distance between them, v airspeed is the airspeed of the aircraft in a windless environment, v ground is the actual speed of the aircraft relative to the ground, v airspeed and v ground The resultant velocity of, G(n i ) is the movement cost from the previous node n i-1 to the current node n i combining the wind speed factor.

[0024] Optionally, it further includes a multi-aircraft cooperation module and a control unit,

[0025] The multi-aircraft cooperation module is used to achieve the cooperative control of the UAV swarm;

[0026] The control unit is used to receive the data of the multi-source data perception module and the planning execution module, and control the startup of the data processing module and the UAV attitude.

[0027] The intelligent low-altitude route planning and cooperation method for UAVs based on multi-source perception and edge computing, applied to the intelligent low-altitude route planning and cooperation system for UAVs based on multi-source perception and edge computing described in any one of the above, includes the following steps:

[0028] Install and configure the multi-source data perception module in the UAV and the ground station system to collect environmental data in real time;

[0029] After the control module receives the corresponding environmental data, it controls the data processing module to start, and obtains physical prediction data and obstacle information data;

[0030] Generate a planned path based on the physical prediction data and the obstacle information data, the control module controls the flight of the UAV swarm, and automatically adjusts the flight path according to environmental changes to avoid obstacles and extreme wind speed areas;

[0031] The feedback monitoring module monitors the flight process of the UAV swarm in real time.

[0032] As can be seen from the above technical solutions, compared with the prior art, the present invention provides an intelligent route planning and coordination system and method for low-altitude unmanned aerial vehicles (UAVs) based on multi-source perception and edge computing, which has the following beneficial effects: 1) By integrating various sensor data (such as wind speed, wind direction, air pressure, etc.) and combining with a physics-informed neural network (PINN) for wind field prediction, the present invention not only improves the accuracy of wind field prediction but also effectively avoids the limitations caused by traditional single data sources. By synthesizing multi-source data, the present invention can more accurately capture and predict wind field changes in the low-altitude environment, providing more reliable data support for UAV route planning; 2) The path planning module of the present invention adopts an improved A* algorithm, which can dynamically adjust route planning according to real-time wind field data, optimize the flight route, not only improve flight efficiency, reduce energy consumption, but also can respond to complex wind field changes in real time to ensure safety during flight. At the same time, the multi-UAV coordination module supports collaborative task allocation for multiple UAVs, enabling UAVs to cooperate efficiently in the same route and making full use of limited resources; 3) By introducing a mobile edge computing module, the present invention can realize real-time data processing and decision execution during route planning, effectively reducing the dependence on remote servers, reducing data transmission latency, and ensuring that the system can quickly respond in a dynamically changing environment, enabling UAVs to execute tasks more flexibly and efficiently in a complex low-altitude environment; 4) The flight monitoring and feedback module in the system can monitor the flight state in real time and make intelligent adjustments according to the task execution situation. Once an abnormality or deviation from the preset trajectory is detected during flight, the system will automatically correct it to ensure the accuracy of the route and flight safety. Through real-time feedback, operators can better grasp the flight state and take necessary intervention measures in a timely manner, improving the intelligent level of the system; 5) Combining various sensors, edge computing, and intelligent algorithms, the system of the present invention has strong robustness and fault tolerance capabilities. Even in a complex low-altitude environment, such as under severe wind field changes or equipment failures, the system can still ensure the stability and reliability of UAV route planning and task execution, reducing the operation risks of UAVs; 6) By optimizing path planning and task scheduling, the present invention can minimize energy consumption, optimize flight time, and improve the efficiency of the route. This not only reduces energy consumption during flight but also reduces environmental impacts caused by non-optimal routes, achieving the goals of energy conservation and environmental protection. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only the embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on the provided drawings.

[0034] Figure 1 Block diagram of the intelligent route planning and coordination system for low-altitude UAVs based on multi-source perception and edge computing disclosed in the present invention;

[0035] Figure 2 Schematic diagram of the intelligent route planning and coordination system for low-altitude UAVs based on multi-source perception and edge computing disclosed in the present invention;

[0036] Figure 3 Flow chart of the physical field prediction disclosed in the present invention;

[0037] Figure 4 Wind speed decomposition diagram disclosed in the present invention;

[0038] Figure 5a Comparison diagram of the mean square error prediction results of different models of U-component wind at the test points disclosed in the embodiments of the present invention;

[0039] Figure 5b Comparison diagram of the mean square error prediction results of different models of V-component wind at the test points disclosed in the embodiments of the present invention;

[0040] Figure 6a Comparison diagram of the mean square error prediction results of different models of u-component wind on the entire test set disclosed in the embodiments of the present invention;

[0041] Figure 6b Comparison diagram of the mean square error prediction results of different models of v-component wind on the entire test set disclosed in the embodiments of the present invention;

[0042] Figure 6c Comparison diagram of the structural similarity prediction results of different models of U-component wind on the entire test set disclosed in the embodiments of the present invention;

[0043] Figure 6d Comparison diagram of the structural similarity prediction results of different models of V-component wind on the entire test set disclosed in the embodiments of the present invention;

[0044] Figure 7a Comparison diagram of the A* path and the U-direction wind field disclosed in the embodiments of the present invention;

[0045] Figure 7b Disclosed in the embodiments of the present invention Figure 7a Comparison diagram of the A* path and the U-direction wind field after swapping the start and end points;

[0046] Figure 7c Comparison diagram of the A* path and the V-direction wind field disclosed in the embodiments of the present invention;

[0047] Figure 7d Disclosed in the embodiments of the present invention Figure 7c ) Comparison diagram of the A* path and the V-direction wind field after swapping the start and end points. Detailed implementation manners

[0048] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0049] The present invention discloses an intelligent planning and coordination system for low-altitude air routes of unmanned aerial vehicles based on multi-source perception and edge computing. Referring to Figure 1 and Figure 2 as shown, it includes a multi-source data perception module, a data processing module, a planning execution module, and a feedback monitoring module. The multi-source data perception module, the first input end of the data processing module, the first output end of the planning execution module, and the input end of the feedback monitoring module are connected in sequence. The second output end of the planning execution module is connected to the input end of the multi-source data perception module, and the output end of the feedback monitoring module is connected to the second input end of the data processing module;

[0050] The multi-source data perception module is used to obtain meteorological data and airborne data and generate an initial data set;

[0051] The data processing module is used to generate an environmental model and input the initial data set into the environmental model to obtain physical prediction data and obstacle information data;

[0052] The planning execution module is used to generate a planned path according to the physical prediction data and obstacle information data and control the flight of the UAV swarm;

[0053] The feedback monitoring module is used to monitor the flight state of the UAV swarm in real time and feed back the data to the ground control unit.

[0054] Furthermore, the multi-source data perception module obtains key data such as wind speed, wind direction, temperature, and environmental obstacles by integrating ground meteorological stations, satellite remote sensing, and airborne sensors, and generates an initial data set.

[0055] Specifically, the ground meteorological station provides real-time meteorological data such as wind speed and wind direction in the low-altitude area; the satellite remote sensing system monitors environmental changes and obstacle information in a larger range; the airborne sensors are used to capture the dynamic changes around the UAV, such as wind speed changes and flight attitudes.

[0056] Furthermore, referring to Figure 3 as shown, the environmental model includes a preprocessing unit that applies PredRNN to an encoder-decoder framework, constructs a loss function using a physical information neural network constraint mechanism, and fuses the initial data set.

[0057] Specifically, PredRNN is an advanced neural network framework designed for modeling spatio-temporal sequence data. Its core advantage lies in its ability to simultaneously capture the spatial dependencies and temporal evolution laws in the data. Compared with the traditional Convolutional Long Short-Term Memory Network (ConvLSTM), this model significantly improves the model's ability to model spatio-temporal dynamic features while retaining the spatial feature extraction ability of ConvLSTM by innovatively introducing the spatio-temporal memory unit parameter M.

[0058] Compared with the limitation in the traditional ConvLSTM model where parameters can only be propagated across time steps within the same layer or towards deeper layers at the same time step, PredRNN introduces the spatio-temporal memory unit parameter M and establishes a zigzag memory propagation path. This unique cross-spatio-temporal coupling mechanism not only effectively tracks the dynamic evolution process of spatial features over time but also significantly reduces the loss of feature information in deep networks.

[0059] Furthermore, the encoder-decoder framework includes using the spatio-temporal memory unit parameter M and establishing a zigzag memory propagation path, enabling the bottom layer of the next time step to directly obtain the deepest layer information of the previous time step.

[0060] Furthermore, to further improve the physical rationality and accuracy of physical field prediction, a Physics-Informed Neural Network (PINN) constraint mechanism is integrated. By jointly constructing a composite loss function with the integral form of the governing equation (PDE), observational data, initial conditions, and boundary conditions, the partial derivatives of the PDE-related terms are accurately calculated using automatic differentiation techniques, and the weight coefficients of each term are dynamically adjusted through Bayesian optimization, thus achieving an organic unity of data-driven and physical constraints. Taking wind field prediction as an example, the loss function is set to include:

[0061]

[0062] where u and v are the x-axis and y-axis components of the wind field velocity at time steps from t1 to t10 respectively, t is the corresponding time step, fu is the Coriolis force component (or other external force source term component) in the u direction, fv is the Coriolis force component (or other external force source term component) in the v direction, F u is the residual term of the control equation for the u component, F v is the residual term of the control equation for the v component, MSE is the total mean square error, MSE u is the mean square error corresponding to the wind speed of u, MSE v is the mean square error corresponding to the wind speed of v, is the mean square error of the residual term Fu, is the mean square error of the residual term Fv, N is the total number of data points, u i is the u-component velocity value of the i-th data point, The u - component velocity value of the i - th data point predicted by the neural network, v i is the v - component velocity value of the i - th data point, is the v - component velocity value of the i - th data point predicted by the neural network, is the F of the i - th data point in the numerical solution u residual term, is the F of the i - th data point predicted by the neural network u value (i.e., the PDE residual), is the F of the i - th data point in the numerical solution v residual term, is the F of the i - th data point predicted by the neural network v residual term.

[0063] Specifically, the loss function is set based on PINN. The training input is the physical - field components (such as the wind field, and the wind speed will be decomposed into u and v) from time step t1 to t10. These historical physical - field sequences are processed by the encoder to capture their spatio - temporal patterns. The decoder then uses the encoded information to predict the physical - field components at future time steps (from t11 to t20). This setting enables the model to capture the complex spatio - temporal interactions in the physical field. The data - processing module also includes the processing and fusion of obstacle information to generate an accurate environmental model, providing precise data for path planning.

[0064] Furthermore, the planned path integrates the flight distance, turbulence intensity, battery - energy - consumption coefficient and combines with the improved A* algorithm for path optimization. The core cost - function expression of the improved A* algorithm is:

[0065]

[0066] where H(n k ) is the estimated cost from the current node n k to the target node n m , considering the influence of wind speed. k is the current - node index, representing the k - th node in the search path, m is the index of the end node (i.e., nm is the target node), d i is the distance between nodes n i-1 and n i in spherical coordinates, v airspeed is the airspeed of the aircraft in a windless environment (i.e., the speed of the aircraft itself), v ground is the actual speed of the aircraft relative to the ground, v airspeed and v ground is the resultant velocity of the combination of v i and v i-1 . G(n i ) is the movement cost from the previous node n

[0067] Furthermore, to minimize energy consumption and risks during flight, a comprehensive cost function is designed, which combines distance cost and airflow turbulence cost. Based on wind field prediction data, the system realizes adaptive trajectory planning through the A* algorithm. This algorithm is an improvement of the Dijkstra algorithm. Different from the Dijkstra algorithm that can only blindly search all possible paths, the A* algorithm avoids ineffective search in areas far from the target by introducing a heuristic function to preferentially explore nodes expanding towards the target direction, significantly reducing the computational complexity. At the same time, considering the positive correlation between the energy consumption of the UAV and the duration of each flight stage, the system integrates wind field factors into the cost function calculation through wind speed vector decomposition and synthesis technology. The corresponding wind speed decomposition image is as shown in Figure 4 shown. By dynamically adjusting the flight route, areas with drastic changes in wind speed are avoided in real time, thus ensuring the safety and efficiency of the route planning. In addition, based on wind field prediction, the path planning will also adjust the route in real time to ensure that the UAV can fly along the optimal path.

[0068] Furthermore, during the execution process, the UAV receives the latest environmental data every 15 seconds, triggering the following closed-loop optimization:

[0069] Risk prediction: Compare the difference between the predicted wind field and the real-time monitoring, and divide it into three levels of risk: low (Δ < 10%), medium (10% ≤ Δ < 30%), and high (Δ ≥ 30%);

[0070] Dynamic replanning: In the case of medium and high risks, based on the sliding time window mechanism, locally reconstruct the path, and preferentially avoid strong shear wind areas and sudden obstacles;

[0071] Energy efficiency balance: Automatically select the flight segment with the smallest increase in energy consumption (slope change ≤ 3° / s) in the detour plan.

[0072] Furthermore, the feedback monitoring module is responsible for real-time monitoring of the flight state of the UAV and feeding back the data to the ground control unit. Through the remote connection with the user monitoring terminal, the operator can view the real-time data during the flight at any time and make adjustments according to the feedback information. The embedded anomaly detection and avoidance mechanism in the system can sense environmental changes in real time and automatically adjust the flight route to avoid potential risks. In addition, the feedback monitoring module will also generate reports based on flight data for subsequent task optimization and decision-making reference.

[0073] Furthermore, it also includes a multi-UAV cooperation module and a control unit.

[0074] The multi-UAV cooperation module is used to realize the cooperative control of the UAV swarm;

[0075] The control unit is used to receive data from the multi-source data perception module and the planning execution module, and control the startup of the data processing module and the UAV attitude.

[0076] An intelligent planning and coordination method for low-altitude air routes of unmanned aerial vehicles based on multi-source perception and edge computing, which is applied to the intelligent planning and coordination system for low-altitude air routes of unmanned aerial vehicles based on multi-source perception and edge computing described in any one of the above, includes the following steps:

[0077] Install and configure the multi-source data perception module in the unmanned aerial vehicle and the ground station system. The weather station, satellite remote sensing system, and airborne sensors in the multi-source perception module are arranged independently. The weather station and the remote sensing system are connected to the control module through a wireless network, and the airborne sensors transmit data to the control module through the sensor module. The unmanned aerial vehicle deploys the edge computing module according to the mission requirements to collect environmental data in real time;

[0078] After the control module receives the corresponding environmental data, it controls the data processing module to start, and obtains physical prediction data and obstacle information data;

[0079] Generate a planned path based on the physical prediction data and obstacle information data. The control module controls the flight of the unmanned aerial vehicle group. The path planning module performs dynamic path planning using the A* algorithm according to the processed wind field prediction data and optimizes the air route in real time. During the flight, the control module continuously receives sensor data and flight status information, and automatically adjusts the flight route according to environmental changes to avoid obstacles and extreme wind speed areas;

[0080] The feedback monitoring module monitors the flight process of the unmanned aerial vehicle group in real time.

[0081] Specifically, the system realizes the collaborative operation of multiple unmanned aerial vehicles through the multi-aircraft coordination module, and the task assignment and path planning are synchronously adjusted. The flight monitoring module monitors the flight status in real time, and feeds back the flight data to the remote monitoring end through the control unit. The operator can monitor the flight process of the unmanned aerial vehicle in real time and perform necessary interventions and adjustments. The control module records and stores all the collected data to provide support for subsequent data analysis and decision optimization. The system continuously optimizes the algorithm model according to historical data and task feedback to improve the wind field prediction accuracy and the reliability of path planning.

[0082] In a specific embodiment, the ground weather station, satellite remote sensing, and airborne sensors respectively collect real-time wind speed, obstacle distribution, and local turbulence data (update frequency 1-10 Hz), and perform spatio-temporal alignment and noise filtering through the edge computing node. The data is input into the PredRNN-PINN dual model architecture based on physical constraints, where PredRNN captures the evolution law of the temporal wind field, and PINN embeds the prior knowledge of fluid mechanics through the Navier-Stokes equation, and jointly outputs a high-precision three-dimensional wind field prediction for the next 5-30 minutes (error rate <8%). The improved A* algorithm is used to generate the initial safe path, and the closed-loop optimization is triggered to achieve a response delay of the air route planning in a complex wind field environment ≤1.2 seconds, reducing the energy consumption by more than 17% compared with the traditional method. The verification results are asFigures 5a - 7d as shown

[0083] The foregoing description of the disclosed embodiments enables those skilled in the art to implement or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An intelligent planning and coordination system for low-altitude air routes of unmanned aerial vehicles based on multi-source perception and edge computing, characterized in that, It includes a multi-source data perception module, a data processing module, a planning and execution module, and a feedback monitoring module. The multi-source data perception module, the first input end of the data processing module, the first output end of the planning and execution module, and the input end of the feedback monitoring module are connected in sequence. The second output end of the planning and execution module is connected to the input end of the multi-source data perception module. The output end of the feedback monitoring module is connected to the second input end of the data processing module; The multi-source data perception module is used to obtain meteorological data and airborne data and generate an initial data set; The data processing module is used to generate an environmental model and input the initial data set into the environmental model to obtain physical prediction data and obstacle information data; The planning and execution module is used to generate a planned path according to the physical prediction data and obstacle information data and control the flight of the UAV swarm; The feedback monitoring module is used to monitor the flight state of the UAV swarm in real time and feedback the data to the ground control unit.

2. The intelligent UAV low-altitude route planning and coordination system based on multi-source perception and edge computing according to claim 1, characterized in that The multi-source data perception module obtains corresponding key data through integrating a ground meteorological station, satellite remote sensing, and airborne sensors and generates an initial data set.

3. The intelligent UAV low-altitude route planning and coordination system based on multi-source perception and edge computing according to claim 1, characterized in that The environmental model includes a preprocessing unit that applies PredRNN to an encoder-decoder framework, constructs a loss function using a physical information neural network constraint mechanism, and fuses the initial data set.

4. The intelligent UAV low-altitude route planning and coordination system based on multi-source perception and edge computing according to claim 3, characterized in that The encoder-decoder framework includes adopting the parameters M of the spatio-temporal memory unit and establishing a zigzag memory propagation path, enabling the bottom layer of the next time step to directly obtain the deepest layer information of the previous time step; The setting of the loss function includes: Among them, u and v are the x-axis and y-axis components of the wind field wind speed from time step t1 to t10, t is the corresponding time step, fu is the Coriolis component force in the u direction, fv is the Coriolis component force in the v direction, F u is the residual term of the control equation for the u component, F v is the residual term of the control equation for the v component, MSE is the total mean square error, MSE u is the mean square error corresponding to the wind speed of u, MSE v is the mean square error corresponding to the wind speed of v, is the mean square error of the residual term Fu, is the mean square error of the residual term Fv, N is the total number of data points, u i is the u-component velocity value of the i-th data point, is the u-component velocity value of the i-th data point predicted by the neural network, v i is the v-component velocity value of the i-th data point, is the v-component velocity value of the i-th data point predicted by the neural network, is the F of the i-th data point in the numerical solution u residual term, is the F of the i-th data point predicted by the neural network u value (i.e., the PDE residual), is the F of the i-th data point in the numerical solution v residual term, is the F of the i-th data point predicted by the neural network v residual term.

5. The intelligent UAV low-altitude route planning and coordination system based on multi-source perception and edge computing according to claim 1, characterized in that The planned path fuses the flight distance, turbulence intensity, and battery energy consumption coefficient and combines an improved A* algorithm for path optimization. The core cost function expression of the improved A* algorithm is: Among them, H(n k ) is the predicted cost from the current node n k to the target node n m considering the influence of wind speed. k is the index of the current node, representing the k-th node in the search path. m is the index of the end node. d i is the distance between nodes n i-1 and n i in spherical coordinates. v airspeed is the airspeed of the aircraft in a windless environment. v ground is the actual speed of the aircraft relative to the ground. v airspeed and v ground is the combined speed. G(n i ) is the moving cost from the previous node n i-1 to the current node n i considering the wind speed factor.

6. The intelligent UAV low-altitude route planning and coordination system based on multi-source perception and edge computing according to claim 1, characterized in that It further includes a multi-aircraft coordination module and a control unit, The multi-aircraft coordination module is used to realize the coordinated control of the UAV swarm; The control unit is used to receive the data of the multi-source data perception module and the planning and execution module and control the startup of the data processing module and the attitude of the UAV.

7. A method for intelligent planning and coordination of low-altitude air routes of unmanned aerial vehicles based on multi-source perception and edge computing, which is used to execute the system for intelligent planning and coordination of low-altitude air routes of unmanned aerial vehicles based on multi-source perception and edge computing according to any one of claims 1-6, characterized in that, It includes the following steps: Install and configure the multi-source data perception module in the UAV and ground station system to collect environmental data in real time; After the control module receives the corresponding environmental data, it controls the data processing module to start and obtains physical prediction data and obstacle information data; Generate a planned path according to the physical prediction data and obstacle information data. The control module controls the flight of the UAV swarm and automatically adjusts the flight route according to environmental changes to avoid obstacles and extreme wind speed areas; The feedback monitoring module monitors the flight process of the UAV swarm in real time.

Images