Multi-algorithm unmanned aerial vehicle intelligent route planning method

Abstract

A multi-algorithm unmanned aerial vehicle intelligent route planning method comprises the steps that a route planning task is received, and a navigation target and constraint conditions of an unmanned aerial vehicle are determined; analyzing required route planning algorithm characteristics according to the navigation target and the constraint condition of the unmanned aerial vehicle; according to the analysis result, dynamically selecting one or more route planning algorithms suitable for the current task demand from a preset algorithm library; dynamically loading the selected route planning algorithm; generating an initial path of the route; defining target functions and loss functions for different navigation targets; optimizing the objective function according to the defined loss function, and adjusting the weight value of the influence factor of the navigation target; and determining an optimal route. The method solves the problem that a single algorithm is difficult to adapt to a complex scene, and improves the accuracy and efficiency of route planning. Through a dynamic loading and switching mechanism, an optimal algorithm strategy can be selected according to actual requirements, dynamic adjustment of a route is realized, and flight safety and task completion quality are improved.

Description

Technical Field

[0001] The present invention relates to unmanned aerial vehicle (UAV) route planning technology, and in particular to a multi-algorithm unmanned aerial vehicle (UAV) intelligent route planning method. Background Art

[0002] With the development of drone technology, drones are increasingly used in various industries. Route planning is an important issue in drone applications, which is related to the flight efficiency, safety and mission completion quality of drones. In the prior art, there are a variety of route planning algorithms for different application scenarios and needs, such as Dijkstra algorithm, A* algorithm, genetic algorithm, etc. However, different algorithms have different advantages and disadvantages in different scenarios, and a single algorithm is difficult to adapt to changing application needs.

[0003] It should be noted that the information disclosed in the above background technology section is only used for understanding the background of the present application, and therefore may include information that does not constitute prior art known to ordinary technicians in the field. Summary of the invention

[0004] The main purpose of the present invention is to solve the problems existing in the above-mentioned background technology and provide a multi-algorithm UAV intelligent route planning method.

[0005] To achieve the above object, the present invention adopts the following technical solutions:

[0006] In a first aspect of the present invention, a multi-algorithm UAV intelligent route planning method comprises the following steps:

[0007] S1: Receive the route planning task and determine the navigation target and constraints of the UAV;

[0008] S2: Analyze the required route planning algorithm characteristics based on the UAV’s navigation goals and constraints;

[0009] S3: According to the analysis result of step S2, dynamically select one or more route planning algorithms suitable for the current task requirements from a preset algorithm library;

[0010] S4: dynamically loading the route planning algorithm selected in step S3;

[0011] S5: Generate a preliminary path for the route using the algorithm loaded in step S4;

[0012] S6: define the objective function and loss function for different navigation goals;

[0013] S7: Applying a machine learning method to optimize the objective function according to the loss function defined in step S6, and adjusting the weight values ​​of the factors affecting the navigation target;

[0014] S8: Based on the optimization result in step S7, determine the final optimal route.

[0015] In a second aspect of the present invention, a computer program product includes a computer program, and when the computer program is executed by a processor, the multi-algorithm unmanned aerial vehicle intelligent route planning method is implemented.

[0016] The present invention has the following beneficial effects:

[0017] The present invention provides a multi-algorithm UAV intelligent route planning method, which can select different algorithm strategies according to actual needs through dynamic loading and switching mechanisms, and use multi-level algorithm strategies: first, use the initial selected algorithm such as the Dijkstra algorithm to generate a preliminary path, then optimize the path through a machine learning method, and finally generate the optimal route. The multi-algorithm UAV intelligent route planning method of the present invention solves the problem that a single algorithm is difficult to adapt to the needs of complex scenarios, and improves the accuracy and efficiency of UAV intelligent route planning.

[0018] The method of the present invention can use a unified API interface to integrate route planning algorithms, can dynamically load and switch different algorithms according to actual needs, and support the application of multi-level algorithms, including a route path planning layer under different objectives and a route weight local optimization layer, thereby improving the accuracy and efficiency of UAV route planning.

[0019] Based on the method of the present invention, a drone route engine can be provided that can integrate multiple algorithms and dynamically select the optimal algorithm according to specific circumstances. It is a multi-algorithm drone intelligent route engine that integrates multiple route intelligent planning algorithms through a unified API interface and supports pluggable algorithm output. Through this engine, different algorithms can be dynamically loaded and switched according to actual needs, and an expansion interface can be reserved to support the addition of factors affecting the route and the replacement of algorithm applications at different levels.

[0020] Other beneficial effects of the embodiments of the present invention will be further described below. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 This is an interactive diagram of a multi-algorithm UAV intelligent route engine according to an embodiment of the present invention.

[0022] Figure 2 Flow chart of preliminary path planning (first-layer path planning) of an embodiment of the present invention

[0023] Figure 3 Flow chart of path weight optimization (second layer path optimization) according to an embodiment of the present invention

[0024] Figure 4 It is a schematic diagram of weight optimization using the gradient descent method according to an embodiment of the present invention.

[0025] Figure 5 This is a data-driven optimization framework for an embodiment of the present invention.

[0026] Figure 6 The following is a pseudo code describing the implementation process of the Dijkstra algorithm according to an embodiment of the present invention.

[0027] Figure 7 Schematic diagram of machine learning objective function evaluation according to an embodiment of the present invention. DETAILED DESCRIPTION

[0028] The following is a detailed description of the embodiments of the present invention. It should be emphasized that the following description is only exemplary and is not intended to limit the scope and application of the present invention.

[0029] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of the embodiments of the present invention, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0030] See also Figures 1 to 4 The embodiment of the present invention provides a multi-algorithm UAV intelligent route planning method, comprising the following steps:

[0031] Step S1: receiving the route planning task and determining the navigation target and constraints of the UAV;

[0032] Step S2: Analyze the mission requirements and analyze the required route planning algorithm characteristics according to the UAV’s navigation objectives and constraints;

[0033] Step S3: Dynamically select an algorithm, according to the analysis result of step S2, dynamically select one or more route planning algorithms suitable for the current task requirements from a preset algorithm library;

[0034] Step S4: dynamically loading the algorithm, the route planning algorithm selected in step S3 may be dynamically loaded through the algorithm manager;

[0035] Step S5: Preliminary route generation, using the algorithm loaded in step S4 to generate a preliminary route for the route;

[0036] Step S6: path optimization preparation, defining objective functions and loss functions for different navigation goals;

[0037] Step S7: route optimization execution, applying machine learning methods, optimizing the objective function according to the loss function defined in step S6, and adjusting the weight values ​​of the factors affecting the navigation target;

[0038] Step S8: Generate the optimal route. Based on the optimization result in step S7, determine the final optimal route.

[0039] In some embodiments, the navigation goals include optimal time, shortest distance or minimum risk; wherein, the characteristic parameters related to optimal time include the distance of the horizontal distance, the true height of the route, the congestion of the route, the speed of the drone, etc., and the weight of each route section is related to the characteristic parameters related to optimal time; the characteristic parameters related to the shortest distance include the distance of the horizontal distance, the true height of the route, etc., and the weight of each route section is related to the characteristic parameters related to the shortest distance; the characteristic parameters related to the minimum risk include weather conditions, the complexity of the route terrain, etc., and the weight of each route section is related to the characteristic parameters related to the minimum risk.

[0040] In a preferred embodiment, the generated route can also be dynamically adjusted in combination with real-time mission data and environmental perception data to improve flight safety and mission completion quality.

[0041] Figure 1 The drone route engine provided by the method of the present invention is shown, which can integrate multiple algorithms and dynamically select the optimal algorithm according to specific circumstances. It is a multi-algorithm drone intelligent route engine that integrates multiple route intelligent planning algorithms through a unified API interface and supports pluggable algorithm output. Through this engine, different algorithms can be dynamically loaded and switched according to actual needs, and an extended interface can be reserved to support the addition of factors affecting the route and the replacement of algorithm applications at different levels. As a further optimization, an extensible design is adopted, and an extended interface is reserved to support the addition of new influencing factors and the integration of new route planning algorithms.

[0042] See also Figure 1 In a preferred embodiment, a target influencing factor library is pre-built to store the corresponding relationship between different navigation targets and related influencing factors; in step S2, when performing the operation of analyzing task requirements, the navigation target and related influencing factors are input through a unified API interface, and the influencing factors and weight values ​​are returned. A target influencing factor weight library is pre-built to store and optimize the path weight; in step S7, when performing the operation of path optimization execution, the path weight optimization layer is called through a unified API interface to optimize the path weight, and the path weight optimization layer is configured to adjust and optimize the weights of various influencing factors in the route path.

[0043] See also Figure 2In a preferred embodiment, in step S5, the preliminary path generation uses the Dijkstra algorithm as the default basic algorithm, calculates the weight value of each route segment according to the user-set target and the weight factor configured by the system, and generates a preliminary path. The preliminary path generation specifically includes: determining the navigation target, selecting one of the targets with the best time, the shortest distance or the lowest risk; obtaining the influencing factors affecting the navigation target, including various parameters that may affect the route planning; assigning an initial weight to each influencing factor, and the weight represents the importance of each influencing factor in the route planning; according to the navigation target, the influencing factors and their weights, using the weight function to calculate the weight value of each route segment; wherein the weight of the influencing factor evolves through an iterative process to obtain the optimal value; applying the Dijkstra algorithm, using the calculated weight value of each route segment to generate a preliminary path from the starting point to the end point of the route.

[0044] Further preferably, during the execution of the Dijkstra algorithm, each time the node closest to the starting node is selected from the priority queue for expansion, and the shortest distances of other nodes are dynamically updated. The result finally returned includes the shortest distance of each node and the predecessor node of each node, which can be used to trace back the shortest path (see Figure 6 Pseudocode shown).

[0045] See also Figure 3 In a preferred embodiment, in step S7, the path optimization execution specifically includes: defining the objective function, preliminarily deriving the objective function based on the navigation goal and related influencing factors; constructing a loss function, using the loss function to evaluate the quality of the objective function, and determining the deviation between the objective function and the actual navigation goal; executing the gradient descent method, iteratively optimizing the objective function through the gradient descent method, and gradually adjusting the weight value of each influencing factor; evaluating the optimization result, using the loss function to evaluate the optimization result after each iteration to determine whether it is close to the optimal solution; determining the optimal weight value, determining the optimal weight value of each influencing factor through the gradient descent method to form an optimal solution function; generating an optimized path, applying the optimal weight value to adjust the preliminary path, and generating an optimized path that meets the navigation goal.

[0046] The gradient descent method specifically includes the following process: calculating the gradient of the weight and bias relative to the loss function for each training data; calculating the sum of the gradients of all training data weights; calculating the sum of the gradients of all training data biases; calculating the gradient average of the weights and biases of all samples; updating the weights and biases according to the gradient average; repeating the gradient calculation and weight update process until the loss function converges. Figure 4 It is a schematic diagram of weight optimization using the gradient descent method according to an embodiment of the present invention.

[0047] See also Figure 5In some embodiments, the method of the present invention is implemented based on a data-driven optimization framework, including the following data-driven implementation process: data set segmentation, dividing the collected data into a training data set and a test data set for model training and verification; preliminary analysis, performing preliminary analysis on the training data to identify key features that affect route planning; feature engineering, performing feature selection and construction based on the results of the preliminary analysis; model selection, selecting a suitable machine learning model according to the navigation objectives and the results of feature engineering; setting parameters, setting necessary parameters for the selected model to adapt to specific route planning objectives; model training, training the model using the training data set; model testing, testing the trained model using the test data set; model prediction, applying the trained and tested model to predict and optimize the initially generated path to generate an optimized path that better meets the navigation objectives; model evaluation and iteration, evaluating the performance of the model according to the test results, reselecting the model or adjusting the parameters if necessary, and repeating the training and testing process.

[0048] The multi-algorithm UAV intelligent route planning method of the present invention can adapt to changing application requirements and complex scenarios. The method realizes the integration and dynamic loading of multiple route planning algorithms, so that the UAV can flexibly switch and select the most suitable route planning algorithm according to actual flight conditions and mission requirements. The method adopts a multi-level algorithm strategy, first using an initial selected algorithm such as the Dijkstra algorithm to generate a preliminary path, and then using a machine learning method to optimize these paths to generate the optimal route. This method not only improves the accuracy and efficiency of route planning, but also enhances the flight efficiency and safety of the UAV, while ensuring the quality of task completion. An extensible interface is designed based on the route planning engine of the present invention to support the integration of new technologies and the application of new algorithms in the future. In general, the present invention significantly improves the flexibility and optimization of the application of UAVs in various industries through intelligent route planning, showing a wide range of application prospects and significant market value.

[0049] The specific embodiments of the present invention are further described below.

[0050] The present invention provides a multi-algorithm unmanned aerial vehicle intelligent route planning method, designs a multi-algorithm unmanned aerial vehicle intelligent route engine, including a dynamic loading and switching mechanism, an extensibility design, and a multi-level algorithm application strategy, which can integrate multiple route planning algorithms through a unified API interface to achieve pluggable algorithm output. The dynamic loading and switching mechanism can dynamically load and switch the corresponding route planning algorithm through configuration items. The extensibility design reserves an extension interface to support the addition of factors affecting the route and algorithms at different levels. The multi-level algorithm application strategy includes two levels: preliminary path generation and path optimization. The preliminary path generation layer preferably adopts the Dijkstra algorithm, calculates the weight value of each route according to the user-set target and the weight factor configured by the system, and generates a preliminary path. The path optimization layer optimizes the objective function under different navigation targets and the optimal solution function between various factors through a machine learning method to obtain the final optimal route.

[0051] 1. Dynamic loading and switching mechanism: Since it is a drone route planning engine that integrates multiple algorithms, the corresponding algorithms can be dynamically loaded and switched through configuration items to observe the advantages and disadvantages of different algorithms in route calculation under different circumstances. The present invention configures the Dijkstra algorithm in the first layer of route path planning layer by default (it can also be switched to other path planning algorithms), and the second layer of route weight local optimization layer uses the gradient descent method to optimize the route weight W (it can also be switched to other machine learning optimization algorithms).

[0052] 2. Scalability: An expansion interface is reserved to support adding factors that affect the route and adding algorithms at different levels.

[0053] 3. Multi-level algorithm is applicable:

[0054] The first layer is used for preliminary route generation, which is calculated using the basic Dijkstra algorithm. This layer first determines the navigation goal (optimal time / shortest distance / minimum risk), and generates the weight value W of each route according to the goal determined by the user and the weight factor configured by the system, that is, the weight W of each route. The weight function is:

[0055] W=w1x1+w2x 2+ w3x 3+......+ w n x n

[0056] Among them, x1, x2, ..., xn are the factors affecting the target, w1, w2, ..., wn are the weights of the factors, and the weights of the factors will be given default values ​​at the beginning, and then the optimal values ​​will be evolved through iteration. Then, the optimal route of the route composition is calculated according to the Dijkstra algorithm, and the route is recommended.

[0057] The second layer is used for path optimization, which is mainly to optimize the objective function under different navigation goals (optimal time / shortest distance / minimum risk) and the optimal solution function between various factors. According to the general practice of machine learning, the objective function is first preliminarily derived based on the factor x and the target y, and then the loss function is used to evaluate the quality of the objective function. The gradient descent method is used to continuously approach the optimal solution neighborhood, accelerate the convergence to the optimal solution, and obtain the optimal solution function, that is, the weight value of each factor.

[0058] Dynamic loading and switching mechanism

[0059] The engine has a dynamic loading and switching mechanism, which can flexibly load different route planning algorithms through configuration items. Configuration items include algorithm name, algorithm parameters, etc., which users can configure according to actual needs. When the engine is running, it dynamically loads the corresponding algorithm according to the configuration items to perform route planning calculations. In different application scenarios, by switching different algorithms, the advantages and disadvantages of their calculation results are compared, and the optimal algorithm is selected for practical application.

[0060] Scalable design

[0061] In order to adapt to new requirements and new algorithms that may appear, the engine is designed with reserved extension interfaces. The extension interfaces include the interface for factors affecting the route and the algorithm extension interface. The interface for factors affecting the route allows users to add new influencing factors, such as wind speed, obstacle distribution, etc. The algorithm extension interface supports users to add new route planning algorithms. New algorithms can be integrated into the engine through simple interface calls to achieve integration with existing algorithms.

[0062] Multi-level algorithm application

[0063] The multi-algorithm UAV intelligent route engine adopts a multi-level algorithm application strategy in the route planning process, including:

[0064] 1. Preliminary path generation

[0065] Using the Dijkstra algorithm, the weight value W of each route is calculated according to the user-set goals (such as optimal time, shortest distance, and minimum risk) and the weight factors configured by the system.

[0066] The initial weight values ​​are set based on empirical values ​​and optimized in subsequent iterations.

[0067] The preliminary path is calculated through the Dijkstra algorithm, providing a basis for subsequent optimization.

[0068] Figure 6The pseudo code shown describes the implementation process of Dijkstra's algorithm. During the execution of the algorithm, the node closest to the starting node is selected from the priority queue for expansion each time, and the shortest distances of other nodes are dynamically updated. The final result returned includes the shortest distance of each node and the predecessor node of each node, which can be used to trace back the shortest path.

[0069] The use of Dijkstra's algorithm for initial path generation has the following benefits: Considering the weight of the edge, and selecting the edge with the shortest distance from the current node to the starting node at each step. The concept of weight is introduced, which provides a unified mathematical standard for weight optimization of different goals in the later stage. In addition, the Dijkstra algorithm has better performance in the absence of negative weight edges, and its time complexity is usually O((V+E)logV), where V is the number of nodes and E is the number of edges.

[0070] 2. Path Optimization

[0071] Based on the initially generated path, further optimization is performed for different navigation goals (optimal time, shortest distance, minimum risk). First, the specific goal of the route navigation is determined, and the feature engineering is preliminarily analyzed according to the goal, so as to select different models and set the feature parameters different from the goal for training (see Figure 6 ).

[0072] Using machine learning methods, the objective function is initially derived, and the quality of the objective function is evaluated by constructing a loss function and test data in the test data set. Figure 7 shown.

[0073] The following distance loss functions, L1 and L2 loss functions can be used to optimize the objective function during machine learning model training:

[0074]

[0075] When solving the minimum value of the loss function, the gradient descent method (GD) can be used to iterate step by step to obtain the minimized loss function and model parameter value, continuously adjust the weight value of each influencing factor, gradually approach the optimal solution, and finally determine the optimal path.

[0076] The gradient descent method iteratively updates the parameters along the negative direction of the gradient to minimize the objective function. It is an optimization algorithm used to find the minimum value of a function. Its core idea is to use the gradient information of the objective function to guide the update of the parameters, thereby gradually approaching the minimum value of the function.

[0077] Suppose there is a differentiable objective function f(θ), where θ is a parameter vector. The goal of the gradient descent method is to find a set of parameters θ such that f(θ) reaches the minimum value. It points to the direction where the function grows fastest at θ, so the function value can be gradually reduced by updating θ in the opposite direction of the gradient.

[0078] The update formula of the gradient descent method is:

[0079]

[0080] Where α is the learning rate, which is used to control the step size of the update. It can be obtained by taking partial derivatives:

[0081]

[0082] The calculation process of the gradient descent method is as follows:

[0083] for i=0to the number of training data:

[0084] 1. Calculate the gradient of the weight and bias b of the i-th training data relative to the loss function. Finally, the gradient value of the weight and bias of each training data will be obtained.

[0085] 2. Calculate the sum of the gradients of all training data weights.

[0086] 3. Calculate the sum of the gradients of all training data deviations.

[0087] After completing the above calculations, start the following calculations:

[0088] 1. Using the results obtained in steps 2 and 3 above, calculate the average of the gradients of the weights and biases of all samples.

[0089] 2. Use the following formula to update the weight and bias value of each sample.

[0090]

[0091] 3. Repeat the above process until the loss function converges and remains unchanged.

[0092] Examples

[0093] Example 1: Route planning based on time optimization

[0094] In a drone delivery system, the delivery time is required to be shortened as much as possible, and the goal of the system configuration is time optimization. First, the weight value W of each route is calculated based on the goal set by the user and the weight factor configured by the system. Then, the Dijkstra algorithm is used to calculate the preliminary path. Finally, for the time optimization goal, the path is further optimized through machine learning methods to obtain the final optimal route.

[0095] The characteristic parameters that are more relevant to time are factors such as the distance of the horizontal distance, the true height of the route, the congestion of the route, the speed of the drone, etc. Therefore, the weight W of each route is related to these characteristic parameters.

[0096] Example 2: Route planning based on the shortest distance

[0097] In another application scenario, drones are used for long-distance inspections, which requires the flight distance to be shortened as much as possible. The system is configured to achieve the shortest distance. First, the weight value W of each route is calculated, and the preliminary path is calculated using the Dijkstra algorithm. Then, for the goal of the shortest distance, machine learning methods are applied to optimize the path, and finally the optimal route for the shortest path is determined.

[0098] The characteristic parameters that are more relevant to the route are factors such as the distance of the horizontal route and the true altitude of the route. Therefore, the weight W of each route segment is related to these characteristic parameters.

[0099] Example 3: Route planning based on minimum risk

[0100] For drone missions that require crossing complex environments, minimizing risk is the main goal. The system sets initial weights based on environmental factors and uses the Dijkstra algorithm to calculate a preliminary path. On this basis, machine learning methods are used to optimize the goal of minimizing risk and obtain the safest route.

[0101] The characteristic parameters that are more relevant to safety are factors such as weather conditions and route terrain complexity. Therefore, the weight W of each route is related to these characteristic parameters.

[0102] The present invention integrates multiple route planning algorithms through a unified API interface, and realizes the dynamic loading and switching, scalability design and multi-level algorithm application strategy of the multi-algorithm UAV intelligent route engine. The engine can select the optimal algorithm according to actual needs and generate the optimal route, which improves the flexibility and optimization of UAV route planning and has broad application prospects and market value.

[0103] The performance of the multi-algorithm UAV intelligent route engine can be further optimized, and more machine learning algorithms and data analysis technologies can be combined to continuously improve the intelligence and adaptability of route planning. By introducing real-time data and environmental perception technology, dynamic adjustment of UAV route planning can be achieved to improve flight safety and mission completion quality.

[0104] An embodiment of the present invention further provides a storage medium for storing a computer program, which at least performs the above method when executed.

[0105] An embodiment of the present invention further provides a control device, comprising a processor and a storage medium for storing a computer program; wherein the processor is configured to execute at least the method described above when executing the computer program.

[0106] An embodiment of the present invention further provides a processor, wherein the processor executes a computer program and at least executes the method described above.

[0107] The storage medium can be implemented by any type of non-volatile storage device, or a combination thereof. Among them, the non-volatile memory can be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), a magnetic random access memory (FRAM), a ferromagnetic random access memory (Flash Memory), a magnetic surface memory, an optical disc, or a compact disc read-only memory (CD-ROM); the magnetic surface memory can be a disk memory or a tape memory. The storage medium described in the embodiments of the present invention is intended to include, but is not limited to, these and any other suitable types of memory.

[0108] In the several embodiments provided by the present invention, it should be understood that the disclosed systems and methods can be implemented in other ways. The device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as: multiple units or components can be combined, or can be integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the components shown or discussed can be through some interfaces, and the indirect coupling or communication connection of the devices or units can be electrical, mechanical or other forms.

[0109] The units described above as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units; some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0110] In addition, all functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may be separately used as a unit, or two or more units may be integrated into one unit; the above-mentioned integrated units may be implemented in the form of hardware or in the form of hardware plus software functional units.

[0111] A person skilled in the art can understand that: all or part of the steps of implementing the above method embodiment can be completed by hardware related to program instructions, and the aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps of the above method embodiment; and the aforementioned storage medium includes: mobile storage devices, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), disks or optical disks, etc. Various media that can store program codes.

[0112] Alternatively, if the above-mentioned integrated unit of the present invention is implemented in the form of a software function module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiment of the present invention can be essentially or partly reflected in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as mobile storage devices, ROM, RAM, magnetic disks or optical disks.

[0113] The methods disclosed in the several method embodiments provided by the present invention can be arbitrarily combined without conflict to obtain new method embodiments.

[0114] The features disclosed in several product embodiments provided by the present invention can be arbitrarily combined without conflict to obtain new product embodiments.

[0115] The features disclosed in several method or device embodiments provided by the present invention can be arbitrarily combined without conflict to obtain new method embodiments or device embodiments.

[0116] The above contents are further detailed descriptions of the present invention in combination with specific preferred embodiments, and it cannot be determined that the specific implementation of the present invention is limited to these descriptions. For those skilled in the art of the present invention, several equivalent substitutions or obvious variations can be made without departing from the concept of the present invention, and the performance or use is the same, which should be regarded as belonging to the protection scope of the present invention.

Claims

1. A multi-algorithm UAV intelligent route planning method, characterized in that: The following steps are involved: S1: Receive the route planning task and determine the navigation target and constraints of the UAV; S2: Analyze the required route planning algorithm characteristics based on the UAV’s navigation goals and constraints; S3: According to the analysis result of step S2, dynamically select one or more route planning algorithms suitable for the current task requirements from a preset algorithm library; S4: dynamically loading the route planning algorithm selected in step S3; S5: Generate a preliminary path for the route using the algorithm loaded in step S4; S6: define the objective function and loss function for different navigation goals; S7: Applying a machine learning method to optimize the objective function according to the loss function defined in step S6, and adjusting the weight values ​​of the factors affecting the navigation target; S8: Based on the optimization result in step S7, determine the final optimal route.

2. The multi-algorithm UAV intelligent route planning method according to claim 1, characterized in that: Also includes: Pre-build a target influencing factor library to store the corresponding relationship between different navigation targets and related influencing factors; In step S2, when performing the operation of analyzing the task requirements, the navigation target and related influencing factors are input through the unified API interface, and the influencing factors and weight values ​​are returned; Pre-build a target influencing factor weight library to store and optimize path weights; In step S7, when performing the path optimization operation, the path weight optimization layer is called through the unified API interface to optimize the path weight, and the path weight optimization layer is configured to adjust and optimize the weights of various influencing factors in the route path.

3. The multi-algorithm UAV intelligent route planning method according to claim 1 or 2, characterized in that: In step S5, the initial path is generated using the Dijkstra algorithm as the default basic algorithm, and the weight value of each route segment is calculated according to the target set by the user and the weight factor configured by the system, and the initial path is generated.

4. The multi-algorithm UAV intelligent route planning method according to claim 3, characterized in that: In step S5, the preliminary path generation specifically includes: Determine the navigation target and choose one with the best time, shortest distance or least risk; Obtain the factors that affect the navigation objectives, including various parameters that may affect route planning; Assigning an initial weight to each influencing factor, wherein the weight represents the importance of each influencing factor in route planning; According to the navigation objectives, influencing factors and their weights, the weight value of each route is calculated using a weight function; wherein the weight of the influencing factors is evolved through an iterative process to obtain the optimal value; The Dijkstra algorithm is applied to generate a preliminary path from the starting point to the end point of the route using the calculated weight value of each route segment.

5. The multi-algorithm UAV intelligent route planning method according to claim 4, characterized in that: During the execution of the Dijkstra algorithm, the node closest to the starting node is selected from the priority queue for expansion each time, and the shortest distances of other nodes are dynamically updated. The final returned result includes the shortest distance of each node and the predecessor node of each node, which can be used to trace back the shortest path.

6. The multi-algorithm UAV intelligent route planning method according to any one of claims 1 to 5, characterized in that: In step S7, the path optimization execution specifically includes: Define the objective function and preliminarily derive the objective function based on the navigation goal and related influencing factors; Construct a loss function, use the loss function to evaluate the quality of the objective function, and determine the deviation between the objective function and the actual navigation goal; Execute the gradient descent method, iteratively optimize the objective function through the gradient descent method, and gradually adjust the weight value of each influencing factor; Evaluate the optimization results and use the loss function to evaluate the optimization results after each iteration to determine whether they are close to the optimal solution; Determine the optimal weight value, and determine the optimal weight value of each influencing factor through the gradient descent method to form the optimal solution function; Generate an optimized path, apply the optimal weight value to adjust the preliminary path, and generate an optimized path that meets the navigation objectives.

7. The multi-algorithm UAV intelligent route planning method according to claim 6, characterized in that: In step S7, the gradient descent method is specifically The process includes: Calculate the gradient of weights and biases with respect to the loss function for each training example; Calculate the sum of the gradients of all training data weights; Calculate the sum of the gradients of all training data deviations; Calculate the average gradient of weights and biases for all samples; Update weights and biases based on the average of gradients; The gradient calculation and weight update process are repeated until the loss function converges.

8. The multi-algorithm UAV intelligent route planning method according to any one of claims 1 to 7, characterized in that: Said navigation objectives include optimal time, shortest distance or minimum risk; Among them, the characteristic parameters related to time optimization include the distance of the horizontal distance, the true height of the route, the congestion of the route, and the speed of the drone. The weight of each route is related to the characteristic parameters related to time optimization. Among them, the characteristic parameters related to the shortest distance include the distance of the horizontal distance and the true height of the route. The weight of each route is related to the characteristic parameters related to the shortest distance. Among them, the characteristic parameters related to the minimum risk include weather conditions and route terrain complexity, and the weight of each route is related to the characteristic parameters related to the minimum risk.

9. The multi-algorithm UAV intelligent route planning method according to any one of claims 1 to 8, characterized in that: The method is implemented based on a data-driven optimization framework, including the following data-driven implementation process: Split the data set into training data sets and test data sets for model training and verification; Preliminary analysis, conduct preliminary analysis of the training data to identify key features that affect route planning; feature Engineering, based on the results of the preliminary analysis, feature selection and construction; Select a model, and choose a suitable machine learning model based on the navigation objectives and the results of feature engineering; Set parameters, set necessary parameters for the selected model to suit specific route planning objectives; Model training, using the training data set to train the model; Model testing: Use the test data set to test the trained model; Model prediction: Use the trained and tested model to predict and optimize the initially generated path to generate an optimized path that better meets the navigation objectives; Model evaluation and iteration: Evaluate the performance of the model based on the test results. If necessary, reselect the model or adjust the parameters and repeat the training and testing process.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the multi-algorithm UAV intelligent route planning method as described in any one of claims 1 to 9 is implemented.

Images