Unmanned aerial vehicle path planning method and device based on dynamic weight and medium

Abstract

The invention relates to an unmanned aerial vehicle path planning method and device based on dynamic weight and a medium, and belongs to the technical field of unmanned aerial vehicle path planning, and the method comprises the following steps: carrying out the three-dimensional modeling of an unmanned aerial vehicle flight area, obtaining an environment three-dimensional model, determining an obstacle point, a starting point and a target point of the environment three-dimensional model, and dividing a no-fly area according to the obstacle point, the starting point and the target point. And then, an improved A * algorithm is adopted to plan a global initial path, the improvement is that a dynamic weight is obtained based on the starting point, the target point and the no-fly zone, and the dynamic weight is used to optimize a cost function of the A * algorithm. And training a Gaussian process regression model by using the global initial path to obtain a global path. Meanwhile, a gravitational repulsion strategy is formulated, and the unmanned aerial vehicle flies to a target point from a starting point according to the strategy and a global path. The dynamic weight can be adaptively adjusted according to dynamic factors such as the obstacle density and the starting point-end point distance, the problems of blind search and node explosion of a traditional fixed weight in an obstacle dense area are solved, and the efficiency and accuracy of path planning are improved.

Description

Technical Field

[0001] This invention relates to a method, device, and medium for unmanned aerial vehicle (UAV) path planning based on dynamic weights, belonging to the field of UAV path planning technology. Background Technology

[0002] With the rapid development of the low-altitude economy and the increasing urgency of intelligent needs, the large-scale application of drones in scenarios such as logistics delivery, emergency rescue, topographic mapping, and urban security has become an inevitable trend. However, efficient and safe flight in complex three-dimensional environments remains the core bottleneck restricting the large-scale deployment of drones: the dense presence of dynamic obstacles such as urban canyons, mountain forests, and temporary control zones makes it difficult for traditional two-dimensional plane or static map path planning methods to meet real-time and safety requirements; at the same time, the limited payload, endurance, and computing power of drones themselves place higher demands on the computational efficiency and smoothness of path algorithms.

[0003] In recent years, there has been a lot of research on path planning for UAVs, such as the A* algorithm. However, the heuristic cost function of the traditional A* algorithm usually uses fixed weights, which cannot be adaptively adjusted according to dynamic factors such as obstacle density and the distance between the start and end points. It is easy to get stuck in "blind search" in areas with dense obstacles, resulting in node explosion. At the same time, the 4 / 8 neighborhood search strategy has low efficiency in expanding in three-dimensional space, and the reverse search domain wastes a lot of computing power. Summary of the Invention

[0004] To address the problems existing in the prior art, this invention proposes a method, device, and medium for UAV path planning based on dynamic weights.

[0005] The technical solution of the present invention is as follows:

[0006] On the one hand, this invention provides a UAV path planning method based on dynamic weights, comprising the following steps:

[0007] A 3D environmental model is obtained by performing 3D modeling on the flight area of ​​the UAV;

[0008] Determine obstacle points in the 3D environmental model, as well as the starting and target points of the UAV, and delineate no-fly zones based on the obstacle points;

[0009] An improved A* algorithm is used to obtain the global initial path from the starting point to the target point; wherein, the improvement of the A* algorithm includes: obtaining dynamic weights based on the starting point, the target point and the no-fly zone, and using the dynamic weights to optimize the cost function of the A* algorithm;

[0010] A Gaussian process regression model is trained using a global initial path, and a global path is obtained based on the Gaussian process regression model.

[0011] Develop a gravity-repulsion strategy to guide drones to avoid no-fly zones;

[0012] The drone starts from the starting point and executes the gravity repulsion strategy and global path to fly to the target point.

[0013] Preferably, a three-dimensional rasterized map method is used to perform three-dimensional modeling of the UAV's flight area.

[0014] Preferably, improvements to the A* algorithm include:

[0015] Discard the search area in the opposite direction between the drone's current point and the target point, and expand the search area in the same direction between the drone's current point and the target point.

[0016] Preferably, dynamic weights are obtained based on the starting point, target point, and no-fly zone, and the cost function of the A* algorithm is optimized using these dynamic weights. The specific steps are as follows:

[0017] Obtain the proportion of the no-fly zone in the 3D environmental model;

[0018] Calculate the distance from the starting point to the target point and the distance from the current point of the drone to the target point;

[0019] Dynamic weights are obtained based on the aforementioned ratio, the distance from the starting point to the target point, and the distance from the current point of the UAV to the target point.

[0020] During the cost function calculation process of the A* algorithm, the dynamic weights are assigned to the estimated cost from the current point to the target point.

[0021] Preferably, a Gaussian process regression model is trained using a global initial path, and a global path is obtained based on the Gaussian process regression model. The specific steps are as follows:

[0022] The initial path arc length is obtained based on the global initial path.

[0023] Among them, the node s with initial path arc length i With node P of the global initial path i correspond;

[0024] The initial path arc length is used as the training set to train the Gaussian process regression model;

[0025] Among them, the covariance function of the Gaussian process regression model is a radial basis function constructed based on the initial path arc length;

[0026] The initial path arc length is predicted using a trained Gaussian process regression model to obtain the smooth path arc length;

[0027] Obtain the curvature and curvature variance of each node in the arc length of the smooth path;

[0028] Set curvature threshold and curvature variance threshold, and delete the nodes in the global initial path corresponding to the smooth path arc length whose curvature is less than the curvature threshold or whose curvature variance is less than the curvature variance threshold;

[0029] Construct the remaining nodes of the global initial path into a global path.

[0030] Preferably, the gravitational repulsion strategy includes a gravitational strategy and a repulsion strategy;

[0031] The gravity strategy guides the UAV to fly along a global path by calculating gravity.

[0032] The repulsion strategy guides the drone to avoid no-fly zones by calculating the repulsion force.

[0033] Preferably, the method for calculating the gravitational force is as follows:

[0034] Calculate the Euclidean distance between the current point of the drone and the target point;

[0035] Set the saturation speed control coefficient and the radius of influence of no-fly zones on drones;

[0036] Gravity is obtained based on the Euclidean distance between the current point and the target point of the UAV, the saturation velocity control coefficient, and the influence radius of the no-fly zone on the UAV.

[0037] Preferably, the method for calculating the repulsive force is as follows:

[0038] Calculate the Euclidean distance between the drone's current point and the no-fly zone;

[0039] Set the repulsive force variation control coefficient;

[0040] The repulsive force is obtained based on the influence radius of the no-fly zone on the drone, the repulsive force change control coefficient, the Euclidean distance between the drone's current point and the target point, and the Euclidean distance between the drone's current point and the no-fly zone.

[0041] In another aspect, the present invention also provides an electronic device having a computer program stored thereon, which, when executed by a processor, implements the UAV path planning method based on dynamic weights as described in any embodiment of the present invention.

[0042] In another aspect, the present invention also provides a computer-readable storage medium for storing one or more programs, which, when executed by one or more processors, cause the one or more processors to implement the UAV path planning method based on dynamic weights as described in any embodiment of the present invention.

[0043] The present invention has the following beneficial effects:

[0044] 1. This invention obtains dynamic weights based on the starting point, target point, and no-fly zone, and assigns these weights to the estimated cost from the current point to the target point in the A* algorithm's cost function. These dynamic weights can adaptively adjust according to dynamic factors such as obstacle density and the distance between the starting and ending points, avoiding the problem of traditional fixed weights leading to node explosion due to "blind search" in densely obstacle-filled areas, thus further improving the efficiency and accuracy of path planning.

[0045] 2. This invention utilizes a global initial path to train a Gaussian process regression model and obtains a global path based on this model. Parameters are optimized using the marginal likelihood maximization method, with the radial basis function constructed based on the initial path arc length serving as the covariance function to predict the smoothed path arc length. Then, the curvature and curvature variance of each node along the smoothed path arc length are obtained, and thresholds are set to filter nodes and remove redundant nodes. The resulting global path is smoother, reducing frequent turns and speed fluctuations during UAV flight, thus improving flight stability and safety.

[0046] 3. In the gravity-repulsion strategy formulated in this invention, the gravity strategy guides the UAV to fly along a global path by calculating gravity, while the repulsion strategy guides the UAV to avoid no-fly zones by calculating repulsion. When calculating gravity and repulsion, various factors are comprehensively considered, including the Euclidean distance between the UAV's current point and the target point, the no-fly zone, the saturation velocity control coefficient, the influence radius of the no-fly zone, and the repulsion change control coefficient. This allows the UAV to more accurately avoid obstacles during flight without deviating too far from the planned path, improving the safety and reliability of path planning. Attached Figure Description

[0047] Figure 1 This is a flowchart illustrating the implementation of the method of the present invention.

[0048] Figure 2 This is a schematic diagram illustrating the search field of this invention. Detailed Implementation

[0049] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0050] It should be understood that the step numbers used in the text are for ease of description only and are not intended to limit the order in which the steps are performed.

[0051] It should be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0052] The terms “comprising” and “including” indicate the presence of the described feature, whole, step, operation, element and / or component, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or collections thereof.

[0053] The term “and / or” refers to any combination of one or more of the associated listed items, as well as all possible combinations, and includes these combinations.

[0054] Example 1:

[0055] See Figure 1 This invention provides a method for UAV path planning based on dynamic weights, comprising the following steps:

[0056] A 3D environmental model is obtained by performing 3D modeling on the flight area of ​​the UAV;

[0057] Determine obstacle points in the 3D environmental model, as well as the starting and target points of the UAV, and delineate no-fly zones based on the obstacle points;

[0058] An improved A* algorithm is used to obtain the global initial path from the starting point to the target point; wherein, the improvement of the A* algorithm includes: obtaining dynamic weights based on the starting point, the target point and the no-fly zone, and using the dynamic weights to optimize the cost function of the A* algorithm;

[0059] A Gaussian process regression model is trained using a global initial path, and a global path is obtained based on the Gaussian process regression model.

[0060] Develop a gravity-repulsion strategy to guide drones to avoid no-fly zones;

[0061] The drone starts from the starting point and executes the gravity repulsion strategy and global path to fly to the target point.

[0062] Preferably, a three-dimensional rasterized map method is used to perform three-dimensional modeling of the UAV's flight area.

[0063] Using a 3D rasterized map method to create a 3D model of the UAV's flight area can accurately represent the terrain and obstacle distribution in a complex 3D environment, providing detailed and accurate environmental information for subsequent path planning. This enables the technical solution to adapt to various complex flight scenarios, such as urban canyons and mountain forests.

[0064] The A* algorithm is an efficient heuristic search algorithm for finding the shortest path in a graph or grid. Its core idea is to use a heuristic function to guide the search direction, ensuring the shortest path is found while reducing unnecessary computation.

[0065] See Figure 2 Preferably, improvements to the A* algorithm include:

[0066] Discard the search area in the opposite direction between the drone's current point and the target point, and expand the search area in the same direction between the drone's current point and the target point.

[0067] Figure 2 In the diagram, yellow nodes represent the drone's current location, blue nodes marked with the word "target" represent the drone's target location, several red nodes represent the extended search area, and several blue nodes without text represent no-fly zones.

[0068] The original A* algorithm had 4 or 8 expansion directions for each node, referred to as 4-neighborhood and 8-neighborhood search. This embodiment improves the A* algorithm's search neighborhood to a "fathead fish" neighborhood, such as... Figure 2 The red node shown resembles a fathead fish.

[0069] By discarding the search domain in the opposite direction between the current point and the target point of the UAV and expanding the search domain in the same direction, the traditional 4-neighborhood and 8-neighborhood search is improved into "fathead fish" neighborhood search, thereby reducing unnecessary search directions and nodes, improving search efficiency, and enabling the global initial path from the starting point to the target point to be found more quickly.

[0070] In one embodiment, a "fathead fish" neighborhood search is established at the current point and the target point of the drone, respectively. The search starts from the starting point and the ending point, respectively, and two open sets are maintained. Then, they try to approach each other in two directions until they meet at some intermediate node, which improves the search efficiency.

[0071] Preferably, dynamic weights are obtained based on the starting point, target point, and no-fly zone, and the cost function of the A* algorithm is optimized using these dynamic weights. The specific steps are as follows:

[0072] Obtain the proportion of the no-fly zone in the 3D environmental model;

[0073] Calculate the distance from the starting point to the target point (specifically, the Euclidean distance) and the distance from the current point of the UAV to the target point (specifically, the Euclidean distance);

[0074] The dynamic weight is obtained based on the aforementioned ratio, the distance from the starting point to the target point, and the distance from the current point of the UAV to the target point, and is expressed by the formula:

[0075]

[0076] In the formula, ω represents the dynamic weight, D represents the distance from the starting point to the target point, and d now The distance from the current point of the drone to the target point is represented by α, the proportion of the no-fly zone in the 3D environmental model is represented by β, and the empirical adjustment coefficient is represented by β (adjusted multiple times according to the actual situation to the optimal value, or it can be set in segments by weight).

[0077] In the cost function calculation process of the A* algorithm, the dynamic weights are assigned to the estimated cost from the current point to the target point, expressed by the formula:

[0078] F sum =G now +ωH now ;

[0079] In the formula, F sum Let G represent the cost function. now H represents the actual cost from the starting point to the current point. now This represents the estimated price from the current point to the target point.

[0080] The dynamic weights are obtained based on multiple factors such as the starting point, the target point, and no-fly zones. They can be dynamically adjusted according to different flight missions and environmental conditions, making the path planning method highly flexible and adaptable, and able to meet the path planning needs of UAVs in different scenarios.

[0081] Preferably, a Gaussian process regression model is trained using a global initial path, and a global path is obtained based on the Gaussian process regression model. The specific steps are as follows:

[0082] The initial path arc length is obtained based on the global initial path, specifically as follows:

[0083] The nodes that calculate the initial path arc length based on the nodes of the global initial path are then connected in index order to obtain the initial path arc length, expressed by the formula:

[0084] s i =s i-1 +||P i -P i-1 ||;

[0085] s1 = 0;

[0086] In the formula, s i P represents the i-th node of the initial path arc length. i Let node P represent the i-th node of the global initial path. i Records the three-dimensional coordinate information used for this point, where |||| represents the norm;

[0087] Among them, the node s with initial path arc lengthi With node P of the global initial path i correspond;

[0088] The initial path arc length is used as the training set to train a Gaussian process regression model, and the parameter σ is optimized by maximizing the marginal likelihood method. f σ n And l;

[0089] The covariance function of the Gaussian process regression model is a radial basis function constructed based on the initial path arc length, expressed by the formula:

[0090]

[0091] In the formula, k represents the covariance function, and s j The j-th node represents the initial path arc length, l represents the length scale, and σ f σ represents the signal variance. n σ represents the noise variance. ij This represents a binary variable; if i = j, the value is 1, otherwise it is 0.

[0092] The initial path arc length is predicted using a trained Gaussian process regression model to obtain the smooth path arc length;

[0093] Obtain the curvature and curvature variance of each node in the arc length of the smooth path;

[0094] Set curvature threshold and curvature variance threshold, and delete the nodes in the global initial path corresponding to the smooth path arc length whose curvature is less than the curvature threshold or whose curvature variance is less than the curvature variance threshold;

[0095] The remaining nodes of the initial global path are connected in index order to form a global path. Obstacle collision detection is performed on the straight segments between nodes of the global path (if a collision occurs, a new node is added near the collision point). The final result is a smooth curve that reduces redundant nodes and does not intersect with obstacles.

[0096] Preferably, the gravitational repulsion strategy includes a gravitational strategy and a repulsion strategy;

[0097] The gravity strategy guides the UAV to fly along a global path by calculating gravity.

[0098] The repulsion strategy guides the drone to avoid no-fly zones by calculating the repulsion force.

[0099] Preferably, the method for calculating the gravitational force is as follows:

[0100] Calculate the Euclidean distance between the current point of the drone and the target point;

[0101] Set the saturation speed control coefficient and the radius of influence of no-fly zones on drones;

[0102] Gravity is obtained based on the Euclidean distance between the current point and the target point of the UAV, the saturation velocity control coefficient, and the influence radius of the no-fly zone on the UAV. The gravity formula is expressed as:

[0103]

[0104] In the formula, F att k represents gravity. att d(p) represents the gravitational potential constant, θ represents the saturation velocity control parameter, and d(p) represents the gravitational potential constant. now ,p goal ) represents calculating the current point p of the drone. now With target point p goal The Euclidean distance, d0 represents the radius of influence of the no-fly zone on the drone.

[0105] Preferably, the method for calculating the repulsive force is as follows:

[0106] Calculate the Euclidean distance between the drone's current point and the no-fly zone;

[0107] Set the repulsive force variation control coefficient;

[0108] The repulsive force is obtained based on the influence radius of the no-fly zone on the drone, the repulsive force change control coefficient, the Euclidean distance between the drone's current point and the target point, and the Euclidean distance between the drone's current point and the no-fly zone. The repulsive force field function is expressed by the formula:

[0109]

[0110] The repulsive force experienced by the drone is the negative gradient of the repulsive field function, expressed by the formula:

[0111]

[0112] In the formula, U rep Represents the repulsive field function. F represents the gradient. rep k represents repulsive force. rep d(p) represents the repulsive potential field constant. now (p0) represents the current point p of the drone. now Euclidean distance from the no-fly zone p0, represents the repulsive force change control coefficient, and m represents the amplitude adjustment coefficient.

[0113] When the drone approaches the target point, reduce the gravitational force F. att To avoid oscillations, the repulsive force F is applied. repIt also gradually decreases to avoid the problem of target points being unreachable due to nearby obstacles; when the drone moves away from the target point, the gravitational force F... att Enlarge it to guide the drone to move quickly toward the target point.

[0114] Example 2:

[0115] This embodiment provides an electronic device that stores a computer program, which, when executed by a processor, implements a UAV path planning method based on dynamic weights as described in any embodiment of the present invention.

[0116] Example 3:

[0117] This embodiment provides a computer-readable storage medium for storing one or more programs, which, when executed by one or more processors, cause the one or more processors to implement the UAV path planning method based on dynamic weights as described in any embodiment of the present invention.

[0118] In this application embodiment, "at least one" refers to one or more, and "more than one" refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent the existence of A alone, A and B simultaneously, or B alone. A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one of the following" and similar expressions refer to any combination of these items, including any combination of singular or plural items. For example, at least one of a, b, and c can represent: a, b, c, a and b, a and c, b and c, or a and b and c, where a, b, and c can be single or multiple.

[0119] Those skilled in the art will recognize that the units and algorithm steps described in the embodiments disclosed herein can be implemented using electronic hardware, computer software, or a combination of electronic hardware and software. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0120] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0121] In the several embodiments provided in this application, any function, if implemented as a software functional unit and sold or used as an independent product, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0122] The above description is merely an embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A method for UAV path planning based on dynamic weights, characterized in that, Includes the following steps: A 3D environmental model is obtained by performing 3D modeling on the flight area of ​​the UAV; Determine obstacle points in the 3D environmental model, as well as the starting and target points of the UAV, and delineate no-fly zones based on the obstacle points; An improved A* algorithm is used to obtain the global initial path from the starting point to the target point; wherein, the improvement of the A* algorithm includes: obtaining dynamic weights based on the starting point, the target point and the no-fly zone, and using the dynamic weights to optimize the cost function of the A* algorithm; A Gaussian process regression model is trained using a global initial path, and a global path is obtained based on the Gaussian process regression model. Develop a gravity-repulsion strategy to guide drones to avoid no-fly zones; The drone starts from the starting point and executes the gravity repulsion strategy and global path to fly to the target point.

2. The UAV path planning method based on dynamic weights according to claim 1, characterized in that, A 3D rasterized map method was used to create a 3D model of the UAV's flight area.

3. The UAV path planning method based on dynamic weights according to claim 1, characterized in that, Improvements to the A* algorithm include: Discard the search area in the opposite direction between the drone's current point and the target point, and expand the search area in the same direction between the drone's current point and the target point.

4. The UAV path planning method based on dynamic weights according to claim 1, characterized in that, Based on the starting point, target point, and no-fly zone, dynamic weights are obtained. These dynamic weights are then used to optimize the cost function of the A* algorithm. The specific steps are as follows: Obtain the proportion of the no-fly zone in the 3D environmental model; Calculate the distance from the starting point to the target point and the distance from the current point of the drone to the target point; Dynamic weights are obtained based on the aforementioned ratio, the distance from the starting point to the target point, and the distance from the current point of the UAV to the target point. During the cost function calculation process of the A* algorithm, the dynamic weights are assigned to the estimated cost from the current point to the target point.

5. The UAV path planning method based on dynamic weights according to claim 1, characterized in that, The Gaussian process regression model is trained using a global initial path, and the global path is obtained based on the Gaussian process regression model. The specific steps are as follows: The initial path arc length is obtained based on the global initial path, and the node s with the initial path arc length is given. i With node P of the global initial path i correspond; The initial path arc length is used as the training set to train the Gaussian process regression model. The covariance function of the Gaussian process regression model is a radial basis function constructed based on the initial path arc length. The initial path arc length is predicted using a trained Gaussian process regression model to obtain the smooth path arc length; Obtain the curvature and curvature variance of each node in the arc length of the smooth path; Set curvature threshold and curvature variance threshold, and delete the nodes in the global initial path corresponding to the smooth path arc length whose curvature is less than the curvature threshold or whose curvature variance is less than the curvature variance threshold; Construct the remaining nodes of the global initial path into a global path.

6. The UAV path planning method based on dynamic weights according to claim 1, characterized in that, The gravitational repulsion strategy includes both gravitational and repulsion strategies. The gravity strategy guides the UAV to fly along a global path by calculating gravity. The repulsion strategy guides the drone to avoid no-fly zones by calculating the repulsion force.

7. The UAV path planning method based on dynamic weights according to claim 6, characterized in that, The method for calculating the gravity is as follows: Calculate the Euclidean distance between the current point of the drone and the target point; Set the saturation speed control coefficient and the radius of influence of no-fly zones on drones; Gravity is obtained based on the Euclidean distance between the current point and the target point of the UAV, the saturation velocity control coefficient, and the influence radius of the no-fly zone on the UAV.

8. The UAV path planning method based on dynamic weights according to claim 7, characterized in that, The method for calculating the repulsive force is as follows: Calculate the Euclidean distance between the drone's current point and the no-fly zone; Set the repulsive force variation control coefficient; The repulsive force is obtained based on the influence radius of the no-fly zone on the drone, the repulsive force change control coefficient, the Euclidean distance between the drone's current point and the target point, and the Euclidean distance between the drone's current point and the no-fly zone.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the UAV path planning method based on dynamic weights as described in any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by the processor, the program implements the UAV path planning method based on dynamic weights as described in any one of claims 1 to 8.

Images