A UAV path planning method and system based on artificial intelligence

Abstract

The present invention relates to the field of drone technology, and in particular to a drone path planning method and system based on artificial intelligence. The method comprises: obtaining multiple obstacles in front of the drone during its travel; calculating the behavior complexity of each obstacle; the behavior complexity is positively correlated with the shape complexity and the direction change rate; updating the repulsive field function in the artificial potential field, and generating the path of the current drone through the updated artificial potential field; that is, the solution of the present invention can avoid obstacles in time when the drone is performing path planning, thereby improving the safety of the drone.

Description

Technical Field

[0001] The present invention relates to the field of unmanned aerial vehicle technology, and more specifically, to an artificial intelligence-based unmanned aerial vehicle path planning method and system. Background Art

[0002] With the rapid development of drone technology, the application scope of drones is constantly expanding, such as agricultural monitoring, logistics distribution, environmental monitoring, disaster relief, etc. When planning the path of drones, traditional path planning algorithms include Algorithm (A-star Algorithm), Dijkstra algorithm and artificial potential field method and bionic ant colony algorithm. Among them, the artificial potential field method is a simple and effective method in path planning algorithm.

[0003] In the application of UAV path planning, the basic idea of ​​the artificial potential field method is to construct an artificial potential field in the working environment of the UAV. The potential field includes repulsive poles and attractive poles. The areas and obstacles that the UAV is not allowed to enter are defined as repulsive poles, and the target and the area where the UAV is recommended to enter are defined as gravitational poles. The UAV in the potential field is affected by the combined effect of the gravitational field of its target position and the repulsive field around the obstacles, and moves towards the target.

[0004] As a common path planning algorithm, the artificial potential field method is applied to vehicle obstacle avoidance planning and UAV path planning due to its low algorithm complexity and simple calculation. However, the existing algorithm structure has obvious shortcomings. It is very easy to fall into the local minimum, resulting in planning failure, and it is unable to handle complex obstacles and iterative optimization. These shortcomings make the artificial potential field method only applicable to local path planning and not suitable for global path planning in complex obstacle environments.

[0005] Although the above methods are widely used, they still have problems such as local optimal solutions, insufficient obstacle handling, and inability to respond to collisions with dynamic obstacles in a timely manner in complex and dynamic environments.

[0006] For example, the patent application document with publication number CN118838362A discloses a multi-UAV artificial potential field design method and a multi-UAV collaborative collision avoidance method, which optimizes the speed-related terms of the repulsion function and the position-related terms of the repulsion function, and synthesizes the two terms into an artificial potential field repulsion function, thereby realizing multi-UAV formation coordination and collision avoidance control.

[0007] Although the above scheme calculates the position-related and speed-related terms of the repulsion function between each obstacle and the current UAV, and then solves the problem of local optimality and unreachable target of the UAV by increasing the repulsion; however, since the influence range of the obstacle is fixed, when the obstacles are compactly distributed, their response time is short and they cannot be avoided by applying force in advance. Therefore, there is a problem that the UAV cannot avoid obstacles in time and collides. Summary of the invention

[0008] The purpose of the present invention is to propose an artificial intelligence-based drone path planning method and system to solve the problem that the drone path planning in the prior art cannot avoid obstacles in time and has safety hazards; to this end, the present invention provides solutions in the following two aspects.

[0009] In a first aspect, the present invention provides an artificial intelligence-based unmanned aerial vehicle path planning method, comprising:

[0010] Obtain multiple obstacles in front of the drone during its flight;

[0011] Calculate the behavior complexity of each obstacle; the behavior complexity is positively correlated with the shape complexity and the direction change rate;

[0012] Update the repulsive field function in the artificial potential field, and generate the path of the current UAV through the updated artificial potential field; the repulsive field function is:

[0013] ; represents the value of the repulsive field function of obstacle k, represents the repulsion scale factor, Indicates the position of the drone. represents the position of obstacle k, is a vector whose magnitude is the distance between the current position of the drone and the position of obstacle k, and whose direction is from obstacle k to the current position of the drone. is a constant, represents the behavioral complexity of obstacle k, is the distance from the current position of the drone to the end point.

[0014] In the above scheme, by performing complexity analysis on the obstacles in front of the drone, the complexity of the obstacle area itself can be obtained, and then the complexity itself can be used to improve the repulsion range (safety range) in the repulsion function in the artificial potential field method, so that different obstacles have an adaptive repulsion range, so that when planning the drone path, collisions can be avoided in time to ensure the safety of the drone.

[0015] Optionally, the behavior complexity for: ;

[0016] in, represents the moving speed of obstacle k, is the shape complexity of obstacle k, Indicates the maximum moving speed of multiple obstacles in front of the flight space. represents the rate of change of the direction of obstacle k.

[0017] In the above scheme, the behavior complexity of the obstacle is determined by combining the shape complexity and the moving speed, which can accurately characterize the situation of the corresponding obstacle.

[0018] Optionally, the behavior complexity for: ;

[0019] in, Indicates the maximum length of multiple obstacles ahead in the flight space. , , are the length, shape complexity, and moving speed of obstacle k, respectively. Indicates the maximum moving speed of multiple obstacles in front of the flight space. represents the rate of change of the direction of obstacle k.

[0020] In the above scheme, the length of the obstacle is introduced on the basis of shape complexity and moving speed, which can highlight the relative situation of the shapes of multiple obstacles in front.

[0021] Optionally, the shape complexity is: ;

[0022] in, represents the shape complexity of obstacle k, represents the aspect ratio of obstacle k, , They are the number of corner points representing obstacle k and the number of corner points where the sharp internal angle is less than 45°.

[0023] In the above scheme, the structure of the obstacle is characterized by obtaining the size of the obstacle and the number of corner points on the obstacle.

[0024] Optionally, the direction change rate for: ;

[0025] in, represents the number of times the obstacle k changes its motion direction during the observation time, T represents the observation time, It represents the sum of the angles of the transformation direction of obstacle k during the observation time.

[0026] In the above scheme, the regularity of the obstacle's movement can be represented by the rate of change of the obstacle's direction.

[0027] Optionally, the gravitational field function of the artificial potential field is: ;

[0028] in, represents the value of the gravitational field function, Represents a vector whose size is the Euclidean distance from the current position of the drone to the end position, and whose direction is from the current position of the drone to the end position. is the proportional gain coefficient, Indicates the position of the drone. is the end position.

[0029] Optionally, the process of generating the path of the current UAV through the updated artificial potential field is:

[0030] The artificial potential field function is obtained by adding the updated repulsive field function to the original gravitational field function;

[0031] The gradient descent algorithm is used on the artificial potential field function to find the optimal path to the drone.

[0032] In a second aspect, an artificial intelligence-based unmanned aerial vehicle path planning system comprises:

[0033] processor;

[0034] A memory stores a computer instruction for drone path planning based on artificial intelligence. When the computer instruction is executed by the processor, the system executes the above-mentioned drone path planning method based on artificial intelligence.

[0035] The beneficial effects of the present invention are:

[0036] The solution of the present invention can calculate the behavioral complexity of obstacles according to the static characteristics and dynamic characteristics of obstacles in the environment in front of the drone, and obtain the influence range of the repulsive field of different obstacles according to the behavioral complexity of the obstacles and the difference in motion vectors, determine the gravitational field function of the front environment and the repulsive field function of each obstacle, realize the path planning of the drone in an environment with mixed dynamic and static obstacles, and avoid collisions during the driving process of the drone. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] By reading the following detailed description with reference to the accompanying drawings, the above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:

[0038] Figure 1 A flowchart of the steps of a drone path planning method based on artificial intelligence in this embodiment is schematically shown;

[0039] Figure 2 The structural block diagram of an artificial intelligence-based UAV path planning system in this embodiment is schematically shown. DETAILED DESCRIPTION

[0040] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0041] The present invention is aimed at the path planning of a UAV, that is, when the UAV is driving automatically, there are static obstacles and dynamic obstacles in front.

[0042] Specifically, Figure 1 As shown, a drone path planning method based on artificial intelligence in this embodiment includes the following steps:

[0043] Step S1, obtaining real-time data of multiple obstacles in front of the drone during its travel and the obstacles. In this embodiment, the real-time data of obstacles is obtained by a laser radar arranged on the drone, wherein the real-time data includes the size (length and width), real-time position, moving direction, moving speed, moving acceleration, etc. of the obstacles.

[0044] The above-mentioned obstacle sizes are collected because the sizes of different obstacles affect the relative safety distance of the drone. For example, trees, other drones, and large birds are large in size. When avoiding obstacles, if the relative safety distance is small, it is difficult for the drone to avoid them.

[0045] The above-mentioned collection of moving direction, moving speed and moving acceleration is due to the fact that the movement states of obstacles during obstacle avoidance are diverse and different. Failures with more frequent changes in moving direction may lead to higher safety risks, and faster moving speeds may cause the drone to fail to respond in time and cause a collision.

[0046] Step S2, calculating the behavior complexity of each obstacle.

[0047] In this embodiment, considering the obstruction of various obstacles in the path planning of the drone, it is necessary to first calculate the behavior complexity of each obstacle. The specific process is as follows:

[0048] Step S21, calculating the shape complexity of each obstacle.

[0049] The shape complexity is: ;

[0050] in, represents the shape complexity of obstacle k, represents the aspect ratio of obstacle k, , They are the number of corner points representing obstacle k and the number of corner points where the sharp internal angle is less than 45°.

[0051] The above number of corner points can be used to collect the image of the obstacle in front and obtain it through Harris corner point detection.

[0052] The sharp inner angle at the location of the above corner point is obtained by obtaining the size and direction of the inner angle of the corner point. Specifically, the line between the left adjacent point of the corner point and the corner point is in the direction of the corner point pointing to the left adjacent point, and the line between the corner point and the right adjacent point and the corner point is in the direction of the corner point pointing to the right adjacent point. If the angle between the two lines is less than 45°, it is a sharp inner angle.

[0053] It should be noted that obstacles of different shapes have different collision risks. For example, geese have a large transverse diameter, curved wings, and a large length-to-width ratio; sparrows have a small transverse and vertical ratio, short wings, and a relatively simple shape; balloons are approximately spherical in volume, have smooth edges and no edges, and are relatively simple in shape. Therefore, different shape characteristics may cause the length of some directions of obstacles to exceed the safety distance when avoiding obstacles based on a fixed safety distance, resulting in collisions. Therefore, the shape complexity of the obstacle is calculated based on the static shape characteristics of the obstacle.

[0054] Step S22, calculating the direction change rate of each obstacle.

[0055] The rate of change of direction is: ;

[0056] in, represents the rate of change of the direction of obstacle k, represents the number of times the obstacle k changes its motion direction during the observation time, T represents the observation time, It represents the sum of the angles of the transformation direction of obstacle k during the observation time.

[0057] Among the above The larger it is, the more frequently the direction of the obstacle k changes during the observation time, and the greater the degree of change.

[0058] Due to the diversity of the movement direction and speed of obstacles (the direction change rates of different obstacles are also significantly different, such as balloons, drones, birds, etc., the factors affecting the direction change rates of the three are completely different), which leads to the phenomenon of misjudgment of the obstacle avoidance route and collision when avoiding obstacles based on a fixed safety distance. Therefore, the direction change rate of the obstacle is calculated based on the number of times the obstacle changes direction within a period of time to characterize the movement characteristics of the obstacle ahead.

[0059] Step S23, obtaining the behavior complexity of the corresponding obstacle according to the shape complexity and direction change rate of the obstacle.

[0060] In one embodiment, the behavior complexity for:

[0061] ;

[0062] in, represents the moving speed of obstacle k, is the shape complexity of obstacle k, Indicates the maximum moving speed of multiple obstacles in front of the flight space. represents the rate of change of the direction of obstacle k.

[0063] In another embodiment, the behavior complexity You can also:

[0064] ;

[0065] in, Indicates the maximum length of multiple obstacles in the moving space, , , are the length, shape complexity, and moving speed of obstacle k, respectively. Indicates the maximum moving speed of multiple obstacles in front of the flight space. represents the rate of change of the direction of obstacle k.

[0066] Above is the correction term for the behavioral complexity of obstacle k. The longer the obstacle k is, the more it affects safety. Since there may be multiple obstacles ahead, it is necessary to determine the correction term for each obstacle separately. Therefore, The velocity term and the direction change rate term are corrected upward to obtain the complexity of the obstacle.

[0067] Step S3, updating the repulsive field function in the artificial potential field, and generating the flight path of the current UAV through the updated artificial potential field.

[0068] In this embodiment, an artificial potential field is constructed based on the starting point, the end point, the position of the drone, the position of the obstacle, the motion vector difference and the behavior complexity of the obstacle to perform the path planning of the drone, as follows:

[0069] First, the repulsive field function in the artificial potential field is constructed by using the behavioral complexity of each obstacle. Specifically, the repulsive field function and the attractive field function are constructed through the artificial potential field method; the value of the repulsive field function is positively correlated with the behavioral complexity.

[0070] Among them, Artificial Potential Field (APF) is a common path planning and obstacle avoidance method, which achieves path planning and obstacle avoidance by simulating the interaction between objects.

[0071] In this embodiment, the repulsive field function is obtained according to the obtained behavior complexity of each obstacle, specifically:

[0072] ;

[0073] represents the value of the repulsive field function of obstacle k, represents the repulsion scale factor, Indicates the position of the drone. represents the position of obstacle k, is a vector whose magnitude is the distance between the current position of the drone and the position of obstacle k, and whose direction is the direction from the current position of the drone to obstacle k. is a constant, which represents the influence coefficient of the repulsive force of obstacle k on the UAV; represents the behavioral complexity of obstacle k, is the distance from the current position of the drone to the end point.

[0074] in, It represents the influence range of the corrected obstacle k's repulsive force on the UAV. The larger it is, the greater the range of influence of the obstacle k on the drone's repulsive force. Indicates the distance between the drone and the end point. When the relative distance between the drone and the obstacle k exceeds , it is determined that the obstacle has no effect on the drone, and its repulsive field function is 0.

[0075] The value of the above proportional coefficient is usually A feasible positive proportionality coefficient is 50. The value of is usually , a feasible The value is 15 meters.

[0076] The above-mentioned behavior complexity is used to measure the behavior of obstacles. That is, due to the different distribution and behavior complexity of dynamic obstacles and static obstacles in the environment, local optimality or direct collision with moving obstacles after obstacle avoidance may occur when using traditional static artificial potential fields for obstacle avoidance in complex flight spaces. Therefore, the behavior complexity is used in the embodiment to adjust the safety range in the repulsive field function.

[0077] Among them, the gravitational field function is: ;

[0078] in, represents the gravitational field function, Represents a vector whose size is the Euclidean distance from the current position of the drone to the end position, and whose direction is from the current position of the drone to the end position. is the proportional gain coefficient, Indicates the position of the drone. is the end position.

[0079] The above gravitational potential field function is mainly related to the distance between the current position of the drone and the end position. The farther the distance, the greater the gravitational value, and the closer the distance, the smaller the gravitational value.

[0080] Secondly, after determining the repulsive field function and the gravitational field function, the total potential field function of the artificial potential field is calculated. When the value of the total potential field function is the largest, the optimal path of the UAV can be obtained.

[0081] Among them, since the specific implementation process of the artificial potential field method is an existing technology, this embodiment only introduces the improved repulsive field function, and the others are not repeated.

[0082] When constructing the repulsive field function and the gravitational field function, the complexity of the behavior of the obstacle ahead is taken into consideration. In order to improve the repulsive field function, the repulsive range of the drone can be obtained, that is, the repulsive range of different obstacles (determined by the safety range) is different. According to the adaptive repulsive range of different obstacles, not only the accuracy of path planning can be improved, but also the problem of drone collision can be avoided in time. Compared with the traditional artificial potential field, which will have the problem of local optimality or unable to respond in time to avoid dynamic obstacles, the solution of this embodiment can avoid the phenomenon of collision due to untimely response.

[0083] The solution of the present invention can quickly obtain the safety range between the current drone and the obstacles ahead, and avoid them in time when planning the route, thus avoiding the risk of collision and improving safety.

[0084] The present invention also provides a UAV path planning system based on artificial intelligence. Figure 2 As shown, the system includes a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, an artificial intelligence-based drone path planning method according to the present invention is implemented.

[0085] The system also includes other components familiar to those skilled in the art, such as a communication bus and a communication interface, whose configuration and functions are known in the art and thus will not be described in detail here.

[0086] In the present invention, the aforementioned memory may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, apparatus or device. For example, a computer-readable storage medium may be any appropriate magnetic storage medium or magneto-optical storage medium, such as a resistive random access memory RRAM (Resistive Random Access Memory), a dynamic random access memory DRAM (Dynamic Random Access Memory), a static random access memory SRAM (Static Random-Access Memory), an enhanced dynamic random access memory EDRAM (Enhanced Dynamic Random Access Memory), a high-bandwidth memory HBM (High-Bandwidth Memory), a hybrid memory cube HMC (Hybrid Memory Cube), etc., or any other medium that can be used to store the required information and can be accessed by an application, a module, or both. Any such computer storage medium may be part of a device or accessible or connectable to a device. Any application or module described in the present invention may be implemented using computer-readable / executable instructions that may be stored or otherwise maintained by such a computer-readable medium.

[0087] In the description of this specification, “plurality” means at least two, such as two, three or more, etc., unless otherwise clearly and specifically defined.

[0088] Although this specification has shown and described a number of embodiments of the present invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art will conceive of many modifications, changes and alternatives without departing from the ideas and spirit of the present invention. It should be understood that in the practice of the present invention, various alternatives to the embodiments of the present invention described herein may be employed.

Claims

1. A UAV path planning method based on artificial intelligence, characterized in that: include: Obtain multiple obstacles in front of the drone during its flight; Calculate the behavioral complexity of each obstacle; The complexity of the behavior for: , represents the moving speed of obstacle k, represents the shape complexity of obstacle k, Indicates the maximum moving speed of multiple obstacles in front of the flight space. represents the rate of change of the direction of obstacle k; where the shape complexity is: , represents the aspect ratio of obstacle k, , They represent the number of corner points of obstacle k and the number of corner points where the sharp internal angles are less than 45°; Update the repulsive field function in the artificial potential field, and generate the path of the current UAV through the updated artificial potential field; the repulsive field function is: ; represents the value of the repulsive field function of obstacle k, represents the repulsion scale factor, Indicates the position of the drone. represents the position of obstacle k, is a vector whose magnitude is the distance between the current position of the drone and the position of obstacle k, and whose direction is from obstacle k to the current position of the drone. represents a constant, represents the behavioral complexity of obstacle k, Indicates the distance from the current position of the drone to the end point.

2. The method for unmanned aerial vehicle path planning based on artificial intelligence according to claim 1, characterized in that: The complexity of the behavior for: ; in, Indicates the maximum length of multiple obstacles ahead in the flight space. , , They represent the length, shape complexity, and moving speed of obstacle k respectively. Indicates the maximum moving speed of multiple obstacles in front of the flight space. represents the rate of change of the direction of obstacle k.

3. The method for unmanned aerial vehicle path planning based on artificial intelligence according to claim 1, characterized in that: The rate of change of direction for: ; in, represents the number of times the obstacle k changes its motion direction during the observation time, T represents the observation time, It represents the sum of the angles of the transformation direction of obstacle k during the observation time.

4. The method for unmanned aerial vehicle path planning based on artificial intelligence according to claim 1, characterized in that: The gravitational field function of the artificial potential field is: ; in, represents the value of the gravitational field function, Represents a vector whose size is the Euclidean distance from the current position of the drone to the end position, and whose direction is from the current position of the drone to the end position. is the proportional gain coefficient, Indicates the position of the drone. is the end position.

5. The method for unmanned aerial vehicle path planning based on artificial intelligence according to claim 1, characterized in that: The process of generating the path of the current UAV through the updated artificial potential field is as follows: The artificial potential field function is obtained by adding the updated repulsive field function to the original gravitational field function; The gradient descent algorithm is used on the artificial potential field function to find the optimal path to the drone.

6. An artificial intelligence-based drone path planning system, characterized in that: include: processor; A memory storing a computer instruction for drone path planning based on artificial intelligence. When the computer instruction is executed by the processor, the system executes a drone path planning method based on artificial intelligence according to any one of claims 1 to 5.

Images