CN-121680472-B - Unmanned aerial vehicle control method and electronic equipment

CN121680472BCN 121680472 BCN121680472 BCN 121680472BCN-121680472-B

Abstract

The application provides an unmanned aerial vehicle control method and electronic equipment, wherein the method comprises the steps of obtaining first position information of a first unmanned aerial vehicle and second position information of each second unmanned aerial vehicle, determining first neighborhood information and a first local map according to the first position information and the second position information of each second unmanned aerial vehicle, constructing a target graph structure corresponding to the first unmanned aerial vehicle according to the first position information, each second position information, the first neighborhood information and the first local map, determining node characteristic information and side characteristic information of the target graph structure, inputting the target graph structure corresponding to the first unmanned aerial vehicle, the node characteristic information and the side characteristic information of the target graph structure into a pre-trained flight decision model, and obtaining flight control information of the first unmanned aerial vehicle, and controlling the first unmanned aerial vehicle to fly based on the flight control information of the first unmanned aerial vehicle. The application can realize the stable flight of the first unmanned aerial vehicle under the conditions of intensive meeting of multiple unmanned aerial vehicles, complex obstacle and the like.

Inventors

LIU XING
GAO BO

Assignees

深圳信息职业技术大学

Dates

Publication Date: 20260505
Application Date: 20260209

Claims (8)

1. The unmanned aerial vehicle control method is characterized by being applied to a first unmanned aerial vehicle in a multi-unmanned aerial vehicle scene, wherein the multi-unmanned aerial vehicle scene comprises the first unmanned aerial vehicle and a plurality of second unmanned aerial vehicles, and the method comprises the following steps: Acquiring first position information of a first unmanned aerial vehicle and second position information of each second unmanned aerial vehicle; determining first neighborhood information and a first local map of the first unmanned aerial vehicle according to the first position information of the first unmanned aerial vehicle and the second position information of each second unmanned aerial vehicle; Constructing a target graph structure corresponding to the first unmanned aerial vehicle according to the first position information, the second position information, the first neighborhood information and the first local map, determining node characteristic information and side characteristic information of the target graph structure, wherein the target graph structure comprises a plurality of nodes and sides among the nodes, and each node represents the first unmanned aerial vehicle or a second unmanned aerial vehicle positioned in the neighborhood of the first unmanned aerial vehicle; Inputting a target graph structure corresponding to the first unmanned aerial vehicle, node characteristic information and side characteristic information of the target graph structure into a pre-trained flight decision model, and predicting the flight decision model to obtain flight control information of the first unmanned aerial vehicle, wherein the flight decision model is a model obtained based on graph attention strategy network training; Controlling the first unmanned aerial vehicle to fly based on the flight control information of the first unmanned aerial vehicle; The determining node characteristic information and side characteristic information of the target graph structure comprises the steps of carrying out centering processing on a current local map of a current unmanned aerial vehicle to obtain a current local map after centering processing, wherein the current unmanned aerial vehicle is an unmanned aerial vehicle corresponding to any node in the target graph structure; the method comprises the steps of inputting a current local map after centering into a map convolutional neural network obtained by training in advance, extracting and encoding map features of the current local map after centering by the map convolutional neural network to generate current local map encoding features of the current unmanned aerial vehicle, obtaining a target destination of the current unmanned aerial vehicle, determining current destination direction features according to the target destination of the current unmanned aerial vehicle and current position information of the current unmanned aerial vehicle, determining current movement direction features of the current unmanned aerial vehicle at the previous moment, determining current neighborhood features of the current unmanned aerial vehicle according to the current position information of the current unmanned aerial vehicle and current neighborhood information of the current unmanned aerial vehicle, and taking the current local map encoding features, the current destination direction features, the current movement direction features and the current neighborhood features of the current unmanned aerial vehicle as node feature information corresponding to the current unmanned aerial vehicle; Traversing the nodes corresponding to the second unmanned aerial vehicles in the target graph structure, calculating the direction vector and the distance between the first unmanned aerial vehicle and the second unmanned aerial vehicle corresponding to the current node according to the first position information and the second position information of the second unmanned aerial vehicle corresponding to the current node, and determining the side characteristic information between the current node and the nodes corresponding to the first unmanned aerial vehicle according to the calculated direction vector and distance.
2. The method of claim 1, wherein the determining the first neighborhood information and the first local map of the first drone according to the first location information of the first drone and the second location information of each of the second drones includes: determining first neighborhood information of the first unmanned aerial vehicle according to the first position information of the first unmanned aerial vehicle, the second position information of each second unmanned aerial vehicle and a preset communication radius; And acquiring a global map of the multi-unmanned aerial vehicle scene, and determining the first local map according to the global map, the visual field information of the first unmanned aerial vehicle and the first position information of the first unmanned aerial vehicle.
3. The unmanned aerial vehicle control method of claim 1, wherein the constructing the target graph structure corresponding to the first unmanned aerial vehicle and determining node feature information and edge feature information of the target graph structure according to the first location information, each of the second location information, the first neighborhood information, and the first local map comprises: constructing a target graph structure corresponding to the first unmanned aerial vehicle according to the first neighborhood information; Determining node characteristic information of the target graph structure according to the first position information, the second position information, the first local map and the first neighborhood information; And determining side characteristic information of the target graph structure according to the first position information, the second position information and the target graph structure.
4. The unmanned aerial vehicle control method of claim 1, wherein the determining the current neighborhood feature of the current unmanned aerial vehicle according to the current location information of the current unmanned aerial vehicle and the current neighborhood information of the current unmanned aerial vehicle comprises: Determining the number of the neighborhoods of the current unmanned aerial vehicle according to the current neighborhood information of the current unmanned aerial vehicle; Determining a neighborhood distance average value corresponding to the current unmanned aerial vehicle according to the current neighborhood information of the current unmanned aerial vehicle, the current position information of the current unmanned aerial vehicle and the position information of each unmanned aerial vehicle in the multi-unmanned aerial vehicle scene; and determining the current neighborhood characteristics of the current unmanned aerial vehicle according to the neighborhood number and the neighborhood distance average value.
5. The unmanned aerial vehicle control method of claim 1, wherein the flight decision model comprises a linear transformation layer, a plurality of multi-head graph attention strategy networks, a plurality of post-processing layers and an output layer which are connected in sequence; Inputting the target graph structure corresponding to the first unmanned aerial vehicle, node characteristic information and side characteristic information of the target graph structure into a pre-trained flight decision model, and predicting by the flight decision model to obtain flight control information of the first unmanned aerial vehicle, wherein the method comprises the following steps: Inputting node characteristic information and side characteristic information of the target graph structure into the linear transformation layer to perform linear transformation to obtain linear node characteristic information and linear side characteristic information; Inputting the target graph structure, the linear node characteristic information and the linear side characteristic information into a plurality of multi-head graph attention strategy networks and a plurality of post-processing layers, performing multi-head attention processing by the multi-head graph attention strategy networks, and performing output processing by each post-processing layer to obtain characteristic vectors of each node in the target graph structure; And inputting the feature vector of each node in the target graph structure into the output layer to obtain scores corresponding to the first unmanned aerial vehicle in a plurality of flight directions, and taking each flight direction and the score of each flight direction as the flight control information.
6. The unmanned aerial vehicle control method of claim 1, wherein the flight control information comprises a plurality of flight directions and a score for each flight direction; The controlling the flight of the first unmanned aerial vehicle based on the flight control information of the first unmanned aerial vehicle includes: determining flight track constraint information corresponding to the first unmanned aerial vehicle in each flight direction and target guiding constraint information corresponding to the first unmanned aerial vehicle in each flight direction; determining performance parameters of the first unmanned aerial vehicle in all flight directions according to the scores corresponding to all flight directions, the flight track constraint information and the target guiding constraint information; determining target flight actions of the first unmanned aerial vehicle according to the performance parameters of the first unmanned aerial vehicle in all flight directions; And controlling the first unmanned aerial vehicle to fly based on the target flying action of the first unmanned aerial vehicle.
7. The method of claim 6, wherein determining the target flight actions of the first drone based on the performance parameters of the first drone in each flight direction, comprises: sequencing all flight directions according to performance parameters of the first unmanned aerial vehicle in all flight directions, and determining a plurality of candidate flight directions according to a preset quantity threshold; determining cost values of the candidate flight directions according to a preset cost map, and determining a target flight direction of the first unmanned aerial vehicle according to the cost values of the candidate flight directions; And determining the target flight action of the first unmanned aerial vehicle according to the target flight direction of the first unmanned aerial vehicle.
8. An electronic device comprising a processor and a memory, the memory storing machine-readable instructions executable by the processor, the processor executing the machine-readable instructions to perform the steps of the drone control method of any one of claims 1 to 7 when the electronic device is operating.

Description

Unmanned aerial vehicle control method and electronic equipment Technical Field The application relates to the technical field of unmanned aerial vehicle control, in particular to an unmanned aerial vehicle control method and electronic equipment. Background Along with the rapid development of unmanned aerial vehicle technology, the application of a multi-unmanned aerial vehicle system in the fields of logistics distribution, emergency rescue, agricultural plant protection, city inspection and the like is increasingly wide. In these application scenarios, multiple unmanned aerial vehicles often need to cooperatively execute tasks in the same airspace, and inevitably face the risks of approaching each other and even collision. Therefore, how to realize efficient and safe autonomous avoidance decision has become a key technical challenge for guaranteeing the reliable operation of the multi-unmanned aerial vehicle system. In the prior art, the avoidance of the multiple unmanned aerial vehicles mainly adopts two methods, namely a centralized global path planning method and a track scheduling method and a distributed local avoidance method. Specifically, the centralized global path planning and track scheduling method relies on a central controller to acquire starting and ending point information and a global environment map of all unmanned aerial vehicles, and A\is utilized、D\Or an optimization algorithm generates a collision-free global flight path and responds to environmental changes or state deviations by periodic re-planning. According to the distributed local avoidance method, each unmanned aerial vehicle calculates the course or speed adjustment quantity in real time in each control period by combining local strategies such as a manual potential field method, a speed obstacle method (VO), a dynamic window method (DWA) and the like according to local environment information sensed by a sensor of the unmanned aerial vehicle and the relative state of the adjacent unmanned aerial vehicle, so that the obstacle and other unmanned aerial vehicles can be avoided in real time. However, in the prior art, the methods lack an explicit and structured modeling mechanism for dynamic interaction relation among multiple unmanned aerial vehicles, so that environmental adaptability and system expandability are limited, and meanwhile, the problem that instantaneity and decision stability are difficult to be achieved is also existed. Disclosure of Invention The application aims to provide an unmanned aerial vehicle control method and electronic equipment aiming at the defects in the prior art, so as to solve the problems that the environmental adaptability and the system expandability in the prior art are limited, and the real-time performance and the decision stability are difficult to be simultaneously realized. In order to achieve the above purpose, the technical scheme adopted by the embodiment of the application is as follows: In a first aspect, an embodiment of the present application provides a method for controlling an unmanned aerial vehicle, which is applied to a first unmanned aerial vehicle in a multi-unmanned aerial vehicle scene, where the multi-unmanned aerial vehicle scene includes a first unmanned aerial vehicle and a plurality of second unmanned aerial vehicles, and the method includes: Acquiring first position information of a first unmanned aerial vehicle and second position information of each second unmanned aerial vehicle; determining first neighborhood information and a first local map of the first unmanned aerial vehicle according to the first position information of the first unmanned aerial vehicle and the second position information of each second unmanned aerial vehicle; Constructing a target graph structure corresponding to the first unmanned aerial vehicle according to the first position information, the second position information, the first neighborhood information and the first local map, determining node characteristic information and side characteristic information of the target graph structure, wherein the target graph structure comprises a plurality of nodes and sides among the nodes, and each node represents the first unmanned aerial vehicle or a second unmanned aerial vehicle positioned in the neighborhood of the first unmanned aerial vehicle; Inputting a target graph structure corresponding to the first unmanned aerial vehicle, node characteristic information and side characteristic information of the target graph structure into a pre-trained flight decision model, and predicting by the flight decision model to obtain flight control information of the first unmanned aerial vehicle; and controlling the first unmanned aerial vehicle to fly based on the flight control information of the first unmanned aerial vehicle. In one possible implementation manner, the determining, according to the first location information of the first unmanned aerial vehicle and the second location information