CN-121995947-A - Unmanned aerial vehicle control method and device and electronic equipment

CN121995947ACN 121995947 ACN121995947 ACN 121995947ACN-121995947-A

Abstract

The application relates to the technical field of unmanned aerial vehicles, and provides an unmanned aerial vehicle control method, an unmanned aerial vehicle control device and electronic equipment. The method comprises the steps of obtaining visual perception information and flight state parameters of the unmanned aerial vehicle, obtaining flight action parameters of the unmanned aerial vehicle at each decision moment through a trained reinforcement learning model according to the visual perception information and the flight state parameters, performing flight control evaluation on a target object according to the visual perception information to obtain flight control credibility of the target object, performing second-order smooth constraint on the flight action parameters under the condition that the flight control credibility meets preset credibility, and determining flight control instructions for controlling the unmanned aerial vehicle according to the flight action parameters after the second-order smooth constraint. The control method and the control device can improve the control precision of the unmanned aerial vehicle in a complex environment scene and improve the flight stability of the unmanned aerial vehicle.

Inventors

Yi Nengmin
WANG CHAO
LU JIA

Assignees

四川傲势科技有限公司

Dates

Publication Date: 20260508
Application Date: 20260408

Claims (10)

1. A method of unmanned aerial vehicle control, comprising: acquiring visual perception information and flight state parameters of the unmanned aerial vehicle; Obtaining flight action parameters of the unmanned aerial vehicle at each decision moment through a trained reinforcement learning model according to the visual perception information and the flight state parameters, wherein a reward function of the reinforcement learning model is determined by the center alignment degree, the tracking stability, the safety distance constraint, the observation scale constraint and the flight stability; Performing flight control evaluation on a target object according to the visual perception information to obtain the flight control reliability of the target object; And under the condition that the flight control reliability meets the preset reliability, performing second-order smoothness constraint on the flight action parameters, and determining a flight control instruction for controlling the unmanned aerial vehicle according to the flight action parameters after the second-order smoothness constraint.
2. The method according to claim 1, wherein performing flight control evaluation on the target object according to the visual perception information to obtain the flight control reliability of the target object comprises: According to the visual perception information, determining a target track stability index and a target prediction deviation of the target object in a first preset time window, and determining a target detection confidence mean value of the target object in the first preset time window and an aspect ratio change rate of a target boundary frame; and performing flight control evaluation on the target object according to the target track stability index, the target prediction deviation, the target detection confidence coefficient mean value and the aspect ratio change rate of the target boundary box to obtain the flight control reliability of the target object.
3. The method according to claim 2, wherein determining the target trajectory stability index and the target prediction bias of the target object within a first preset time window according to the visual perception information comprises: according to the visual perception information, determining the overlapping degree of target boundary frames in adjacent image frames in a first preset time window, and determining the target track stability index according to the overlapping degree; according to the visual perception information, determining a historical track of the target object in a first preset time window, and determining an actual position of the target object in a current image frame; And predicting a target prediction position of the target object in the current image frame according to the historical track, and determining the target prediction deviation according to the target prediction position and the actual position.
4. The method of claim 1, wherein the visual perception information comprises a target relative position and a target detection confidence, and wherein the flight status parameters comprise a flight speed, a flight attitude angle and a flight altitude; the method for obtaining the flight action parameters of the unmanned aerial vehicle at each decision moment through the trained reinforcement learning model according to the visual perception information and the flight state parameters comprises the following steps: constructing a state space vector for reinforcement learning according to the target relative position, the target detection confidence, the flying speed, the flying attitude angle and the flying height; and obtaining flight action parameters of the unmanned aerial vehicle at each decision moment through a trained reinforcement learning model according to the state space vector.
5. The method according to claim 1, wherein the obtaining, according to the visual perception information and the flight state parameters, the flight action parameters of the unmanned aerial vehicle at each decision time by using the trained reinforcement learning model further comprises: Determining a sample object according to sample perception information obtained by the unmanned aerial vehicle in the flight process, determining the center alignment degree, tracking stability, observation scale constraint and safety distance constraint of the unmanned aerial vehicle in the flight process according to the sample object, and determining the flight stability of the unmanned aerial vehicle in the flight process; And determining the rewarding function according to the center alignment degree, the tracking stability, the safety distance constraint, the observation scale constraint and the flight stability.
6. The method of claim 5, wherein determining center alignment, tracking stability, viewing scale constraints, safety distance constraints of the unmanned aerial vehicle during flight from the sample object comprises: Determining the offset of the sample object and the center of the sample image according to the center pixel coordinate of the sample object, the center coordinate of the sample image and the height and width of the sample image, carrying out normalization processing on the offset, and determining the center alignment degree according to the offset after normalization processing and preset adjustment parameters; determining the position change of a sample object in an adjacent image frame in a second preset time window, and determining the tracking stability according to the position change; Acquiring the actual distance between the unmanned aerial vehicle and the sample object, and determining the safety distance constraint according to the actual distance, a preset minimum distance and a preset maximum distance by using a logarithmic barrier function; and acquiring the actual occupied area of the sample object in the sample image picture, and determining the observation scale constraint according to the actual occupied area and the preset occupied area.
7. The method of claim 5, wherein the determining the flight stability of the drone during the flight comprises: Acquiring the current flight control action and the flight control action of the unmanned aerial vehicle at the last moment; and determining the flight stability according to the current flight control action and the flight control action at the last moment.
8. The method of claim 1, wherein determining flight control instructions for controlling the drone based on the second order smoothness constrained flight motion parameters comprises: Carrying out safety constraint processing on flight action parameters after second-order smooth constraint according to preset flight constraint conditions; And determining a flight control instruction for controlling the unmanned aerial vehicle according to the flight action parameters after the safety constraint processing.
9. A method of unmanned aerial vehicle control, comprising: the parameter acquisition module is configured to acquire visual perception information and flight state parameters of the unmanned aerial vehicle; The action decision module is configured to obtain flight action parameters of the unmanned aerial vehicle at each decision moment through a trained reinforcement learning model according to the visual perception information and the flight state parameters, wherein a reward function of the reinforcement learning model is determined by the center alignment degree, the tracking stability, the safety distance constraint, the observation scale constraint and the flight stability; The control evaluation module is configured to perform flight control evaluation on the target object according to the visual perception information to obtain the flight control reliability of the target object; And the flight control module is configured to carry out second-order smoothness constraint on the flight action parameters under the condition that the flight control reliability meets the preset reliability, and determine flight control instructions for controlling the unmanned aerial vehicle according to the flight action parameters after the second-order smoothness constraint.
10. An electronic device comprising one or more processors and a memory, the memory having stored thereon a computer program which, when executed by the one or more processors, causes the electronic device to perform the steps of the drone control method of any one of claims 1 to 8.

Description

Unmanned aerial vehicle control method and device and electronic equipment Technical Field The application belongs to the technical field of unmanned aerial vehicles, and particularly relates to an unmanned aerial vehicle control method, an unmanned aerial vehicle control device and electronic equipment. Background At present, with the development of artificial intelligence technology, the unmanned aerial vehicle field starts to introduce a visual recognition technology to detect a target, and then the flight state of the unmanned aerial vehicle is controlled based on a target detection result, so that the autonomous flight control of the unmanned aerial vehicle is realized. However, in a scene with complex environment, the control precision requirement on the unmanned aerial vehicle is high, the flight state control in the related technology is difficult to ensure the flight stability of the unmanned aerial vehicle, and the high-precision control requirement in practical application cannot be met. Disclosure of Invention In view of the defects in the prior art, the application provides an unmanned aerial vehicle control method, an unmanned aerial vehicle control device and electronic equipment, which are used for solving the problem that in the related art, unmanned aerial vehicles cannot meet high-precision control under a complex environment scene, so that flight stability is poor. In a first aspect, the present application provides a method for controlling an unmanned aerial vehicle, including: acquiring visual perception information and flight state parameters of the unmanned aerial vehicle; Obtaining flight action parameters of the unmanned aerial vehicle at each decision moment through a trained reinforcement learning model according to visual perception information and flight state parameters, wherein a reward function of the reinforcement learning model is determined by center alignment degree, tracking stability, safety distance constraint, observation scale constraint and flight stability; Performing flight control evaluation on the target object according to the visual perception information to obtain the flight control reliability of the target object; Under the condition that the flight control reliability meets the preset reliability, performing second-order smooth constraint on the flight action parameters, and determining a flight control instruction for controlling the unmanned aerial vehicle according to the flight action parameters after the second-order smooth constraint. In an embodiment of the application, flight control evaluation is performed on a target object according to visual perception information to obtain flight control reliability of the target object, and the flight control evaluation is performed on the target object according to the visual perception information to obtain the flight control reliability of the target object, wherein the flight control reliability of the target object is obtained by determining a target track stability index and a target prediction deviation of the target object in a first preset time window, determining a target detection confidence mean value and an aspect ratio change rate of a target boundary frame of the target object in the first preset time window according to the visual perception information, and performing flight control evaluation on the target object according to the target track stability index, the target prediction deviation, the target detection confidence mean value and the aspect ratio change rate of the target boundary frame. In an embodiment of the application, the target track stability index and the target prediction bias of the target object in the first preset time window are determined according to the visual perception information, and the method comprises the steps of determining the overlapping degree of target boundary frames in adjacent image frames in the first preset time window according to the visual perception information, determining the target track stability index according to the overlapping degree, determining the historical track of the target object in the first preset time window according to the visual perception information, determining the actual position of the target object in the current image frame, predicting the target prediction position of the target object in the current image frame according to the historical track, and determining the target prediction bias according to the target prediction position and the actual position. In an embodiment of the application, the visual perception information comprises a target relative position and a target detection confidence coefficient, the flight state parameters comprise a flight speed, a flight attitude angle and a flight height, and the flight action parameters of the unmanned aerial vehicle at each decision moment are obtained through a trained reinforcement learning model according to the visual perception information and the flight stat