CN-122018298-A - Unmanned aerial vehicle intelligent control method and system based on machine learning

CN122018298ACN 122018298 ACN122018298 ACN 122018298ACN-122018298-A

Abstract

The invention discloses a machine learning-based unmanned aerial vehicle intelligent control method and a machine learning-based unmanned aerial vehicle intelligent control system, which relate to the technical field of unmanned aerial vehicle control and comprise the following steps: judging whether the unmanned aerial vehicle enters a complex flight mode according to flight environment parameters, switching corresponding intelligent control modes according to different flight states of the unmanned aerial vehicle in the complex flight stage, wherein the intelligent control modes comprise an initial adaptation stage, a dynamic adjustment stage and a stable optimization stage, and respectively executing different control strategies to realize accurate control of the unmanned aerial vehicle. The machine learning model is used for deep mining and analysis of flight data, the open loop control is combined with the quick response of environmental change, the closed loop control is used for dynamically correcting control parameters, the problems of poor control timeliness, low precision and weak anti-interference capability of the traditional unmanned aerial vehicle in a complex environment are solved, the flight stability and task execution reliability of the unmanned aerial vehicle in complex terrain and severe weather conditions are improved, and the flight deviation and safety risk are reduced.

Inventors

LI GUIPENG
ZHU LIN
Cheng Wanyi
CHEN XIANGBAO
WANG XIANGWEI
ZHANG KAI
LIU ZIZHANG
ZHANG GANGQIANG
ZHENG ZHIZHONG
WANG XINYU
DONG XIAOXUAN

Assignees

华能新能源股份有限公司河北分公司

Dates

Publication Date: 20260512
Application Date: 20260225

Claims (10)

1. The intelligent control method of the unmanned aerial vehicle based on machine learning is characterized by comprising the following steps of: Judging whether a user inputs a flight task instruction of the unmanned aerial vehicle, if the user inputs the flight task instruction, acquiring flight environment parameters, judging whether the unmanned aerial vehicle is in a complex flight scene, and if the unmanned aerial vehicle is in the complex flight scene, controlling the unmanned aerial vehicle to enter an initial adaptation stage of the complex flight mode; Judging whether the stability of the flight attitude of the unmanned aerial vehicle is greater than a preset stability threshold, and if the stability of the flight attitude is greater than the preset stability threshold, controlling the unmanned aerial vehicle to enter a dynamic adjustment stage of a complex flight mode; judging whether the operation time length of the dynamic adjustment stage is longer than a preset adjustment time length, and if the operation time length of the dynamic adjustment stage is longer than the preset adjustment time length, controlling the unmanned aerial vehicle to enter a stable optimization stage of a complex flight mode; judging whether the task execution deviation of the unmanned aerial vehicle is smaller than a preset deviation threshold value, and if the task execution deviation is smaller than the preset deviation threshold value, controlling the unmanned aerial vehicle to exit from a complex flight mode and enter into a conventional flight control mode.
2. The machine learning based unmanned aerial vehicle intelligent control method of claim 1, wherein the initial adaptation phase is configured to execute a first control strategy comprising performing open loop control of unmanned aerial vehicle flight parameters based on a pre-trained environmental adaptation model; The dynamic adjustment stage is used for executing a second control strategy, and the second control strategy comprises executing semi-closed loop control of flight parameters based on real-time flight data and dynamic output of a machine learning model; The stability optimization stage is configured to execute a third control strategy that includes closed-loop optimization control of flight parameters based on a reinforcement learning algorithm.
3. The intelligent control method of the unmanned aerial vehicle based on machine learning according to claim 2, wherein the open loop control of the flight parameters comprises controlling the power actuator, the attitude actuator and the navigation actuator of the unmanned aerial vehicle by a flight control unit to control the unmanned aerial vehicle to operate with a pre-generated initial set of flight parameters.
4. A machine learning based unmanned aerial vehicle intelligent control method according to claim 3, wherein the initial set of flight parameters comprises a flight speed reference value, an attitude angle reference value, a lift reference value, and a navigation path reference value.
5. The intelligent control method of the unmanned aerial vehicle based on machine learning according to claim 4, wherein the initial flight parameter set is obtained by calculation of an environment adaptation model by a flight control unit, the environment adaptation model is a neural network model trained based on a deep learning algorithm, input parameters of the neural network model comprise wind speed, wind direction, air pressure, visibility, terrain gradient and obstacle density, and output parameters of the neural network model are corresponding initial flight parameter sets.
6. The machine learning based unmanned aerial vehicle intelligent control method of claim 5, wherein the training data of the environment adaptation model comprises flight sample data under different complex environments, and the flight sample data comprises environmental parameters, flight parameters and flight effect evaluation values.
7. The intelligent control method of the unmanned aerial vehicle based on machine learning according to claim 2, wherein the semi-closed loop control of the flight parameters comprises the steps of collecting flight state data of the unmanned aerial vehicle in real time through a flight control unit, inputting the flight state data into a dynamic adjustment model, outputting a flight parameter correction amount based on a gradient descent algorithm by the dynamic adjustment model, and dynamically correcting an initial flight parameter set.
8. The machine learning based unmanned aerial vehicle intelligent control method of claim 7, wherein the flight status data comprises an actual flight speed, an actual attitude angle, an actual position coordinate, a battery remaining power, and a task execution progress.
9. The intelligent control method for the unmanned aerial vehicle based on machine learning according to claim 2, wherein the closed-loop optimization control of the flight parameters comprises the steps of constructing a reinforcement learning agent through a flight control unit, optimizing the flight parameters in real time based on a Markov decision process by maximizing task execution precision and minimizing energy consumption, wherein a state space of the reinforcement learning agent comprises an environment dynamic change parameter, an unmanned aerial vehicle flight state parameter and a task state parameter, and an action space comprises a flight parameter adjustment quantity set.
10. Unmanned aerial vehicle intelligent control system based on machine learning, characterized by comprising: The initial adaptation module is used for judging whether a user inputs a flight task instruction of the unmanned aerial vehicle, acquiring flight environment parameters and judging whether the unmanned aerial vehicle is in a complex flight scene or not if the user inputs the flight task instruction, and controlling the unmanned aerial vehicle to enter an initial adaptation stage of the complex flight mode if the unmanned aerial vehicle is in the complex flight scene; the dynamic adjustment module is used for judging whether the stability of the flight attitude of the unmanned aerial vehicle is greater than a preset stability threshold value, and if the stability of the flight attitude is greater than the preset stability threshold value, controlling the unmanned aerial vehicle to enter a dynamic adjustment stage of a complex flight mode; the stability optimization module is used for judging whether the operation time length of the dynamic adjustment stage is longer than the preset adjustment time length, and if the operation time length of the dynamic adjustment stage is longer than the preset adjustment time length, controlling the unmanned aerial vehicle to enter a stability optimization stage of the complex flight mode; And the conventional flight module is used for judging whether the task execution deviation of the unmanned aerial vehicle is smaller than a preset deviation threshold value, and if the task execution deviation is smaller than the preset deviation threshold value, controlling the unmanned aerial vehicle to exit the complex flight mode and enter the conventional flight control mode.

Description

Unmanned aerial vehicle intelligent control method and system based on machine learning Technical Field The invention relates to the technical field of unmanned aerial vehicle control, in particular to an unmanned aerial vehicle intelligent control method and system based on machine learning. Background Along with the rapid development of unmanned aerial vehicle technology, the application scene of the unmanned aerial vehicle has been expanded from simple aerial mapping to the fields of material transportation, electric power inspection, emergency rescue and the like under complex environments. In complex flight scenes, such as mountainous terrain, strong wind environments, low-visibility conditions and the like, unmanned aerial vehicles face challenges of rapid environmental dynamic change, multiple interference factors and high flight precision requirements. The existing unmanned aerial vehicle control method mostly adopts a traditional PID control algorithm, and flight control is realized through preset fixed parameters or simple parameter self-tuning strategies. However, the traditional control method has the defects that firstly, the adaptability to complex dynamic environments is poor, fixed control parameters are difficult to cope with sudden situations such as sudden changes of wind speed, fluctuation of topography and the like, the flying gesture is easy to be unstable, secondly, the control precision is limited, accurate parameter adjustment cannot be carried out based on dynamic changes of environments and tasks, the task execution deviation is large, thirdly, the autonomous learning capability is lacking, the control strategy cannot be optimized through historical flying data, the control effect depends on initial parameter setting, and the anti-interference capability is weak. In addition, although a simple intelligent algorithm is introduced into part of unmanned aerial vehicles in the prior art, the problems of poor timeliness of data processing and insufficient generalization capability of models exist, real-time accurate control under a complex scene is difficult to realize, and the application of the unmanned aerial vehicle in a high-requirement scene is limited. Therefore, how to provide a machine learning-based unmanned aerial vehicle intelligent control method and system, which overcomes the defects existing in the prior art is a problem that needs to be solved by the skilled person. Disclosure of Invention In view of the above, the invention provides a machine learning-based unmanned aerial vehicle intelligent control method and system, which realize accurate and stable control of the unmanned aerial vehicle in a complex environment by deep fusion of a staged intelligent control strategy and a machine learning algorithm. The method solves the problems of low control precision, poor adaptability, weak anti-interference capability and the like in the prior art. In order to achieve the above purpose, the present invention adopts the following technical scheme: a robot intelligent control method based on machine learning comprises the following steps: Judging whether a user inputs a flight task instruction of the unmanned aerial vehicle, if the user inputs the flight task instruction, acquiring flight environment parameters, judging whether the unmanned aerial vehicle is in a complex flight scene, and if the unmanned aerial vehicle is in the complex flight scene, controlling the unmanned aerial vehicle to enter an initial adaptation stage of the complex flight mode; Judging whether the stability of the flight attitude of the unmanned aerial vehicle is greater than a preset stability threshold, and if the stability of the flight attitude is greater than the preset stability threshold, controlling the unmanned aerial vehicle to enter a dynamic adjustment stage of a complex flight mode; judging whether the operation time length of the dynamic adjustment stage is longer than a preset adjustment time length, and if the operation time length of the dynamic adjustment stage is longer than the preset adjustment time length, controlling the unmanned aerial vehicle to enter a stable optimization stage of a complex flight mode; judging whether the task execution deviation of the unmanned aerial vehicle is smaller than a preset deviation threshold value, and if the task execution deviation is smaller than the preset deviation threshold value, controlling the unmanned aerial vehicle to exit from a complex flight mode and enter into a conventional flight control mode. Optionally, the initial adaptation stage is configured to execute a first control strategy, where the first control strategy includes performing open loop control of unmanned aerial vehicle flight parameters based on a pre-trained environmental adaptation model; The dynamic adjustment stage is used for executing a second control strategy, and the second control strategy comprises executing semi-closed loop control of flight parameters base