UAV Flight Control Using Zeroing Neural Networks for Stable Power Allocation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional PID controllers for unmanned aircraft lack stability and effective power allocation, leading to suboptimal control performance.
Innovation Solution
A method utilizing real-time flight data and a multi-layer zeroing neural network to construct neurodynamic equations, which determine output control quantities for motor powers, ensuring stable flight by optimizing power allocation through a power allocation matrix.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If PID controllers are used for unmanned aircraft control, then the control design is simple and implementation is easy, but the aircraft stability and control performance are insufficient
Solution Approach 1:
The patent replaces the conventional PID control algorithm with a neural network-based control system. The neural network learns optimal control strategies through training data, substituting the fixed mathematical model of PID with an adaptive computational model that can handle the nonlinear dynamics of the aircraft, thereby improving stability while maintaining implementation feasibility through software-based control
Solution Approach 2:
The patent transforms the control approach by changing from fixed gain parameters in PID to dynamic, adaptive parameters learned by the neural network. The control outputs are no longer based on fixed proportional-integral-derivative calculations but on variable parameters adjusted through neural network inference, enabling better adaptation to varying flight conditions and improving overall aircraft stability
2Ease of manufacture
If conventional PID control algorithms are used, then the control system is easy to implement, but the power allocation scheme does not achieve desired stability
Solution Approach 1:
The patent substitutes the conventional PID power allocation algorithm with a neural network-based power distribution system. The neural network processes flight state inputs and generates optimized motor power commands, replacing the linear control law of PID with a nonlinear mapping that better captures the complex relationships between flight conditions and optimal power allocation, thereby achieving desired stability
Solution Approach 2:
The patent implements preliminary action by pre-training the neural network controller offline with simulated flight data to learn optimal power allocation strategies. This pre-learning phase allows the controller to be deployed with pre-encoded knowledge of stable power distribution patterns, eliminating the need for real-time iterative adjustment and ensuring stable operation from the start of actual flight operations
Data Source
AI summary
Disclosed is a method for controlling stable flight of an unmanned aircraft, comprising the following steps: acquiring real-time flight operation data of the aircraft itself by means of an attitude sensor, a position sensor and an altitude sensor mounted to the unmanned aircraft, performing corresponding analysis on a kinematic problem of the aircraft by a processor mounted thereto, and establishing a dynamics model of the aircraft (S1); designing a controller of the unmanned aircraft according to a multi-layer zeroing neurodynamic method (S2); solving output control quantities of motors of the aircraft by the designed multi-layer zeroing neural network controller using the acquired real-time operation data of the aircraft and target attitude data (S3); and transferring solution results to a motor governor of the aircraft, and controlling powers of the motors according to a relationship between the control quantities solved by the controller and the powers of the motors of the multi-rotor unmanned aircraft, so as to control the motion of the unmanned aircraft (S4). Based on the multi-layer zeroing neurodynamic method, a correct solution to the problem can be approached rapidly, accurately and in real time, and a time-varying problem can be significantly solved.