Design And Software Development Of Attitude Estimation System Based On Neural Network PID

Today,drones are widely used in military,commercial,civil,and scientific fields,and countries,companies and research institutions are beginning to conduct research on drones.As the control core of the drone,the flight controller is responsible for the functions of control signal reception,attitude calculation,and output control.Based on the practical application and engineering design,this paper improves the algorithm of the attitude estimation part in the flight control,and designs,develops and verifies the hardware and software.According to the embedded engineering design process,the system functional requirements are analyzed,and an embedded hardware platform based on STM32 is designed.This paper designs the interface circuit and power management of the hardware system,including the peripheral circuit of the microcontroller,the power supply step-down circuit,the USB interface circuit,the sensor and the digital interface circuit.And the issues that need attention during PCB fabrication and component soldering are discussed.Based on the embedded development environment under Linux,this paper plans the software project file structure and develops the task scheduler in the Cotex-M3 kernel.Design the data structure in the scheduler,write the relevant scheduling function,study the operation principle of the scheduler,and analyze the impact of the interrupt on the scheduling,and realize the context switching function under the ARM system.Based on the scheduler,this paper studies the attitude representation and update method in strapdown inertial navigation,analyzes the advantages and disadvantages of various sensors and common data fusion algorithms,and gives a complementary filter fusion algorithm based on neuron PID and implement it as a process in the software.The data transmission protocol is implemented,and the attitude angle is uploaded to the upper computer by using wireless communication,and the stability,anti-interference and precision are analyzed.Finally,the software size and real-time performance were tested to verify the software performance.In general,this paper studies the flight control attitude estimation system,and realizes a miniaturized,low-cost,real-time attitude solution software and hardware system using neuron PID complementary filter.