Research On Attitude Control Of Quadrotor Based On Neural Network

The control performance of quadrotor aircraft is of great significance to flight safety.Because the controlled object is complex and vulnerable to various interferences during flight,it is difficult to establish an accurate model for a quadrotor aircraft.Therefore,how to design a reasonable and effective control method is particularly important.Neural network has good learning ability and calculation ability,can achieve better control effect,but the slow speed of software implementation limits its development and application.The FPGA hardware implementation has the advantages of fast running speed,low power consumption,and flexible use.This paper focuses on the optimization of neural network control algorithms and FPGA hardware acceleration to achieve steady-state tracking control of flight attitude.Aiming at the problem that it is difficult to establish accurate models for quadrotors,a fuzzy neural network PID controller based on particle swarm-genetic algorithm optimization is designed.In order to obtain the ideal control parameters,the fuzzy neural network adaptively adjusts the control parameters of the PID.The particle swarm-genetic algorithm was used to optimize the initial values of the parameters of the fuzzy neural network,and the gradient descent method was used to adjust the parameters.Simulation results show that the controller output has good dynamic performance,small steady-state error,short adjustment time,and strong anti-interference ability.Aiming at the problem that the neural network serial implementation is slow and cannot meet the real-time control of complex systems,a BP neural network PID controller implemented in parallel with FPGA is designed.In order to facilitate design and save resources,a combination of the Cordic IP core and the division IP core is used to implement a non-linear activation function,and 16-bit fixed-point numbers are used to complete data operations.The control simulation and comprehensive optimization of pitch,roll and yaw channels were completed through Vivado.Simulation results show that the design implementation significantly improves the running speed and computing performance of neural network control.In order to verify the effectiveness and feasibility of realizing PID control of BP neural network based on FPGA,the board-level verification experiment of FPGA is implemented by software and hardware collaborative design method,and the output data is transmitted to Matlab via Ethernet for analysis.Experimental results show that the design is feasible in practical control applications.Compared with serial implementation,FPGA achieves parallel operation with fast operation speed and low power consumption.The response speed of the control system is significantly improved,which meets the real-time control requirements of flight attitude.