Research On Error Correction Algorithm Of Attitude Calculation Based On MEMS Sensor

Three-dimensional space attitude information is an important parameter information reflecting the state of an object in real time.Attitude measurement technology is also widely used in defense military,industrial control,and civil consumer electronics.Attitude measurement technology based on MEMS sensor due to its low cost,low power consumption and light weight,which is very suitable for attitude measurement of consumer electronics.However,due to the low accuracy of consumer-grade MEMS sensors,the attitude angle solved by the traditional attitude calculation algorithm is prone to shift,resulting in the increasing error of attitude angle calculation.In this paper,based on the comparative analysis of attitude solution algorithms such as Euler angle method,directional cosine method,and quaternion method,the steady-state drift of angle,the delay and large error during fast motion and the adaptation of the attitude angle solution Conduct research on the issue of weak regulation.The main work of this article is as follows:(1)In order to solve the problem of angle steady state drift,a fourth-order Runge-Kuta attitude correction algorithm based on ARMA model-Kalman filtering is proposed.The algorithm first establishes the ARMA model of the MEMS gyroscope drift data,on this basis,an ARMA model-Kalman filter algorithm is established to modify the MEMS gyroscope?_x,?_y,?_z data;for the quaternion method,direct digital integration is used to solve the attitude.For the problem of error accumulation,a fourth-order Runge-Kutta algorithm based on quaternion is introduced to perform posture calculation on the corrected data.The comparison analysis of the attitude angle steady-state drift information between the modified algorithm proposed in this paper and the unmodified algorithm is realized through Matlab,which shows that the algorithm proposed in this paper suppresses the steady-state drift of attitude angle.(2)In order to solve the problem of delay and large error in the attitude solution during fast movement,the attitude angle accuracy of the single gyroscope solution cannot meet the requirements.In this paper,the multi-sensor data fusion attitude solution algorithm is improved and compared.According to the complementary characteristics of the three sensors in the frequency domain,such as gyroscope,accelerometer and magnetometer,the gyro drift error is corrected by establishing a nonlinear observer,and the fourth-order Runge-Kutta algorithm based on quaternion is combined to solve the attitude,A Quaternion-based Nonlinear Complementary Filter Attitude Fusion Algorithm(Quaternion Nonlinear Complementary Filter,QNCF)is established.And implemented a quaternion-based gradient descent attitude fusion algorithm(Quaternion Gradient Descent,QGD).The data set is used to experiment the QNCF and QGD algorithms,and compared with the calculation effect of the attitude algorithm based on the single MEMS sensor.The experiment shows that the QNCF algorithm has a shorter attitude delay and a higher accuracy of attitude angle.(3)In order to solve the problem that the adjustment coefficient of the nonlinear observer in the QNCF algorithm is fixed and cannot be adjusted in real time,the adaptive adjustment ability is weak.According to the Elman neural network(Neural Networks,NN)has the function of storing the internal state so that it has the dynamic characteristics of the mapping,this paper proposes an Elman NN compensation quaternion nonlinear complementary filter attitude correction algorithm(Elman NN-QNCF).The algorithm designed Elman NN-QCNF attitude solution model and Elman NN error compensation model.Aiming at the problem of slow convergence speed and easy to fall into local minimum point in NN training,Levenberg-Marquardt(L-M)NN learning algorithm is introduced.In view of the certain error noise due to NN prediction,a one-dimensional median filter algorithm is introduced to denoise the compensated posture information.The experiment compares the attitude information solved by the Elman NN-QNCF algorithm with the QNCF and QGD algorithms.The experiment shows that the Elman NN-QNCF algorithm has a small attitude solution error,a short attitude solution delay,and a strong adaptive adjustment capability.Experimental results show that the algorithm proposed in this paper has an ideal effect to solve the problems of angular steady-state drift error in attitude calculation,long delay of attitude calculation during fast movement,low accuracy of angle calculation,and weak adaptive adjustment ability.It has certain practical value for improving the attitude measurement accuracy of low-cost MEMS sensors.