Algorithm Implementation Of IMU Zero Offset Suppression And Dynamic Zero Compensation Based On DSP/FPGA

The inertial measurement unit is a device that uses inertial sensors to measure and sense the motion of the carrier.It is one of the core components in the fields of defense,aerospace and other fields.With the development of MEMS technology,miniaturization and low power consumption have become the development direction of inertial sensors,which makes it possible to realize the development of portable devices based on inertial measurement units.The performance of the inertial measurement unit depends on the final attitude angle resolution accuracy.The factors affecting its accuracy are: 1.The update frequency of the attitude angle;2.The random error common to inertial sensors,the most important of which is the zero drift error.Therefore,in order to improve the performance of the Inertial Measurement Unit(IMU),it is necessary to design an embedded hardware platform with fast processing capability,and perform effective zero offset suppression and zero compensation on the platform according to the zero error characteristics.This paper first demonstrates the advantages of using a Digital Signal Processor(DSP)and a Field-Programmable Gate Array(FPGA)as the inertial sensor measurement system processor for DSP,FPGA,inertial sensor,and serial port.Chips and other major chip selection,using Altium Designer to draw printed circuit board(PCB)layout according to the package signal of each chip.The data communication architecture between FPGA and DSP is constructed.The functions of SPI,FIFO and serial port required by this topic are realized in FPGA.The resource modules required by DSP are configured.The inertia is collected through hardware in-circuit debugging and serial port debugging.Secondly,according to the zero error characteristics of inertial sensors,a zero offset suppression algorithm model is proposed.In this model,the original signal is wavelet decomposed.This process transforms the time series information analysis problem into frequency domain analysis problem.According to the characteristics of the gyro signal,the soft threshold processing is used to filter out the high frequency information.After the wavelet reconstruction,the coarse filter function is obtained.Then,the least squares fitting of the coarse filtered data is performed to obtain the ideal zero offset suppression data.The data obtained from the high and low normal temperature experiments show that the three axial gyro zero data are improved by about 40% compared with the original data after the zero offset suppression algorithm proposed in this paper.Finally,since the zero position of the sensor after zero-bias suppression will change nonlinearly with the change of ambient temperature and running time,in order to achieve higher precision attitude calculation,dynamic zero compensation algorithm needs to be designed.In the fifth chapter,the characteristics of the LSTM neural network model are introduced.The gyro dynamic zero compensation algorithm based on the LSTM model is established.The zero-bit data collected at different operating temperatures and different running times are used as learning quantities to train the network model.The function body solving process is derived.Through the variable temperature with or without zero compensation algorithm,the calculation of the heading angle error comparison experiment is carried out.The obtained data shows that after the zero compensation algorithm is solved,the heading angle accuracy can be improved by 9.4%,and the heading angle stability can be increased by an order of magnitude;Through the normal temperature three-axis rotary table motion with or without zero compensation algorithm,the solution attitude angle error comparison experiment is obtained.The data shows that the solution error of the three attitude angles is reduced from the average drift from 7.65° to 1.39° per hour after the zero compensation algorithm is solved.