The Arithmetic Of Attitude Computation And Data Fusion Based On Mems Sensors

In indoor positioning system,direction is always a Research focus.And in Pedestrian Dead Reckoning based on MEMS sensors,traditional direction judgment by magnetometer is difficult to meet the requirements.The direction is related to the attitude of the equipment,so calculating reliable attitude based on MEMS sensors is the key point and is also the research target of this thesis.This thesis aims at designing a micro attitude measurement system by MEMS sensors to output Real-time attitude information.This system is not only used in the field of indoor positioning system,but also various kinds of model airplanes,wearable devices and so on.This thesis introduces and analyses the methods of calculating attitude by MEMS sensors and defects in them,so data fusion algorithms is necessary.Then it researches and analyzes the theory and algorithm model of Unscented Kalman Filter.With the system model of attitude calculating based on MEMS sensors,this thesis analyzes some problems including huge calculated amount,inherent model deviation and no-positive definite matrix's appearance in the Unscented Kalman Filter algorithm.To solve these problems,it comes up with several optimization schemes including reducing the steps of Unscented Kalman Filter model,bringing in adaptive factor to anti interference rapidly and a method to make sure the covariance matrix is a positive definite matrix.At last,this thesis has designed the algorithm flow chart and system scheme according to the optimized Unscented Kalman Filter.This thesis has designed a two-channel attitude measurement system with a microcontroller(CC2640)and a nine-in-one MEMS sensor(BNO055).Connecting with SEC380(a commercial attitude measurement system),the system is tested in static state,dynamic state,complex interference environment and Pedestrian Dead Reckoning application.Test results show that the deviation of this attitude calculation and dates fusion algorithm is within ±0.3 in static state,±2° of heading angle and ±1.5° of pitch angle and roll angle in dynamic state and the algorithm works well in Pedestrian Dead Reckoning application.It meets the design target of this thesis.