Research On MIMUs/GPS Information Processing Based On Adaboost And Cloud Model Optimization

The rapid development of MEMS technology provides a broad space for the application of MEMS inertial measurement units(MEMS IMUs,MIMUs)in low-cost inertial navigation systems.Based on the stand-alone work of MIMUs,this paper combines GPS with Kalman filter technology to carry out research on MIMUs/GPS information fusion strategy.The navigation solution method of work-alone MIMUs is studied.The application of attitude quaternion differential equation solution is analyzed according to different output formation of the gyroscope.Meanwhile,for compensating the accumulation of navigation error in pure inertial mode,the MIMUs/GPS integrated Kalman filter is designed.The navigation accuracy is improved through effective estimation of errors.Taking into account the fact that GPS is vulnerable to outside interference,an optimized Adaboost information fusion algorithm for integrated navigation system is proposed.It is necessary to seek information processing strategies for navigation systems when GPS outages.Through iterative training,the strong learner can effectively predict the Kalman filter's filter observations in the period of GPS outage.To a certain extent,the method above inhibits the rapid divergence of navigation parameter errors caused by GPS outage.During the 50 s GPS outage,the Adaboost-based BP algorithm can effectively predict the observation values of the navigation system filtering in the loosely coupled mode,and ensure that the filter has ideal navigation forecast stability and forecast accuracy without losing the real-time performance of the system.To deal with the problem that the statistical characteristics of measurement noise change with time and the surrounding environment,a normal cloud model based fuzzy adaptive filtering algorithm is proposed.In this algorithm,the consistency degree between the system theoretical residuals and actual ones is updated,and the relationship between the consistency degree and the measurement noise variance matrix is established.The coefficient of the measurement noise variance matrix is adjusted adaptively to make it more in line with the current actual situation of the system.Compared with traditional Kalman filtering method,a better filtering effect is obtained when normal cloud model uncertainty inference is in use.In addition,when the system state is mutated,in order to keep the ability which makes the filter track the state quickly,a fuzzy strong tracking filter based on the normal cloud model is proposed.The relationship between the softening factor and the actual residuals is established through the fuzzy rules,realizing the dynamic updates of softening factor.The simulation results indicate that the method above can improve the navigation accuracy and optimize the system performance in each case.Simultaneously,applying the cloud model theory to the field of navigation information processing provides a new idea for dealing with the information fusion problem of integrated navigation.