Research On MINS/OD Integrated Navigation System Filter Algorithm

Automotive MEMS Inertial Measurement Unit (IMU) and odometer can constitute afully autonomous navigation system. MEMS IMU provides comprehensive navigationinformation, but the navigation results are inaccuracy and with large noise. The navigationerrors accumulate quickly with time. Odometer is a sensor which can measure the distancetravelled, and thus it can be used to restrain the divergence of the inertial navigation error.However, the output information of the odometer is limited and contains the scale factorerror. To combine their advantages, these two approaches are integrated. To solve theproblems in practice, various modified Kalman Filter is used to estimate states such asposition, velocity, attitude, and device errors.This dissertation mainly research the MEMS IMU/odometer integrated navigationsystem data fusion problems and the purpose is to establish a high precision fusionalgorithm. The main research contents include:(1) MINS/Odometer integrated model establishmentIn order to build the MINS and odometer mathematical model for integratednavigation system, IMU and odometer device errors are selected as state variables filter toestablish state equation of the filter, such as gyroscope bias, acceleration bias, attitude errorand position error, velocity error and scale factor error. The filter measurement equation isestablished by the outputs of both systems.(2) Nonholonomic constraint analysis and observability analysis of the system modelAccording to the characteristic of the motion of a wheeled vehicle, a newnonholonomic constraint method is proposed. This method enlarges the observation as wellas promotes the instantaneity of it. The integrated system is a time-varying system, becausethe system model is associated with the movements of the carrier. Piece-Wise ConstantSystem observability analysis is used to analyze the system and thus obtaining theobservation status of the errors.(3) Filter methods researchThe process noise of the MINS varies according to the temperature and differentconditions of the road affect the observation noise easily. The application of Kalman Filteris limited, because it demands that the process noise and the observation noise are knownand constant. For these problems, Multiple Model method is utilized. Several filter models estimate the states together. The best model is matched during the process of filter, thussolving the problem of varying noise efficiently. At the same time, adaptive filter isimproved. Adaptive filter can get the characteristic of noise during the estimation, so thismedthod can also adapt to the variation of different conditions. As for the problems of theinaccuracy of the model and the disturbance of odometer during the travel, the StrongTracking filter forces the residual error to be zero mean white noise. This method is helpfulto the robustness of the system.