This paper has made a comprehensive investigation of the contribution of inertial measurement unit (IMU) signal denoising in terms of navigation accuracy, through theoretical analysis, simulations and real tests. Analysis shows that the integral step in the inertial navigation system (INS) algorithm is essentially equivalent to a super low-pass filter (LPF), whose filtering strength is related to the integral time of the INS. Therefore the contribution of the IMU denoising filter is almost completely overshadowed by the effect of the integral step for normal navigation cases. The theoretical analysis result was further verified by the simulations with an example of inertial angle estimation and by real tests of INS and GPS/INS systems. Results showed that the IMU signal denoising cannot bring observable improvement to INS or GPS/INS systems. This conclusion is strictly valid in the condition that the equivalent cut-off frequency of the integral step (which equals the reciprocal of the INS working alone time) is lower than the cut-off frequency of the denoising filter, which is the usual case for INS applications (except for some static data processing such as the stationary alignment of INS).

Recent navigation systems integrating GPS with Micro-Electro-Mechanical Systems (MEMS) Inertial Measuring Units (IMUs) have shown promising results for several applications based on low-cost devices such as vehicular and personal navigation. However, as a trend in the navigation market, some applications require further reductions in size and cost. To meet such requirements, a MEMS full IMU configuration (three gyros and three accelerometers) may be simplified. In this context, different partial IMU configurations such as one gyro plus three accelerometers or one gyro plus two accelerometers could be investigated. The main challenge in this case is to develop a specific navigation algorithm for each configuration since this is a time-consuming and costly task. In this paper, a universal approach for processing any MEMS sensor configuration for land vehicular navigation is introduced. The proposed method is based on the assumption that the omitted sensors provide relatively less navigation information and hence, their output can be replaced by pseudo constant signals plus noise. Using standard IMU/GPS navigation algorithms, signals from existing sensors and pseudo signals for the omitted sensors are processed as a full IMU. The proposed approach is tested using land-vehicle MEMS/GPS data and implemented with different sensor configurations. Compared to the full IMU case, the results indicate the differences are within the expected levels and that the accuracy obtained meets the requirements of several land-vehicle applications.