High resolution wavelet de-noising for MEMS-based navigation systems

Vehicle navigation systems often employ an inertial measuring unit (IMU) to complement the Global Positioning System (GPS) in the event of satellite signal loss due to blockage or jamming. The added redundancy of the inertial navigation system (INS) can be invaluable to the end user; and as such, the integration of GPS with inertial sensors has become a standard practice. The relatively high cost of INS has been preventing their use for land vehicle applications. Recently, MEMS-based INS has become commercially available at low cost. These relatively low cost inertial sensors have the potential to allow an affordable vehicle navigation system to be developed.; Compared to tactical and navigational grade INS, MEMS-based sensors are less expensive but are more susceptible to the short and long term (low and high frequency) errors that present themselves as correlated noise. Signal processing techniques such as, optimal low-pass filtering and wavelet de-noising have been successful at minimizing the short-term errors, but they have not been able to effectively eliminate the long-term errors mixed with the dynamic motion of the vehicle. Newer techniques such as the Fast Orthogonal Search (FOS) has shown some improvement over other techniques at removing long and short term inertial sensor errors.; The primary objective of this research is to use advanced signal processing techniques such as Wavelet Multi-Resolution Analysis (WMRA), Wavelet Packet Transform (WPT) and high-resolution spectral analysis using Fast Orthogonal Search (FOS) to attempt to remove the short and long-term errors of inertial sensors prior to processing them with GPS. In addition, a variation of FOS known as FOS-Wavelet Transform (FOS-WT) was developed to provide high resolution wavelet analysis. FOS vi WT makes use of exponentially decaying sinusoids as candidate functions to further improve the accuracy of the FOS model. This research will focus on examining the merits and the limitations of the above de-noising techniques when applied to a MEMS-based INS. The removal of the correlated sensor errors should result in a significant increase in accuracy of the overall vehicle position.; Four road test experiments in a land vehicle were conducted. During these tests, real GPS data and MEMS-inertial sensor measurements were collected. Analysis of the data with WMRA and WPT has shown successful removal of the short-term sensor errors but not the long term errors, and that FOS and FOS-WT have successfully removed the short term errors as well as some of the long term errors, resulting in a significant increase in the overall positioning accuracy of the land vehicle.