An evaluation of nonlinear filtering algorithms for integrating GNSS and inertial measurements

The quality of dead reckoning positioning algorithms which are integrating GNSS and INS measurements is a crucial factor for advanced vehicular safety systems. Apart from sensor performances, the most important factor determining this quality is the filter algorithm. Thus, this paper aims to compare and evaluate different nonlinear filtering approaches, focusing on their performance in GNSS/INS integration. The presented approach is based on an advanced motion model which assumes a constant turn rate and longitudinal acceleration of the vehicle. Using this approach, the unscented Kalman Filter is compared to the standard extended Kalman filter. For the evaluation, experimental data obtained by a DGPS receiver with RTK capabilities are used. With this approach, the filter assessment is performed in different scenarios, including urban areas.