Adaptive Particle Filter for INS/GPS Integration

Navigation comprises of the integration of methodologies and systems for estimating the time varying position, velocity and attitude of moving objects. In this paper, an Adaptive Particle Filter (APF), which enables precise characterization of posterior distribution and deals efficiently with non-linear and non-Gaussian models, has been developed. Presently, integrated vehicle navigation systems are implemented by using Kalman Filter, Extended Kalman Filter, and Unscented Kalman Filter. These filters approximate the process and/or measurement models by linear or Gaussian fits. However, these assumptions of fairly linear movement with Gaussian distribution, is unrealistic for highly nonlinear systems such as automotive or airborne integrated navigation systems and target tracking in aircraft or cars. This paper proposes the use of adaptive particle filter to approximate the posterior distribution in these highly nonlinear integrated navigation systems. The objective of this paper is to integrate data from different sensors to accurately estimate the navigation states of a dynamic system. A kinematic experimental GPS/INS dataset including dual frequency carrier phase GPS receiver data and inertial measurements from MEMS grade inertial sensors installed on same vehicle will be used to validate the proposed filtering techniques. The performance of the particle filter will then be compared with the performance of current estimation techniques like Extended Kalman Filter for the same datasets in terms of position errors.