Dempster Shafer neural network algorithm for land vehicle navigation application

Abstract

A Global Positioning System (GPS)–aided Inertial Navigation System (INS) provides a continuous navigation solution with reduced uncertainty and ambiguity. Bayesian approaches like Extended Kalman filter or Particle filter are generally developed for fusing the GPS and INS data. However, these techniques require prior distribution (representing the degree of belief) to be accurately defined for all incorporated parameters−whether known or unknown. If no previous knowledge is obtainable, equal probabilities are assigned to all events, which is questionable. To overcome these limitations, Dempster Shafer (DS) evidence theory is implemented in this paper. In order to effectively fuse GPS and INS data for land vehicle navigation application, we propose an efficient Dempster Shafer Neural Network (DSNN) algorithm by integrating the Dempster Shafer theory and the artificial neural network. Our field test results clearly indicate that the proposed DSNN algorithm effectively compensated and reduced positional inaccuracies during no GPS outage and GPS outage conditions for low cost inertial sensors.