Monte Carlo Initialization for Multi-Sensor Bearing Only Tracking

Abstract

A new algorithm for particle filter initialization for multi-sensor bearing only tracking is developed to enhance tracker performance and stability. Multiple bearing observations are used by a least squares technique to form multiple initial position estimates; these estimates are in turn used to compute the statistics of the initial state distribution. Simulated data is used to demonstrate the performance and efficiency of the algorithm by comparing the new initialization technique to a filter initialized with the true initial state.