Particle filter for target localization and tracking leveraging lack of measurement
Bethany L. Allik · 2017
In this paper a sequential Monte Carlo approach is used to track targets using multiple agents, where lack of measurements by individual agents aide in the estimation procedure. The proposed agents are equipped with cameras or sensors, where the fields of view or dynamic ranges are limited in the measurement model. In current approaches, measurements that are not available due to saturation, occlusion, truncation, etc., provide no net positive in the information gain of the system. However, therein lies a fundamental problem in which these approaches are not using all of the information available. The argument is made that the lack of measurement in this scenario will provide useful information to the state estimator, of which, a particle filter will be employed. Raw measurements will be sent to a fusion center to be processed, where the full likelihood function will be calculated. Simulations of the algorithm will be presented, which show the usefulness of the algorithm in the application of target tracking over a large 2D area.