Exploiting imprecise constraints in particle filtering based target tracking

Abstract

Nonlinear target tracking is a well-known problem for which Bayes optimal solutions based on particle filtering techniques have been developed previously. Traditionally, target tracking is solely based on sensor measurements. The tracking performance can be improved by combining sensor measurements with external information from outside the sensor. An important issue when dealing with external information in practice is that generally it is not perfect. In this paper, we discuss different classes of imperfections that can occur and we focus on imprecision, especially in relation to hard inequality constraints. A generic method is presented for dealing with imprecise hard inequality constraints, i.e. hard inequality constraints that are based on imprecise information, following a model-based approach. Simulations have been performed to demonstrate the method for exploiting imprecise hard inequality constraints in a maritime target tracking example.