A robust and efficient Particle Filter for target tracking with spatial constraints

This paper addresses the problem of including the non-standard information given by hard constraints in a particle filter for target tracking. The methods previously available for including this information work well in most cases. However, there are situations when the performance of these methods can deteriorate. To this end, a new method, built on proposal distributions adapted to the constraints, is developed. Moreover, the derivation of the Rao-Blackwellized Particle Filter is extended to the case of hard constraints. Both these techniques are combined and demonstrated by two illustrative simulations, showing the potential of the developed methods to handle spatial constraints given by road and coastline information.