A new variant of Nonparametric Belief Propagation for self-localization

We consider the problem of relative self-localization of a network of fixed communicating devices that evaluate range measurements between each other. The solution is obtained in two stages: First, a new variant of the Nonparametric Belief Propagation algorithm is used for estimating the beliefs. This variant is based on a Monte-Carlo integration with rejection sampling where a delimited space region is determined for each node in order to reduce the rejection ratio. Then, a new algorithm based on estimation in discrete states space is proposed for solving the flipping ambiguities resulting from the lack of measurements. This solution has the advantage of reducing the amount of communicating particles and the computation cost.