Distributed localization and tracking with coordinated and uncoordinated motion models

This paper introduces the MDL algorithm for distributed localization in a network of mobile agents in R^m, m ¿ 1. The algorithm requires the dynamic agents with unknown locations to lie in the convex hull of at least m + 1 mobile or static anchors that can track their positions over time perfectly. Under minimal assumptions on network connectivity and local triangulation at each agent, we show that the MDL algorithm leads to convergence of the estimated agent locations to the true locations under general deterministic motion models. In addition, MDL exhibits desirable tracking performance in random environments also and we explicitly characterize the steady state tracking error in such environments. A significant feature of the algorithm is its distributed nature, where each agent updates its current location estimate based on its neighbors´ estimates and current local barycentric coordinates, computed using only local information. Simulations verify the effectiveness of the approach.