A Framework for efficient cooperative localization with non-Gaussian ranging error distributions

Cooperative localization takes advantage of the range measurements between neighbouring agents to improve both the availability and accuracy of positioning systems. Distributed belief propagation is a promising technique for data fusion in cooperative localization. Difficulties with belief propagation lie in reducing the required communication overhead and computational complexity. Most of the existing works on this subject are based on Gaussian ranging error models, which are fine for outdoor applications but not suitable for harsh propagation environments such as indoor and urban areas. In this paper, a framework for efficient cooperative localization is proposed, which can be applied to systems with non-Gaussian ranging error distributions. The communication and computational cost is reduced by passing approximate beliefs represented by Gaussian distributions between neighbours and by using an analytical approximation to compute peer-to-peer messages. The proposed scheme is validated experimentally on a deployed indoor localization system, and is shown to achieve high accuracy with low communication and computational cost.