Multihypothesis Gaussian Belief Propagation for Radio Ranging-Based Localization and Mapping

This paper provides a multi-hypothesis Gaussian belief propagation (MGBP) for radio ranging-based localization and mapping systems. To overcome the problem of lacking bearing information, MGBP is implemented via a new multi-hypothesis modeled message passing process which can fully represent the beacon uncertainties in a factor graph. In addition, MGBP does not need an external estimator to initialize the estimated states, which shows a more concise computational framework when compared with the existing graph optimization methods. In the simulation and field trial dataset based on the Ultra-wide Bandwidth (UWB) ranging, the proposed MGBP shows a better performance than many other existing RO-SLAM methods. MGBP framework can be further adapted to other problems that need particle approximations or reasonable initial values, such as the navigation problems with strong non-linearity or high state uncertainty.