A Survey of Information Geometry in Iterative Receiver Design, and Its Application to Cooperative Positioning Master of Science Thesis in Communication Engineering

Belief propagation is a powerful procedure to perform inference, and exhibits near-optimal performance in iterative decoding/demapping. However, a complete analysis of the iterative process is still lacking, and convergence of the procedure has not proved in general cases. Information geometry offers a new view from differential geometry to understand this problem. In this thesis, we try to understand belief propagation as an iterative projection procedure by studying various information-geometric interpretations from the literature. We extend these insights to distributed inference, in particular to cooperative positioning. Cooperative Bayesian algorithms have outstanding performance in many network scenarios. However, they suffers from a large computational complexity when messages are represented by a grid. We propose three techniques to reduce the complexity and improve the accuracy of cooperative Bayesian positioning: geometrical pre-location, dynamic grid, and clipping. The performance of the new algorithm compares favorably with the original algorithm, with considerably complexity and network traffic reduction.