This thesis examines the relationship between a distributed sensor information fusion system and the underlying information transport system for systems with large numbers of simple sensors. Practical methods for the prediction of system performance under changes to system parameters such as capacity, number of sensors, system state function, system and sensor noise are developed. A method for transmitting information more efficiently than measurements but convergent despite message loss is proposed. A method of co-operative transmission is developed such that a set of sensors can shape their RF transmission to allow the fusion process to take place within their co-operative interference allowing better performance than optimal joint encoding with less power and bandwidth on a shared channel. These proposals are demonstrated with the simulation of two reference designs, an HF transmission and EHF satellite, fusing multiple sensor observations.