Qualitative relational maps, which represent space as a set of coarse, relative relationships between landmarks are useful abstractions of space for robots. While qualitative relational maps have be used in several robotic scenarios, the methods for qualitative relational mapping to date are not always practical for modern robotic systems because they cannot incorporate probabilistic informa- tion, cannot adapt qualitative spatial representations to low-level data from an environment, and do not have methods for navigation that generalize to the set of all qualitative relational maps. Three approaches that make qualitative relational maps practical for modern robots are presented: 1) the probabilistic qualitative relational mapping algorithm, which adapts traditional qualitative relational maps for the incorporation of uncertain information, 2) an algorithm for adapting a qualitative spatial relationship to best represent underlying data at desired resolution, and 3) a general method for navigating with any type of qualitative relational information. Each method is tested in simulation studies and in vivo experiments to determine performance in a variety of scenarios with different data.