Grammar-based map compression using Manhattan world priors

In this paper, we study the problem of map compression, that is compressing a given pointset map in the context of robotic mapping and localization applications. In particular, we are interested in grammar-based compression techniques that represent the input data by a context-free grammar generating only that data. A a central contribution, we present a grammar-based map compression framework, as well as an implementation of grammar rules employing the Manhattan world assumption, and then experimentally evaluate the presented techniques in terms of compression ratio as well as compression speed using radish dataset.