In robotics, GraphSLAM is a Simultaneous localization and mapping algorithm which uses sparse information matrices produced by generating a graph of observation interdependencies (two observations are related if they contain data about the same landmark).[1]