In this article we present a novel, hybrid graph spatial representation for robot navigation. This representation enables our mobile robot to build a model of its surroundings which it can then use for navigation. The models or maps that use this representation are hybrid graphs, the nodes being analogical local maps of landmark locations in the robot's environment, the arcs being the actions the robot executes to travel between the locations. This representation yields a reliable navigation tool, one which ensures that the robot can re-orient itself to recover from errors in path execution and encounters with unexpected obstacles. The LOGnet approach also meshes with human's natural approach of mapping with landmarks, instead of using angular and translational data.