LIDAR Signature Based Node Detection and Classification in Graph Topological Maps for Indoor Navigation

Topological map extraction is essential as an abstraction of the environment in many robotic navigation-related tasks. Although many algorithms have been proposed, an efficient and accurate modelling of a topological map is still challenging, especially in complex and symmetrical real environments. In order to detect relevant changes of trajectory or a robotic platform, we propose a feature extraction model based on LIDAR scans to classify the nodes of indoor structures. As a first approach we support the experiments in a preloaded metric map, which has been used as reference for locating the nodes. Experiments are conducted in a real scenario with a differential robot. Results demonstrate that our model is able to establish the graph automatically and with precision, and that it can be used as an efficient tool for patrolling.