A flexible exploration framework for map building

Abstract

Presents a framework for exploration and incremental mapping of unknown environments. The framework allows for evaluation and comparison of different acquisition strategies. During exploration a visibility graph is constructed which holds correct topology information about the environment and provides a means for immediate planning in the partially known map. The framework has been implemented in simulation and on a real platform equipped with a 360 degree laser scanner, an algorithm for line and segment extraction and an extended Kalman filter for localization. Structured environments have been explored and mapped in a fully autonomous mode, simultaneously localizing the robot yielding results of satisfying precision. Limitations and problems of our implementation are discussed as well.